Theoretical and Applied Genetics

, Volume 127, Issue 3, pp 535–547

Genetic analyses of bolting in bulb onion (Allium cepa L.)

Authors

  • Samantha Baldwin
    • New Zealand Institute for Plant and Food Research
  • Roopashree Revanna
    • New Zealand Institute for Plant and Food Research
  • Meeghan Pither-Joyce
    • New Zealand Institute for Plant and Food Research
  • Martin Shaw
    • New Zealand Institute for Plant and Food Research
  • Kathryn Wright
    • New Zealand Institute for Plant and Food Research
  • Susan Thomson
    • New Zealand Institute for Plant and Food Research
  • Leire Moya
    • New Zealand Institute for Plant and Food Research
  • Robyn Lee
    • Biochemistry DepartmentUniversity of Otago
  • Richard Macknight
    • Biochemistry DepartmentUniversity of Otago
    • New Zealand Institute for Plant and Food Research
Original Paper

DOI: 10.1007/s00122-013-2232-4

Cite this article as:
Baldwin, S., Revanna, R., Pither-Joyce, M. et al. Theor Appl Genet (2014) 127: 535. doi:10.1007/s00122-013-2232-4

Abstract

Key message

We present the first evidence for a QTL conditioning an adaptive trait in bulb onion, and the first linkage and population genetics analyses of candidate genes involved in photoperiod and vernalization physiology.

Abstract

Economic production of bulb onion (Allium cepa L.) requires adaptation to photoperiod and temperature such that a bulb is formed in the first year and a flowering umbel in the second. ‘Bolting’, or premature flowering before bulb maturation, is an undesirable trait strongly selected against by breeders during adaptation of germplasm. To identify genome regions associated with adaptive traits we conducted linkage mapping and population genetic analyses of candidate genes, and QTL analysis of bolting using a low-density linkage map. We performed tagged amplicon sequencing of ten candidate genes, including the FT-like gene family, in eight diverse populations to identify polymorphisms and seek evidence of differentiation. Low nucleotide diversity and negative estimates of Tajima’s D were observed for most genes, consistent with purifying selection. Significant population differentiation was observed only in AcFT2 and AcSOC1. Selective genotyping in a large ‘Nasik Red × CUDH2150’ F2 family revealed genome regions on chromosomes 1, 3 and 6 associated (LOD > 3) with bolting. Validation genotyping of two F2 families grown in two environments confirmed that a QTL on chromosome 1, which we designate AcBlt1, consistently conditions bolting susceptibility in this cross. The chromosome 3 region, which coincides with a functionally characterised acid invertase, was not associated with bolting in other environments, but showed significant association with bulb sucrose content in this and other mapping pedigrees. These putative QTL and candidate genes were placed on the onion map, enabling future comparative studies of adaptive traits.

Abbreviations

SD

Short-day

LD

Long-day

DH

Doubled-haploid

CAPS

Cleaved amplified polymorphic sequence

AMAL

Alien monosomic addition line

Introduction

Onion and shallot (Allium cepa L.) are staple functional foods central to most diets, grown from tropical regions to high latitudes. Economic production of dry bulb onion and shallot requires cultivars to be biennial. They must be adapted to local photoperiod and temperature such that a sufficiently large bulb is formed in the growing season without premature transition to flowering (‘bolting’; Kamenetsky and Rabinowitch 2002). Bulb onions are broadly classified into ‘short-day’ (SD) and ‘long-day’ (LD) types depending on the length of day required to initiate bulb formation. In terms of flowering adaptation, all onions produce flowers in the spring. True shallot of species A. cepa is a horticultural type displaying a partial to complete tendency toward asexual reproduction. It is presumed that the shallot phenotype has been selected for practicality of production in tropical regions where florigenesis is compromised by want of vernalization and fungal disease pressure on floral structures (Rabinowitch and Kamenetsky 2002).

There is an extensive literature concerning the physiology of flowering in onion, reviewed most recently by Brewster (2008). Once onion plants reach a critical weight or leaf number, they can be induced to initiate inflorescences by exposure to low temperatures. The critical size and temperature requirements for this vernalization exhibit wide genetic variation. Depending on the conditions and genetic background, there may be ‘competition’ between bulb and inflorescence development (Van Kampen 1970), such that inflorescence development may be suppressed. It has been demonstrated that a vernalization period is all that is required to initiate flowering from onion bulbs, but that temperature and photoperiod affect the time taken for eventual inflorescence appearance and floret opening (Khokhar et al. 2007). De-vernalization in onion is commercially exploited by high-temperature (>25 °C) treatment of cold-stored onion ‘sets’ prior to transplanting in springtime.

The stringent requirement for adaptation of onion cultivars to environment for producing marketable crop places major constraints on breeding and exploitation of germplasm. A deeper understanding of the genetics of temperature and photoperiod adaptation could inform strategies for genetic resource exploitation and development, and provide insights into domestication and dissemination of onion over the past millennia. Most importantly, functional markers to major genes conditioning adaptation could enable more efficient introgression of desirable traits such as disease resistance from wide crosses, as well as enabling seed lot quality control.

Physiological and genetic studies have implicated numerous genes and pathways in control of flowering in Arabidopsis and other species (Matsoukas et al. 2012). Components of the photoperiodic floral induction pathway are conserved between dicots and monocots (e.g. cereals such as rice and barley): for example, the circadian clock is needed to measure day length and downstream components such as Constans (CO; known as Heading date 1/HD1 in rice) play important roles in the photoperiodic flowering of Arabidopsis and rice (Tsuji et al. 2011). The vernalization pathway is less well-conserved and is believed to have evolved separately in grasses and dicots (Greenup et al. 2009). The photoperiodic and vernalization pathways converge on a small number of floral pathway integrators in all plants, including Suppressor of overexpression of constans 1 (SOC1) (Hepworth et al. 2002), and Flowering locus T (FT) genes (HD3a and RiceFT-LIKE1/RFT1 in rice and FT1 in wheat and barley).

Although the extensive genetic and physiological studies of grasses are often assumed to generalise to all monocots, studies of adaptive trait genetics in other monocot clades are very limited. Onion is in the order Asparagales, which diverged 122 m years ago from the monocot grasses (Janssen and Bremer 2004). Therefore, the genetic mechanisms controlling cereal flowering may differ in the Asparagales. Although physiological and genomic studies of flowering are very limited in Asparagales, several recent studies have described transcriptional and functional studies in orchid genera (Chang et al. 2011; Ding et al. 2013; Liang et al. 2012; Xiang et al. 2012). In onion, a period of vernalization was found to be all that was required to initiate flowering from onion bulbs, but that temperature and photoperiod affected the time taken for eventual inflorescence appearance and floret opening (Khokhar et al. 2007). Taylor et al. (2010) hypothesised that components of the photoperiodic pathway might be involved in the induction of onion bulbing. While they found that circadian clock genes appear to have a conserved function based on their expression, whether they are involved in onion bulbing was not established. We recently reported functional characterisation of the FT-like gene family in onion (Lee et al. 2013), identifying three members with distinctive expression patterns associated with bulbing and floral initiation. This work indicated that two antagonistically acting FT-like genes (AcFT1 and AcFT4) are involved in the photoperiodic induction of onion bulbing, while another FT (AcFT2) is involved in the vernalization responsive initiation of flowering (Lee et al. 2013).

Genetic analysis in numerous species has frequently revealed associations between allelic variation in conserved components of these pathways and integrators with flowering time. Natural variation in FT-like genes has been associated with flowering time in dicots such as sunflower and Arabidopsis (Blackman et al. 2010; Laurie et al. 2011), and monocot grasses such as ryegrass (Skøt et al. 2011) and rice (Kojima et al. 2002). Similar associations have been reported for components of the well-conserved photoperiodic pathway such as CO/HD1 in rice (Takahashi et al. 2009).

In this study, we have taken a genetic approach as the first step in understanding the genetic architecture of premature bolting (flowering) in onion. The only reported genetic analysis of onion bolting is that of Hyun et al. (2009) who examined an F2 family segregating for time to flowering using proteomic methods. This revealed that in the family examined, late flowering segregated as a dominant character, and although they identified proteomic differences among the parents they did not examine segregation of these with bolting phenotype. Physiological and genomic studies in onion have been complicated by a lack of freely available reference lines, combined with very limited genomics resources. Recent transcriptome sequencing and mapping studies based on homozygous doubled haploid (DH) lines and RNASEQ have greatly expanded scope for reproducible and comparative approaches (Baldwin et al. 2012b; Duangjit et al. 2013) as well as enabling global searches for candidate genes. We recently functionally characterised members of the FT-like gene family in onion (Lee et al. 2013). In the current study, we sought to place these and other candidate genes on the onion genetic map and to search for QTL conditioning bolting in a wide onion cross (Baldwin et al. 2012b). This revealed that a QTL located on onion chromosome one plays a major role in conditioning bolting in this cross.

Materials and methods

Plant materials

DH onion lines were provided by Cornell University (Alan et al. 2004). Diverse onion genotypes from SD and LD origins were obtained from the EU Allium Gene Bank at Warwick University (http://ealldb.ipk-gatersleben.de/; Table 1). Seed lots were sown in spring either directly as seed or as transplants in Jiffy 7 plugs (Jiffy Corp, Lorain, OH, USA). Crop management followed standard commercial practices.
Table 1

Onion populations used in this study

Population barcode number

NCBI BioSample accession

Label

Description

Type and latitude of origin

Source

Citation or link

1

SRS399772

CUDH2150

Doubled haploid

LD 42 N

Cornell University

Alan et al. (2004)

2

SRS399773

‘Nasik Red’

Open-pollinated line

SD 20 N

USDA-ARS

PI 271311

3

SRS399774

Omani

Omani landrace

SD 23 N

EU Allium Genebank

HRIGRU 8158

http://ealldb.ipk-gatersleben.de/

4

SRS399775

GTM

‘Criolla molonga’ Guatemalan landrace

SD 14 N

EU Allium Genebank

HRIGRU 7293

http://ealldb.ipk-gatersleben.de/

5

SRS399776

ESP

‘Colorada de conservar’

Spanish landrace

LD 43 N

EU Allium Genebank

HRIGRU 9874 http://ealldb.ipk-gatersleben.de/

6

SRS399778

PRT

‘Cebola valenciana’

Portuguese landrace

LD 41 N

EU Allium Genebank

HRIGRU 11567 http://ealldb.ipk-gatersleben.de/

7

SRS399779

‘Texas Grano 438’

Open-pollinated line

Origin SD 26 N, selected at 37 N

Seminis vegetables seeds, Plant and Food Research

McCallum et al. (2006a)

8

SRS399780

W202A

Inbred line

LD 43 N

University of Wisconsin

Goldman (1996)

F2 families (‘2262’ and ‘G2320’; Baldwin et al. 2012b) were derived by self-pollinating individual F1 plants from the cross between the doubled haploid (DH) onion line ‘CUDH2150’ and the heterozygous landrace ‘Nasik Red’. Families were increased further by cage-pollinating clonal topsets from the original F1 umbels. These families were sown in spring in 2008/9 (sowing date 25 Sept 2008) and 2011/12 (sowing date 9 Sept 2011) seasons at Lat 42° S near Christchurch, New Zealand and in 2011/12 (sowing date 29 August 2011) at Lat 37° S near Pukekohe, New Zealand. A plant was scored as bolting if it exhibited a clearly elongated flower scape and expanded umbel at harvest in early March. Bulb carbohydrate analysis of these populations has been described previously (Revanna et al. 2013).

DNA extraction and molecular marker analysis

DNA was isolated from F2 plants of the CUDH2150 × Nasik Red family ‘2262’ in the 2008/9 season from fresh leaf material or freeze-dried bulb tissue of bolting and non-bolting plants as described previously (Baldwin et al. 2012b; McCallum et al. 2006a). For validation in the 2012 season, template DNA was sampled directly from bulbs by stabbing fleshy scales with a 200 μl pipette tip (Axygen). The tip was then placed into a 50 μl volume solution containing 1 × buffer Gold and 1 μL prepGEM reagent (prepGEM™ Tissue kit; ZyGEM Corp, Hamilton, New Zealand). This was incubated for 15 min at 75 °C with a 5 min hold at 95 °C. An aliquot of the solution was then diluted in half with TE buffer (10 mM Tris and 0.1 mM EDTA) and the remainder stored at −20 °C. SSR, CAPS and HRM marker analyses were carried out as described previously (Baldwin et al. 2012b) but the HRM amplification conditions were modified to include a touchdown PCR. The modified conditions were: 95 °C for 15 min followed by seven cycles of 94 °C for 30 s, 62 °C for 30 s (reducing by 1 °C per cycle), and 72 °C for 15 s. Another 45 cycles were carried out at 94 °C for 30 s, 55 °C for 30 s and 72 °C for 15 s. Then, a final hold temperature of 25 °C for 2 min.

Parallel-tagged sequencing

Template genomic DNA was prepared from the homozygous doubled-haploid reference ‘CUDH2150’ and from pooled DNA bulks of at least 25 plants from seven diverse populations adapted to short- to long-day environments (Table 1). Long-range PCR was performed according to the Expand Long Template PCR System (Roche Applied Science) protocol with Expand Long Template buffer 1 and primers (Table 2). Genomic DNA (100 ng) was amplified in a volume of 50 μl by touchdown PCR using the following conditions: 94 °C, 2 min, followed by 8 cycles of 94 °C for 10 s, 65 °C for 30 s (reducing 1 °C per cycle), and 68 °C for 2 min. Another 40 cycles were carried out at 94 °C for 15 s, 57 °C for 30 s and 68 °C for 3 min (increasing by 20 s per cycle), with a final extension at 68 °C for 7 min. PCR products were purified with AMPure XP (Agencourt), and then subjected to Fragmentase treatment to generate sizes in the range of 300–600 bp according to the NEBNext dsDNA Fragmentase (NEB) protocol. The fragmented PCR products were purified as previously, and blunt end repair, adapter ligation and fill-in were carried out according to Meyer et al. (2007) using NEBNext End Repair Module (NEB) and bar-coding adapters from Meyer et al. (2008). Fragmented PCR products were barcoded by genotype (Table 1) then quantified on a Victor3 plate reader (Perkin Elmer) with the Quant-iT™ PicoGreen® dsDNA Assay Kit (Invitrogen). Pooling, dephosphorylation and restriction digestion were carried out according to Meyer et al. (2008). Normalised tagged fragmented PCR products were pooled by amplicon into two libraries for 454 sequencing on separate 1/16 regions of a GS-FLX plate.
Table 2

Primer pairs used for amplicon re-sequencing of flowering candidate genes

Primer set

Forward primer

Reverse primer

ACR036

GCGAAAACTGAAGGACTGAAA

TTAAAGGAAGGTTCGCCAAA

ACR043

CATTTTCTGAGACGCATTGCT

CGGCAAAATTTGAGAGGAAA

ACR045

GGCGTATACCTCTAAGGGTTATGT

GAGAAACAATCGAACTAAACAGCA

ACR047

TCATTTTGTTTCTTCTGCTTCC

CCCATCATCTCCATTTGCTT

ACR048

CATGGCAAGAGAAAGTGACC

TCATCATCAACTTCAACCCAAA

ACR049

ACAATGTTGCTGCTGCTTGA

GGATGGTGACTGATATACCAGAAG

ACR051

AGGGGCGTTATTTGTGAGTA

CCCTGGTGCATACACAGTTT

ACR054

GCAAAAGGCGGTTGTACATT

AAGGTTGCTCATCCAGTGCT

ACR055

TTGTGATTGTCCGAAGTAGCC

TCCGCAGTTTCATTCTTCAA

ACR056

GCAGAGCAGACACAGCAAAG

CCACATGCTAGATTTCGATCC

Sequence reads were separated into populations based on barcodes using the sff_extract tool (https://github.com/JoseBlanca/seq_crumbs.git). A reference assembly was generated for the CUDH2150 homozygous control using Newbler 2.7 (Roche) with default settings. Reference assemblies from CUDH2150 were submitted to Genbank as Accessions KC677631–KC677640 (Table 3). Gene models were obtained by mapping reference cDNA sequences onto these using gmap (Wu and Watanabe 2005). Experimental data was deposited as NCBI SRA study SRP018027.
Table 3

Primer pairs used and nucleotide diversity of genomic loci from parallel-tagged sequencing of candidate genes

Gene Description

Primer Set

Amplicon size bp

CUDH2150 reference assembly Genbank accession

Reference cDNA Genbank accession

S

Tajimas D

Tajimas Pi

AcFT1

ACR048

3,360

KC677631

KC485348

10

−2.41

2.01

AcFT2

ACR043

586

KC677640

KC485349

18

−2.78

2.44

AcFT3

ACR051

2,250

KC677632

KC485350

10

−2.52

1.57

AcFT4

ACR045

2,079

KC677636

KC485351

2

−1.79

0.47

AcFT5

ACR049

1,437

KC677638

KC485352

7

−1.96

2.82

SOC1-like

ACR055

1,983

KC677637

CF449085

3

−2.21

0.61

AcCO

ACR036

1,928

KC677633

JR851107

25

−1.58

3.36

Constans-like

ACR056

2,796

KC677635

GAAO01015443.1

7

−0.32

4.85

Indeterminate-like

ACR047

853

KC677634

GAAO01013966.1

23

−2.81

1.49

VIN3-like

ACR054

933

KC677639

GAAO01012585.1

4

−2.13

0.62

D and Pi were calculated on CDS only

S number of segregating sites, Pi average nucleotide differences per kb

Reads from the seven heterozygous populations were mapped onto the CUDH2150 assemblies using BWA_SW (Li and Durbin 2010; http://www.ncbi.nlm.nih.gov/pubmed/19451168). SNPs were called using samtools mpileup (Li et al. 2009), patterns of nucleotide diversity were determined using PoPoolation (Kofler et al. 2011a) and allele frequencies were compared among genes and populations using PoPoolation2 (Kofler et al. 2011b). A reproducible workflow containing scripts and all settings for read mapping, analysis and visualisation can be accessed at GitHub in the repository https://github.com/cfljam/onion_PTS_workflow. Variant effects were determined using snpEff (Cingolani et al. 2012). Primer sets for indel, HRM or CAPS markers were designed to variants of interest as described previously (Baldwin et al. 2012b).

Candidate gene mapping

Candidate genes that were not segregating in the F2 population ‘2262’ were tested on the ‘W202A × Texas Grano 438’ mapping population (McCallum et al. 2006a, b) or on sets of A. fistulosumA. cepa alien monosomic addition lines (AMALs) developed by Shigyo and colleagues (1996). Linkage maps were produced using JoinMap v4 (Van Ooijen 2006) as described previously (Baldwin et al. 2012b).

QTL analysis

A subset of 43 non-bolting and 50 bolting progeny from the F2 population ‘2262’ from season 2008/9 was tested with molecular markers for both flowering candidate genes and random markers at regular intervals across the genetic map reported previously (Baldwin et al. 2012b) (Table 4). Single marker and interval analysis using a binary model were used to identify genome regions associated with bolting for the ‘Nasik Red × CUDH2150’ mapping populations using R/QTL (Broman et al. 2003). Significantly associated markers were then tested in 2011/12 on large samples of bolting and non-bolting progeny from populations ‘2262’ and ‘G2320’ grown at Lincoln and Pukekohe.
Table 4

Genetic map locations of flowering candidate gene sequences in onion

Flowering-like gene

Marker(s)

Marker type

Reference sequence

Chromosome

Map position (cM)

Closest marker

Population

AcFT1

ACP670

CAPS-Hpy188i

KC677631

1

40

ACM050

W202A × Texas Grano

AcFT2

ACP691

CAPS-HpaII

JR854848

5

92

ACP929

Nasik Red × CUDH2150 (2262)

AcFT3

ACP675 and ACP676

CAPS-AluI and RsaI

KC677632

2

0

ACM066

W202A × Texas Grano

AcFT4

ACP868

PCR-dominant

KC677636

6

  

AMAL

AcFT5

ACP686

CAPS-EcoRV

KC677638

6

74

ACM115

W202A × Texas Grano

AcFT6 (AcFTL)

APCP996

HRM

KC485353

3

96

 

Nasik Red × CUDH2150 (2262)

AcCO (AcCOL)

ACPA16

HRM

KC677633

4

  

AMAL

Constans-like

ACI046 and ACP681

InDel and CAPS-AluI

KC677635

6

42

ACM121

W202A × Texas Grano

SOC1-like

ACPA09

HRM

KC677637

1

  

AMAL

VIN3-like

ACP689

CAPS-RsaI

KC677639

5

111

ACM065

Nasik Red × CUDH2150 (2262)

Indeterminate-like

ACP102

HRM

CF436030

3

  

AMAL

The naming used by (Taylor et al. 2010) is given in brackets where appropriate. With exception of ACP868, which was dominant, all are co-dominant markers. Where the marker type was CAPS the restriction enzyme used is listed. Gene markers were mapped in the ‘Nasik Red × CUDH2150’ population, or alternatively in ‘W202A × Texas Grano 438’ (McCallum et al. 2006a), or assigned to physical chromosomes using AMAL (Shigyo et al. 1996). Map positions based on linkage mapping in Joinmap (Van Ooijen 2006)

Results

Candidate gene identification and re-sequencing

During bioinformatics searches of existing onion EST data for analysis of the FT-like gene family in onion (Lee et al. 2013), we identified other putative candidates by BLASTX matches to Arabidopsis proteins associated with flowering regulation (Srikanth and Schmid 2011). To identify polymorphisms to enable genetic mapping of FT-like genes, and to identify possible differentiation among populations, we surveyed sequence variation in 5 FT-like and 5 other candidate gene loci using amplicon re-sequencing on the 454 platform. We were unable to develop long-range PCR primers for the other FT-like gene we denoted AcFT6 (Lee et al. 2013), which is synonymous with that denoted AcFTL by Taylor et al. (2010). The non-FT candidates we surveyed included the CONSTANS(CO)-like gene described by Taylor et al. (2010) and four others with similarity to CO, SOC1, VIN3 and indeterminate. PCR products of 800–3,000 bp were amplified from the homozygous reference line ‘CUDH2150’ and seven other lines of diverse short-day and long-day origin. Relatively low nucleotide diversity and negative estimates of Tajima’s D were observed in the coding regions of all except Constans-like (Table 3). These observations are consistent with expectations for genes under purifying selection. Comparison of the distribution of significant (Bonferroni corrected P < 0.01) pair-wise differences in allele frequency among the seven non-DH SD and LD populations showed only two regions exhibiting strong evidence of population differentiation, in the FT-like gene AcFT2 and the SOC1 homolog AcSOC1 (Fig. 1). Notably, these two regions include non-synonymous variants (AcFT2 C411T P/S; AcSOC1 C1376A T > K), which were rare in all genes. Non-synonymous variants were also observed in AcFT4 (T127C L > P; C1847T P > L) and AcCO (C54A L > I), which also exhibited significant differences in allele frequency among some populations (Fig. 1).
https://static-content.springer.com/image/art%3A10.1007%2Fs00122-013-2232-4/MediaObjects/122_2013_2232_Fig1_HTML.gif
Fig. 1

Distribution of SNPs in flowering-related genes that exhibit significant allele frequency differences among seven open-pollinated SD and LD onion populations. a FT-like gene family. b Other candidate genes. Scales denote log10 of P value for Fishers exact test of allele frequency equivalence. Points denote pair-wise comparisons with log10 of P greater than Bonferrroni-corrected 1 % significance level (13.97). Arrows denote regions exhibiting strong differentiation among SD and LD populations. Stars denote non-synonymous polymorphisms

PCR-based markers designed to selected variants using Galaxy-based tools (Baldwin et al. 2012b) were either placed on the onion genetic map or assigned to physical chromosomes where mapping was not possible using AMAL (Table 4). The fixation of several candidate loci in the inbred line ‘W202A’ enabled mapping of several variants in the ‘W202A × Texas Grano 438’ mapping population. Candidate genes tested were present on chromosomes 1–6 with chromosome 6 containing three candidates including two FT-like genes (AcFT4 and AcFT5). Using molecular markers all six onion FT-like genes were assigned to the chromosomes of onion along with other major candidates. Map position could not be assigned for AcFT4 along chromosome 6 because none of the markers were polymorphic in the populations tested. Few of the candidates could be mapped in the ‘Nasik Red × CUDH2150’ F2 population.

Segregation of bolting phenotype in mapping populations

All genetic analysis in this study was based on phenotypic evaluations performed on field-grown plants sown in spring, in a temperate intermediate-day environment, to permit parallel analysis of bulb phenotypes (Baldwin et al. 2012b). Bolting was scored simply and conservatively based on whether a plant exhibited an elongated flower scape and umbel at harvest. During development of the ‘Nasik Red × CUDH2150’ populations, extremely vigorous F1 plants were obtained which exhibited a strong tendency (i.e., over 50 %) to bolt in spring-sown field trials at lat. 42° S. The vigour is consistent with the findings of Hyde et al. (2012) who reported excellent combining ability of DH lines from this background. By contrast, bolting of both parents in such trials in this environment was very rare. Since it is a SD variety, ‘Nasik Red’ is poorly adapted to growth at this latitude and forms relatively small bulbs in spring plantings. This absence of visible bolting in this mal-adapted line may reflect consequence of ‘competition’ between bulbing and inflorescence development (Van Kampen 1970). During initial propagation in 2008/9 of the ‘Nasik Red × CUDH2150’ population, a bolting frequency of ~10 % was observed in family ‘2262’, and samples were retained to permit genetic analysis (Table 5). During the 2011/12 validation trials of the two families, observed bolting frequencies were 12–30 % (Table 5). Mean daily temperatures at Lincoln were slightly cooler in early summer 2011 (mid November to mid-December) than in 2008 (Supplementary Table 3; Supplementary Fig. 3).
Table 5

Frequencies of bolting observed at harvest in progeny of two F2 families from the ‘CUDH2150 × Nasik Red’ cross grown in three environments

Population

Environment

Bolt

2262

Lincoln 2008/9

10 % (50/500a)

2262

Lincoln 2011/12

31 % (184/598)

G2320

Lincoln 2011/12

10 % (29/277)

2262

Pukekohe 2011/12

25 % (144/578)

G2320

Pukekohe 2011/12

12 % (54/453

aApproximate total

Molecular marker and QTL analysis

Markers from the low-density map, developed using non-bolting progeny from the ‘2262’ family (Baldwin et al. 2012b), as well as candidate gene markers, were used to conduct selective genotyping of bolting progeny from the 2008/9 trial. Putative associations with bolting (LOD > 3) were identified on chromosomes 1, 3 and 6 (Fig. 2; Table 6). Those genotypes homozygous for the ‘Nasik Red’ allele for the CAPS marker ACP267 on chromosome 1 tended to bolt (23/27). This association was validated by testing F2 lines from the 2011/2012 season that had been grown in two different environments. The association was significant (P < 0.05) and the marker was then tested in an F2 population ‘G2320’ of the same pedigree, again revealing a significant association (Table 7). We denote this locus AcBlt1. The candidate genes present on chromosome 1 include AcFT1 and AcSOC1. These could not be mapped in the ‘Nasik Red × DH2150’ background and comparative mapping with the alternative population ‘W202A × Texas Grano438’ (McCallum et al. 2006a, b) suggested that AcFT1 was not in close linkage with ACP267 (Fig. 3). Whether this association could be with AcSOC1 remains unknown because this gene could not be mapped in either population. QTL analysis in the non-bolting ‘2262’ progeny from the 2008/9 trial revealed evidence of association of AcBlt1 with bulb weight. Bulbs homozygous for the ACP267 ‘CUDH2150’ allele had higher mean bulb weight (T = −1.88; P = 0.07; df = 23).
https://static-content.springer.com/image/art%3A10.1007%2Fs00122-013-2232-4/MediaObjects/122_2013_2232_Fig2_HTML.gif
Fig. 2

Chromosomal locations of QTL affecting bolting frequency and candidate gene loci in ‘Nasik Red × CUDH2150’ families. Green labels denote candidate genes and red bars denote 1–2 LOD support intervals for QTL revealed by interval mapping (colour figure online)

Table 6

Genetic map positions and estimated variance explained by putative QTL associated with bolting in the ‘2262’ population phenotyped in the 2008/9 season at Lincoln

QTL

Chromosome

Map position

Marker

LOD

% variance explained

1 (AcBlt1)

1

10

ACP267

6.56

17

2 (AcSuc)

3

96

ACP996

4.50

13

3

6

96

ACP395

3.15

5

The LOD calculations and QTL positions were calculated using the Haley-Knott method and a binary trait model. The variance explained by each QTL was calculated using the multiple QTL mapping functions in R/qtl (Broman et al. 2003)

Table 7

Genotype counts and Chi-square goodness-of-fit tests for marker ACP267 linked to the QTL AcBlt1 over seasons and populations

Season

Population

Environment

Bolt

Non-bolting

Genotype counts

χ2 2 df

P value

Genotype counts

χ2 2df

P value

‘Nasik red’ homozygote

Heterozygote

‘CUDH2150’ homozygote

‘Nasik red’ homozygote

Heterozygote

‘CUDH2150’ homozygote

2008/9

2262

Lincoln

23

26

1

19.44

0.000

4

25

14

5.79

0.055

2011/12

2262

Lincoln

55

36

9

50.16

0.000

15

36

24

2.28

0.320

2011/12

2262

Pukekohe

69

72

36

18.46

0.000

5

19

25

18.8

0.000

2011/12

2320

Pukekohe

39

76

17

10.36

0.006

17

61

46

13.6

0.000

https://static-content.springer.com/image/art%3A10.1007%2Fs00122-013-2232-4/MediaObjects/122_2013_2232_Fig3_HTML.gif
Fig. 3

Alignment of linkage groups for onion chromosome 1 and 6 from the onion populations ‘Nasik Red × CUDH2150’ (L) and ‘W202A × Texas Grano 438’ (R), showing relationships between putative QTL and candidate gene markers (green) (colour figure online)

The association detected in 2008/9 on chromosome 3 in the region flanked by markers ACP830 and ACP996 includes the functionally characterised (Vijn et al. 1998) acid invertase locus API89 (NCBI Pr006099770). No bolting plants in the selective genotyping screen were homozygous at API89 for the ‘Nasik Red’ allele (Chi square = 20.607, df = 1, P value = 0.000). However, API89 genotyping of samples from both families in the 2011/12 validation trials revealed no significant association with bolting frequency (data not shown). Comparison with other onion genetic maps reveals that this region of chromosome 3 corresponds with the QTL reported by Havey et al. (2004) that conditions variability in bulb soluble solids content. Onion acid invertase activity has been shown to be regulated by photoperiod (Lercari 1982), and temperature (Benkeblia et al. 2004), and sucrose is known to be a signal for florigenesis (Matsoukas et al. 2012). Analysis of bulb composition in a sample of 504 non-bolting progeny from the 2008/9 trial (Revanna et al. 2013) revealed that bulbs homozygous for the ‘Nasik Red’ allele revealed by marker API89-2 (Pr006099770) had significantly higher sucrose levels (T value = −2.49, P value = 0.014, df = 169). Analysis of bulb sucrose composition data from previous studies of inbred families (McCallum et al. 2006a) also revealed the same pattern of recessive inheritance for high sucrose in relation to API89 genotype (Fig. 4). We, therefore, designate this locus AcSuc. The marker ACP102 designed to an indeterminate-like homolog was also located on chromosome 3 although the map position could not be determined.
https://static-content.springer.com/image/art%3A10.1007%2Fs00122-013-2232-4/MediaObjects/122_2013_2232_Fig4_HTML.gif
Fig. 4

Box plots showing distribution of bulb sucrose content in relation to genotype at the chromosome 3 acid invertase locus API89 (NCBI Pr006099770) in inbred F2:3 families ‘W202A × Texas Grano 438’ evaluated in two environments (McCallum et al. 2006a) and in individual F2 bulbs from ‘Nasik Red × CUDH2150’. Box centre-line and bounds denote median, 25th and 75th percentiles, respectively. Whiskers denote 1.5 times the inter-quartile distance. AA denotes homozygous for the high solids parent (‘Nasik Red’ and ‘W202A’, respectively)

On chromosome 6, homozygous ‘Nasik Red’ genotypes for marker ACP395 bolted (14/14). Comparison with the ‘W202A × Texas Grano438’ linkage map would put this marker closest to the candidate genes AcCO-like and AcFT5, which would both be distal to the end of the ‘Nasik Red × DH2150’ chromosome 6 linkage group (Fig. 4). Possible interactions between the loci were detected and tested but the 2008/9 sample size was too limited to draw any accurate conclusions.

Discussion

Based on expression and complementation studies, we previously concluded that there is strong evidence of a functional role in regulation of bulbing and flowering of three of the six FT-like genes we identified in onion, which we designated AcFT1, AcFT2 and AcFT4 (Lee et al. 2013). We hypothesised that AcFT1 and AcFT4, respectively, promote and repress bulbing, and that AcFT2 may be the key signal for florigenesis induced in vernalized meristems. Interactions among these FT-like genes may condition the developmental transitions among the juvenile, ‘thermophase’, and ‘competition’ phases of onion floral development described in the classical model of Van Kampen (1970). The evidence of population differentiation identified for AcSOC1 with the re-sequencing data is interesting as this could not be placed on the ‘Nasik Red × DH2150’ linkage map but was physically located to chromosome 1 and hence, remains a candidate for the AcBlt1 locus. SOC1 is one of six flowering time genes within the days to heading QTL (dth1.1) interval on chromosome 1 in rice (Thomson et al. 2006). MAS was used to produce sub-introgression lines (SILs) containing one or more of these six candidate genes from Oryza rufipogon in an O. Sativa background. Only the SIL with an introgression containing SOC1 and a Flowering time locus T (FT-L8) was consistently early in all experiments (Maas et al. 2010). In onion, the association of AcBlt1with bulb weight in the 2008/9 trial suggests that AcBlt1 may condition variability in the ‘competition’ phase of floral development (Brewster 2008; Van Kampen 1970), so that bulb formation is favoured over inflorescence development.

Our observation that the AcSuc locus conditions variation in bulb sucrose levels and that it may be associated with bolting in some environments suggests that further investigation of its role in adaptive physiology is worthwhile. Physiological studies have demonstrated that acid invertase activity in onion is very strongly regulated by environment. Photoperiodic regulation of acid and neutral invertases was reported by Lercari (1982), and Benkeblia et al. (2004) reported that acid invertase activity in bulbs of cv ‘Tienshan’ is higher at 10 °C storage compared to 20 °C. Further indirect support for a functional role of acid invertase in bolting is suggested by the origins of the tulip acid invertase cDNA clone used by Vijn et al. (1998) to clone the onion gene corresponding to API89. This clone was shown to encode an acid invertase that was strongly up-regulated by cold and shown to be involved in driving flower stalk elongation (Balk and de Boer 1999). Correlation between carbon physiology traits including invertase activities and flowering date have been reported in maize (Thévenot et al. 2005). Based on the reported effects of environmental factors on onion acid invertase, we would expect that our strategy of using large, non-replicated progenies would be inefficient for validating the effects of any locus conditioning genetic variation in this activity. This is emphasised by the large variability for sucrose content observed among F2 bulbs of ‘Nasik Red × CUDH2150’.

The QTL AcBlt1 had the largest estimated effect on bolting explaining 17 % of the additive variation and this is likely to have been over-estimated given the selective genotyping strategy used in this study. Buckler et al. (2009) detected 29–59 additive QTL for flowering time traits in maize. It was suggested that this could be explained by the out-crossing nature of maize compared to the other self-pollinating model species such as Arabidopsis and rice, and that small-effect QTLs may have been advantageous to ensure that offspring where likely to have overlapping flowering times. Interestingly the putative QTL coinciding with AcSuc on chromosome 3 had a dominant effect. This would be difficult to fix in out-breeding crops such as onion without MAS.

The ‘Nasik Red × CUDH2150’ populations employed in this research are the products of a cross between genetically distant (Baldwin et al. 2012a; McCallum et al. 2008) and divergently adapted parent lines from SD and LD backgrounds. Generating F2 onion populations by mass-pollination of F1 bulbils, induced spontaneously or by cytokinin treatment (Andrew 1951; Thomas 1972), allows evaluation of large samples across multiple environments. The principal disadvantage of this strategy is that it does not permit comparison of replicated combinations of the same allelic combinations over environments, such as those conducted in studies of bulb composition (McCallum et al. 2006b). However, developing a sufficient number of inbred or backcross lines from such a cross would be highly challenging due to the severe impact of adaptive loci on bulb storage and seed production. The more limited population sizes practical using inbred lines limit the power to detect epistatic interactions, and also dominance variance, which is likely to be of great significance in adaptive and yield traits (Wardyn et al. 2007). Availability of parent lines from public sources, gene-based maps and falling costs of genotyping mean that our findings may be reproduced and tested by other researchers. Developing a deeper understanding of the genetic architecture of onion adaptation will require use of many such crosses among landrace and domesticated germplasm, as well as diverse panels phenotyped across environments such as that previously conducted in the tropic regions by Currah and Proctor (1990).

Comparative mapping with the ‘W202A × Texas Grano’ population map and the placement of candidates on this map are important as these lines are from genetic backgrounds widely used in breeding. An obvious extension to this research would be an association mapping strategy using the candidate gene sequences and markers identified from the QTL analysis in more diverse SD and LD germplasm. The action of AcBlt1 and AcSuc in developmental physiology could be determined by analyzing sets of progeny classified by QTL-linked marker genotype and then subjected to different environmental regimes under controlled environments or different planting dates. Given the difference in flowering habits across the vegetable Alliums, comparative mapping with shallot (generally asexually propagated) and closely related Allium such as Japanese bunching onion (A. fistulosum) and garlic (clonally propagated) would also help further understand the major genes involved in controlling flowering in the Asparagales.

Utilisation of exotic germplasm for breeding and selection by breeders to introduce novel traits is severely hampered by bolting. Although it is an easy trait to visually select, dominance could mask inferior genotypes and slow the adaptation. MAS for a recessive trait would be highly desirable for breeders trying to utilise crosses between mal-adapted and elite, well-adapted lines. The application of the marker associations detected in this research could be tested in larger breeding collections or genebanks to assess the predictive values for MAS programmes.

Acknowledgments

This research was funded by the NZ Ministry for Business, Innovation and Employment. We gratefully acknowledge provision of germplasm by the EU Allium Genebank (University Of Warwick) and Cornell University Department of Plant Breeding and of monosomic alien addition line DNA samples by Masayoshi Shigyo (Yamaguchi University, Japan). Field trialling support was provided by Enza-Zaden NZ Ltd, Allium Solutions Ltd., Seminis NZ Ltd and FruitFed Supplies. We thank Zygem Corp (Hamilton, New Zealand) for provision of reagents.

Conflict of interest

The authors declare that they have no conflict of interest.

Supplementary material

122_2013_2232_MOESM1_ESM.docx (38 kb)
Supplementary material 1 (DOCX 38 kb)
122_2013_2232_MOESM2_ESM.docx (12 kb)
Supplementary material 2 (DOCX 11 kb)
122_2013_2232_MOESM3_ESM.docx (29 kb)
Supplementary material 3 (DOCX 29 kb)
122_2013_2232_MOESM4_ESM.docx (12 kb)
Supplementary material 4 (DOCX 11 kb)

Copyright information

© Springer-Verlag Berlin Heidelberg 2014