Molecular Breeding

, Volume 21, Issue 3, pp 383–399

Joint analysis for heading date QTL in small interconnected barley populations

Authors

  • Alfonso Cuesta-Marcos
    • Department of Genetics and Plant ProductionAula Dei Experimental Station, CSIC
  • Ana M. Casas
    • Department of Genetics and Plant ProductionAula Dei Experimental Station, CSIC
  • Samia Yahiaoui
    • Department of Genetics and Plant ProductionAula Dei Experimental Station, CSIC
  • M. Pilar Gracia
    • Department of Genetics and Plant ProductionAula Dei Experimental Station, CSIC
  • José M. Lasa
    • Department of Genetics and Plant ProductionAula Dei Experimental Station, CSIC
    • Department of Genetics and Plant ProductionAula Dei Experimental Station, CSIC
Article

DOI: 10.1007/s11032-007-9139-1

Cite this article as:
Cuesta-Marcos, A., Casas, A.M., Yahiaoui, S. et al. Mol Breeding (2008) 21: 383. doi:10.1007/s11032-007-9139-1

Abstract

The purpose of the present work is to validate the effect of the main QTL determining heading date in a set of 281 doubled haploid lines of barley, derived from 17 small interconnected populations, whose parents are cultivars commonly used in the Spanish barley breeding program. We used 72 molecular markers distributed across the seven chromosomes, particularly in regions known to contain flowering time genes or QTL. A combined linkage map over the 17 populations was constructed. The lines were evaluated in four field trials: two autumn sowings and two winter sowings, and in two treatments at a greenhouse trial, under controlled conditions of photoperiod and temperature. We have found that it is possible to carry out QTL detection in a complex germplasm set, representative of the materials used in an active breeding programme. In most cases two alleles per QTL were detected, though polymorphism of flanking markers was notably higher. The results revealed that there is a set of QTL that accounts for an important percentage of the phenotypic variation, suitable for marker assisted selection. Also, the role of the regions carrying the photoperiod response genes Ppd-H1 and Ppd-H2, the vernalization response genes Vrn-H1 and Vrn-H2, and the earliness per se locus Eam6, of which allele-specific or closely linked markers were available, was confirmed. These results support the use of this kind of approach for the validation of QTL found in single cross population studies, or to survey allelic diversity in plant breeding sets of materials.

Keywords

BarleyConsensus mapHeading dateQTL validation

Introduction

Studies on QTL detection in plants have been usually carried out in populations of progenies derived from single crosses. In small grain cereals, these parents are commonly inbred lines. The advantage of this approach is the simplicity of studying only two alleles per polymorphic locus. Although these studies have led to significant advances in knowledge on genetic control of key traits for many crops, when researchers try to use these QTL in breeding programmes, some drawbacks of the approach become evident. In most cases, parents for QTL studies are selected according to the prospects of finding polymorphism in the cross and, therefore, are chosen among the most extreme ones for the trait of interest. This strategy often led to crosses with little relevance from the plant breeding point of view (Swanston et al. 2006), and/or not adequately representative of germplasm involved in breeding programmes (Varshney et al. 2005). Besides, the results of biparental studies cannot be directly extrapolated to other populations in which polymorphisms may be different (Flint-Garcia et al. 2003), and the resolution obtained in QTL detection is low (Melchinger et al. 2004).

Researchers have sought other approaches to detect QTL over wider arrays of more representative germplasm. Several authors proposed methods of QTL analysis in sets of multiple families, either using some kind of mating design (Muranty 1996; Rebaï and Goffinet 2000; Janninck and Jansen 2001; Jansen et al. 2003), or in multiple families of inbred lines crosses (Xu 1998; Liu and Zeng 2000; Crepieux et al. 2004), such as the materials produced in breeding programmes. These approaches were put into practice by Crepieux et al. (2005) and Christiansen et al. (2006) in wheat, and by Rae et al. (2006) in barley. In this last case, the authors used a set of small populations of doubled haploid lines of barley, derived from crosses between elite varieties currently used in the United Kingdom (small cross mapping). The most extreme approach is the search of QTL in sets of individuals with different kinship, even if this kinship is unknown, through association mapping (approach reviewed in Breseghello and Sorrells 2006).

The main advantages of these new approaches are the higher number of alleles explored, and their direct study in target breeding populations (Breseghello and Sorrells 2006). They are also expected to have a deeper impact on marker assisted breeding programmes, detecting marker associations robust enough for successful marker-assisted selection (MAS) beyond its current use for a few genes with outstanding effects.

This study aims at finding loci useful to carry out MAS for time to flowering in barley. Flowering time is one of the main factors for cultivar adaptation in dryland Mediterranean agrosystems (van Oosterom and Acevedo 1992). QTL for this trait under Northern Spanish dryland conditions were previously detected in the Beka × Mogador population (Cuesta-Marcos et al. submitted). Before using these QTL for MAS, it is necessary to validate their effect at the germplasm-pool level, and to survey the variability present for each QTL in the breeding pool. For these purposes, we studied a set of small populations of doubled haploid lines, derived from interconnected crosses made with a set of parental lines frequently used in the Spanish National Breeding Programme, and representative of the genetic diversity used by the breeders.

Materials and methods

Plant material

Heading date was evaluated in a set of 281 doubled haploid (DH) lines of barley, derived from the F1s of 17 small interconnected populations, consisting of biparental crosses among 14 heterogeneous cultivars commonly used in the Spanish barley breeding programme (several parents used more than once). The number of lines was initially set to 20 for each population, to keep a manageable total population size. During the process of doubled haploid production several lines did not survive, and finally the population size ranged from a minimum of 7 to a maximum of 20 DH lines (Table 1).
Table 1

Description of the 17 small populations of doubled haploid (DH) lines of barley used in this study

Population

Parents

 

Growth type

Row type

DH number

No. of analyzed markers

No. of polymorphic markers

1

Seira (SEI)

Orria (ORR)

S–F

2–6

20

72

53

2

Seira (SEI)

Alexis (ALE)

S–S

2–2

18

80

29

3

Seira (SEI)

Tipper (TIP)

S–W

2–2

20

72

44

4

Albacete (ALB)

Monlon (MON)

F–F

6–6

7

72

42

5

Albacete (ALB)

Plaisant (PLA)

F–W

6–6

19

72

46

6

Alexis (ALE)

Pané (PAN)

S–F

2–6

20

72

53

7

Angora (ANG)

Clarine (CLA)

W–W

2–2

20

80

29

8

Barberousse (BAR)

Albacete (ALB)

W–F

6–6

10

72

50

9

Barberousse (BAR)

Monlon (MON)

W–F

6–6

12

80

36

10

Barberousse (BAR)

Plaisant (PLA)

W–W

6–6

20

72

22

11

Barberousse (BAR)

Tipper (TIP)

W–W

2–6

8

72

27

12

Beka (BEK)

Monlon (MON)

S–F

2–6

20

72

56

13

Clarine (CLA)

Plaisant (PLA)

W–W

2–6

20

80

39

14

Gaelic (GAE)

Tipper (TIP)

F–W

2–2

8

72

30

15

Nevada (NEV)

Beka (BEK)

S–S

2–2

20

80

31

16

Pané (PAN)

Plaisant (PLA)

F–W

6–6

19

72

46

17

Plaisant (PLA)

Orria (ORR)

W–F

6–6

20

72

38

Parent name abbreviations in brackets. Growth type is expressed as follows: S (spring); F (facultative); W (winter)

Phenotyping

Four field trials were carried out at two provinces in North-Western Spain in 2003 and 2004: Vedado estate in Zuera, province of Zaragoza, and Lupiñén and Alerre, nearby locations of the province of Huesca, all latitudes around 41.5°N (Table 2). Two of them were sown in autumn in 2002, and harvested in 2003 (coded as AUTVE03 and AUTHU03) and the other two were sown in winter 2004 (coded as WINVE04 and WINHU04). Experimental design followed an alpha lattice with three replicates at each location. Plots consisted of two rows 1.2 m long and 20 cm apart. Days to heading were calculated as the number of days between the 1st of January and the decimal growth stage 49 (DGS49, Zadocs scale), i.e., the day when approximately 2 cm of awns were visible in 50% of stems. Crop husbandry followed local practice at each location. Nine DH lines were not sown in 2002 because of lack of seed. Heading date of each line in the field trials was estimated according to the alpha-lattice design, and also according to several spatial analysis models in order to minimize error due to autocorrelation among adjacent plots. These included bidimensional autoregressive models (AR1 × AR1), either alone or including tiers and columns of the field trial as covariates. The calculation procedure was Restricted Maximum Likelihood (REML), using the Bayesian Information Criterion (BIC, Schwarz 1978) to select for the best model.
Table 2

Environmental conditions for every field and greenhouse trial

Trial/treatment

Location

Year

SD

CDD

TT

HL-Sa

HL-Ha

RF

AUTHU03

Lupiñén

2002–2003

11-08-2002

687

1291

10.8

14.5

316

AUTVE03

Zuera

2002–2003

11-15-2002

623

1283

10.6

14.5

209

WINHU04

Alerre

2004

01-28-2004

468

932

10.6

15.6

196

WINVE04

Zuera

2004

01-22-2004

427

1055

10.5

15.6

152

NV_LP

Greenhouse

2003–2004

1510

17.0

17.0

V_LP

Greenhouse

2003–2004

392

1147

17.0

17.0

Average daily temperatures for field experiments were gathered from nearby meteorological stations. For growth chambers and greenhouses, thermohygrographes were used

aNatural day length includes civil twilight

SD: Sowing date

CDD: Cooling degree-days (from sowing to average heading date of the trial)

TT: Thermal time (°C) (from sowing to average heading date of the trial)

HL-S: Hours of light (in sowing date)

HL-H: Hours of light (in the average heading date of the trial)

RF: Rainfall (mm) from sowing to the average heading date of the trial

Lines were also tested under controlled conditions in two greenhouse treatments that combined presence or absence of vernalization with long photoperiod. For the V_LP treatment (vernalization followed by long photoperiod treatment), four plants per genotype were vernalized for 8 weeks in a growth chamber at 4°C during the light time (17 h), and 7°C in the dark. When the vernalization period was completed, one plant per DH genotype, and three for the parents, were transferred to a greenhouse with day length set to 17 h, and temperature set to 20/10°C (day/night). For the NV_LP treatment (non-vernalized plants and long photoperiod treatment), four seeds per genotype were sown in pots, directly in an adjacent greenhouse with similar conditions of photoperiod and temperature. After 10 days, only one plant per DH genotype and three for the parents were retained. Sowing dates of both treatments were staggered to ensure that, by the end of the vernalization period, vernalized and unvernalized plants reached approximately the same developmental stage. Plants in the greenhouses were randomly distributed and their positions were rotated weekly. For each plant and experiment, heading date and number of leaves were recorded. Response to vernalization (VER) was estimated for each line as the difference in the number of main stem leaves between the NV_LP and the V_LP treatments. No adjustment was made for the number of leaves recorded in the greenhouse experiments, since plants were regularly rotated, and hence no spatial effects were expected.

Genotyping

Genomic DNA was extracted from young leaf tissue of greenhouse-grown plants as described by Casas et al. (1998). A total of 80 markers were used in this study: 44 SSRs, 23 EST derived SSRs, 2 EST InDel, 6 STS and 5 RFLP. Markers were distributed across the seven chromosomes, particularly on regions where the presence of flowering time genes or QTL are known. A total of 72 markers were used for all DH lines and, for the populations that were not segregating for the major vernalization and photoperiod related regions, eight additional markers were placed on chromosome 6H (scssr09398, Bmac316, Bmag500, scssr02093, Bmag009, Bmac018, scssr05599 y scssr00103). This was the only region that was not well covered by the 72 original markers. All markers were selected according to the literature and previous studies carried out with the population Beka/Mogador, tested under similar environmental conditions.

Allele specific markers for the two main vernalization response genes Vrn-H1, and Vrn-H2 were used in this study. Vrn-H1 was characterized with different combinations of primers: HvBM5.42F, HvBM5.43R, HvBM5.55F and HvBM5.56R (von Zitzewitz et al. 2005); HvBM5.85R, HvBM5.88F, HvBM5.66F and HvBM5.67R (P. Szücs personal communication). These combinations of primers amplify different segments on the first intron of the candidate gene (supplementary Table 1). VrnH2 was studied analyzing the presence of the candidate genes ZCCT-Ha and ZCCT-Hb with primers HvZCCT 01F/02R (von Zitzewitz et al. 2005). HvT SNP22, which is the allele-specific marker for the long-photoperiod response gene Ppd-H1 was tested as reported by Turner et al. (2005). HvVRT2, a possible repressor for Vrn-H1 (Kane et al. 2005), was analyzed as published by Szűcs et al. (2006). We also used the CAPS marker aMWG518/Nhe I (Primer F: 5′-AAAGCTGTC ATACGTCAGC-3′ and primer R: 5′-CTTGTATCTTTGCTGCACG-3′), derived from the RFLP MWG518 since it is tightly linked to the photoperiod response gene Ppd-H2 (Laurie et al. 1995).

For the other markers, whose polymorphism is based on amplification product length, alleles of the same size coming from different parents, i.e., identical by state, were also considered as identical by descent.

Consensus map

A consensus map was constructed using the 17 populations. The approximate chromosome location for each marker was previously known. Recombination frequency and LOD score was calculated for each combination of two markers within each chromosome for the entire set of populations, using Joinmap 3.0 (van Ooijen and Voorrips 2001). The number of individuals considered in each combination of two markers was different, as it depended on the number of populations polymorphic for the pair of markers at each comparison. The consensus map was constructed taking into account the recombination frequency of all possible pairs of markers, and weighing them according to their LOD scores (the higher the number of individuals, the higher the LOD score for a similar recombination frequency), using Joinmap 3.0.

QTL detection

QTL analyses, and calculation of allele effects and marker interactions estimates, were performed using a maximum likelihood method (mixed procedure of SAS v9, SAS Institute Inc., Cary, NC, USA), independently for each field trial and greenhouse treatment. Analyses were performed for one marker at a time, using cofactors to account for variation caused at other regions. Each marker was described by a single variable with n levels, representing the n different alleles. Marker main effects were considered as fixed terms in the model. Inclusion of other markers (cofactors) for QTL analysis has been reported by other authors such as Jansen (1993), Zeng (1994), or Jansen and Stam (1994). Alleles present in less than 5% of the individuals were excluded from the analysis.

Cofactor selection was carried out independently for each experiment, following a forward selection and backward elimination method, an approach analogous to procedures implemented in standard QTL analysis programmes such as QTL Cartographer (Basten et al. 1995). In our case, we performed analyses of variance of the trait studied, using the markers as sources of variance in the following manner: at each step, the marker with the lowest P-value of its F-statistic was added to the model. Markers with lowest P-value of the partial F-statistic were then sequentially added to the model until no marker had a P-value below the 0.05 threshold. At this point, we checked whether all cofactors (markers) included in the model were still significant. Markers with P-values above the 0.05 threshold were then sequentially removed until all markers left were significant. The model remaining after this process was completed was declared the cofactors model.

The significance of each marker was tested including it as an additional source of variation in the cofactors model. A likelihood ratio test (LRT) was performed between the maximum likelihood of the model including the marker tested and the cofactors, and the cofactors model, for each marker. The maximum number of parameters in the model was kept below two times the squared root of the number of individuals (Sakamoto et al. 1986). Cofactors that mapped within a 10 cM window at each side of the marker being tested, were excluded from the model. The significance of the LRT was tested using a χ2 with a number of degrees of freedom equal to the difference in the number of parameters between the two models (in this case k−1, being k the number of alleles of the marker being tested). A False Discovery Rate multi-test adjustment (Benjamini and Hochberg 1995) of the Likelihood Ratio Test P-values was performed to declare the significant markers with a genome-wide threshold of 0.05.

The proportion of variance explained by a QTL at a marker position (R2), was calculated as:
$$ {\text{R}}^{2} = ({\mathop {\text{s}}\nolimits_0^2 } - {\mathop {\text{s}}\nolimits_1^2 })/{\mathop {\text{s}}\nolimits_0^2 } $$
where s0 is the standard deviation of the error of the model including only the cofactors and s1 the standard deviation of the error of the model including the cofactors and the marker being tested, (s1)

Allelic effects and interactions

Allele effects of each significant marker were calculated as the adjusted means of days to heading or leaf number, according to the model that included the significant marker and the cofactors for each experiment.

A multilocus model that included the significant markers was tested against the same model plus all possible interactions (between significant markers), added sequentially, and using the BIC to declare significant interactions. In case of significant interactions, these were also included in the definitive model for allele effect estimation, and in the LRT in the same way as cofactors.

Classification of detected QTL

We classified the significant markers as primary or secondary, according to the comparison of the BIC between the multilocus model of all significant markers and the multilocus model of all significant markers but the one being tested. The LRT P-value of markers with higher number of alleles tended to be more significant, which seems unfair. The use of BIC prevents this flaw to some extent, as it penalizes the number of parameters in the model and, to a much lesser extent, the number of individuals. If the removal of a marker from the multilocus model implies a smaller BIC, it means that the improvement of the LRT caused by that marker was small and/or due to a large number of alleles, i.e., number of parameters. Markers presenting this condition were considered as representative of secondary QTL. Conversely, QTL were declared primary when their removal from the multilocus model with all significant markers caused an increment of the BIC value.

Results

Heading date and leaf number

Ranges of variation for days to heading were similar in the winter-sown and autumn-sown field trials, around 20 days for the DH lines and 6 days for the population means (Fig. 1). In all the experiments, most of the populations showed transgressive segregation.
https://static-content.springer.com/image/art%3A10.1007%2Fs11032-007-9139-1/MediaObjects/11032_2007_9139_Fig1_HTML.gif
Fig. 1

Range of days to heading in the field experiments, and leaf number until heading under greenhouse conditions, for the 17 DH populations evaluated in this study. Days to heading of the DH lines are averaged by sowing season. P1 and P2 are the abbreviations of the parents of each line (see Table 1 for description)

The parents were classified as winter, facultative or spring, according to previous knowledge and cultivar recommendations. Overall, there were minimal differences between parents’ heading dates in the autumn-sown trials (averages for the three types of parents differed by less than 1 day). Spring parents flowered late in the winter-sown trials, probably as a consequence of the effect of long photoperiod in late sowings. Spring cultivars are usually photoperiod insensitive, and thus do not show an advancement of heading date as opposed to most of the sensitive facultative and winter cultivars. On the other hand, it seems that temperature conditions at winter-sown trials were sufficient to provide enough vernalization for most winter cultivars, as most were not delayed in heading date (Fig. 1). The exceptions were cultivars Angora and Tipper, rather late in these trials, suggesting a higher vernalization requirement for them. At the greenhouse experiment, winter cultivars were clearly separated from the rest at the NV_LP treatment, where the lack of vernalization made them extremely late. They were also slightly later than spring and facultative parents at the V_LP treatment. Spring parents showed quite homogeneous behaviours across all trials, as expected. Facultative parents presented diverse responses, as they comprise a variety of combinations of vernalization requirement and photoperiod sensitivity. Among them, cultivar Monlon stands out, as it was the latest parent at the autumn-sown trials, and the earliest one at the winter-sown ones. At the V_LP treatment (the most inductive conditions), all populations presented a rather similar response, with a narrow range of variation for all populations. At the NV_LP treatment, the distribution of the populations was influenced by the growth type of the parents, being winter × winter (ANG/CLA, BAR/PLA, BAR/TIP, CLA/PLA) populations the latest and spring × spring (NEV/BEK, SEI/ALE) the earliest populations.

Field trials showed a high correlation coefficient between experiments with similar sowing date. The correlation between the two greenhouse trials was also highly significant (Table 3).
Table 3

Correlation coefficients of heading dates and/or leaf number among the experiments carried out in this study (codes for experiments in the text)

 

AUTHU03

WINVE04

WINHU04

NV_LP

V_LP

AUTVE03

0.88***

0.67***

0.65***

−0.15*

0.18***

AUTHU03

 

0.55***

0.57***

−0.23***

0.23***

WINVE04

  

0.88***

−0.03

0.19***

WINHU04

   

−0.03

0.11*

NV_LP

    

0.86***

*, ***Pearson correlation coefficients significant at P < 0.05, P < 0.001, respectively

Consensus linkage map

A consensus map was constructed with the molecular data of 80 markers analyzed in the 17 DH populations (Fig. 2). A total of 75 markers were distributed across 14 linkage groups. We set a LOD grouping of 3.0. For instance, we got two linkage groups within chromosome 5H (Fig. 2), which means that there were not at least two linkages between markers of the different linkage groups with a LOD over 3.0.
https://static-content.springer.com/image/art%3A10.1007%2Fs11032-007-9139-1/MediaObjects/11032_2007_9139_Fig2_HTML.gif
Fig. 2

Consensus map for the 17 DH populations. Linkage groups are set at a LOD score of 3.0. Distances are in Kosambi cM. Dotted lines bind linkage groups and unlinked markers within each chromosome. Major heading time loci are indicated in bold types, positions according to literature. (*) Markers analyzed only in populations 2, 7, 10, 11, 13, 14 and 15 and excluded from the QTL analysis

Five markers were not grouped since they did not present a significant linkage with at least two other markers. Marker distances within each linkage group agree with other published barley maps. Linkage groups and unlinked markers, as well as their relative positions, were assigned to barley chromosomes according to other previously published barley linkage maps (Pillen et al. 2000; Ramsay et al. 2000; Moralejo et al. 2004; Rostoks et al. 2005).

Quantitative trait loci

Some markers were significantly associated with heading date under all conditions, as Bmac132 and HvBM5, whereas most of the other significant markers had an effect in at least two experiments (Figs. 3 and 4, Table 4). The amount of phenotypic variation explained by the significant QTL and their interactions ranged from 44 to 67% in the field, and was 93% (NV_LP), 70% (V_LP) and 90% (VER) in the greenhouse. Individual R2 of the primary QTL detected in this study are shown in Table 5. The effect of diagnostic markers for loci Vrn-H1, Vrn-H2 and Ppd-H1 reached very high values (over 40%) at some treatments. The number of alleles per locus ranged from 2 to 10. Although there were more than two alleles in most of the markers, in most cases the evidence suggested the presence of a biallelic QTL (Table 4). This will be discussed further in the next sections, where QTL will be presented according to their phenotypic effect.
https://static-content.springer.com/image/art%3A10.1007%2Fs11032-007-9139-1/MediaObjects/11032_2007_9139_Fig3_HTML.gif
Fig. 3

P-values of the Likelihood Ratio Test (LRT) for all the markers in the field experiments, expressed as −log P. Significance threshold for individual QTL detection is based on an experiment-wise error of α = 0.05 (−log α = 1.3). A False Discovery Rate multi-test adjustment (Benjamini and Hochberg 1995) of the LRT P-values was performed. In brackets are written the values of markers whose value was over 15. Tests were made only at the positions of the markers. Full lines join markers only to facilitate visualization, but they are not indicating test values in the intervals. Dotted lines join linkage groups within the same chromosome

https://static-content.springer.com/image/art%3A10.1007%2Fs11032-007-9139-1/MediaObjects/11032_2007_9139_Fig4_HTML.gif
Fig. 4

P-values of the Likelihood Ratio Test (LRT) for all the markers in the greenhouse experiments

Table 4

Significant marker-trait associations detected at each experiment. The table shows the number of lines and average days to heading, or leaf number, for the different alleles of the significant markers for each experiment. Means are adjusted for other significant markers in each experiment. Letters indicate means separation. We used a False Discovery Rate multi-test adjustment for the P-values and confidence limits for the differences of means with alpha = 0.05. In italics, markers considered as secondary QTL according to the methodology described in the M&M section

Chr.

Marker

Allele

AUTVE03

AUTHU03

WINVE04

WINHU04

NV_LP

V_LP

VER

  

(putative gene)

No.

Mean

No.

Mean

No.

Mean

No.

Mean

No.

Mean

No.

Mean

No.

Mean

1H

aMWG518 (Ppd-H2)

1

99

112.2 b

87

110.0 b

101

6.6b

2

164

114.5 a

142

113.4 a

168

7.5 a

1H

WMC1E8

191

66

112.7 b

62

138.5 b

238

197

114.0 a

194

139.4 a

2H

HvTSNP22 (Ppd-H1)

1

114

142.1 a

115

143.1 a

110

11.0 a

119

7.8 a

2

142

135.9 b

142

138.1 b

145

9.1 b

152

6.6 b

2H

Bmac132 (Eam6)

183

27

112.7 b

28

110.4 b

28

138.5 b

28

139.5 b

27

9.8 b

28

6.9 b

189

82

115.5 a

72

114.3 a

80

141.3 a

81

142.3 a

81

10.6 a

85

7.6 a

191

154

111.8 b

129

110.4 b

148

138.2 b

148

139.5 b

147

9.7 b

156

6.6 b

2H

MWG699 (vrs1)

A

97

114.7 a

89

112.9 a

95

139.8 a

D

41

112.8 b

37

110.2 b

38

137.8 b

K

125

112.6 b

103

119.9 b

124

138.1 b

4H

HvZCCT (Vrn-H2)

0

95

138.6 b

98

8.1 b

97

6.5 b

89

1.0 b

1400

162

139.9 a

157

12.0 a

172

7.8 a

163

3.7 a

5H

HvBM5 (Vrn-H1)

0

58

111.7 c

51

110.4 b

57

138.0 c

57

138.6 c

56

7.9 c

60

6.2 c

56

1.0 c

150

25

115.2 ab

28

112.5 ab

27

138.7 bc

28

140.7 b

28

9.5 b

27

7.3 b

25

1.4 bc

1190

17

112.3 c

18

110.8 b

18

137.3 c

18

138.8 c

19

9.1 b

18

6.4 c

16

2.4 b

1200

25

112.8 bc

19

111.4 ab

24

139.6 b

24

140.4 b

30

9.9 b

26

6.8 bc

24

2.3 b

5150

138

114.7 a

113

113.0 a

130

141.2 a

130

142.1 a

122

13.7 a

138

8.5 a

131

4.3 a

6H

Bmag173 (Eam7)

124

134

113.2 ab

118

111.1 bc

135

139.1 a

135

140.4 b

148

35

111.8 b

38

109.6 c

39

137.4 b

39

138.6 c

150

23

113.5 ab

19

111.5 ab

24

138.8 a

25

140.2 b

152

38

114.2 a

37

112.8 a

38

139.8 a

38

141.3 ab

 

158

18

115.5 a

17

113.5 a

20

139.7 a

20

142.3 a

 

7H

Bmag120

230

58

112.6 b

61

140.0 a

62

141.2 ac

 

232

35

110.8 bc

36

138.9 ab

36

140.1 bd

 

236

43

111.9 b

55

139.8 a

55

141.6 a

 

240

11

110.7 bc

13

137.1 c

13

138.5 d

 

254

15

108.7 c

15

137.4 bc

15

139.8 bcd

 

260

57

111.4 b

65

139.5 a

65

141.2 ab

 

262

10

115.7 a

11

140.2 a

11

141.6 ab

 

Table 5

Individual R2 of QTL declared as primary according to the methodology described in the M&M section. R2 were calculated at the positions of the markers

Marker (putative gene)

AUTVE03

AUTHU03

WINVE04

WINHU04

NV_LP

V_LP

VER

aMWG518 (Ppd-H2)

0.05

0.13

0.06

WMC1E8

0.03

0.03

HvTSNP22 (Ppd-H1)

0.51

0.44

0.19

0.15

Bmac132 (Eam6)

0.09

0.14

0.17

0.18

0.04

0.06

MWG699 (vrs1)

0.05

0.10

0.07

HvZCCT (Vrn-H2)

0.04

0.48

0.29

0.25

HvBM5 (Vrn-H1)

0.08

0.05

0.18

0.20

0.57

0.47

0.40

Bmag173 (Eam7)

0.04

0.09

0.08

0.14

Bmag120

0.13

0.10

0.09

Vernalization genes and their interaction

Using different primer combinations, five different products, associated with the length of the first intron of Vrn-H1 were found (supplementary Table 1).

Vrn-H1 showed an important effect in all field trials, with a larger effect in the autumn-sown experiments. Vrn-H2 was significant only at one of the winter-sown experiments. No interaction between both vernalization genes was detected under field conditions. Under greenhouse conditions, however, the vernalization genes presented a strong interaction. The main effects and the interaction among these two loci accounted for most of the genotypic variation: 88% of NV_LP, 77% of V_LP and 91% of the vernalization effect VER (Table 5).

All possible combinations among Vrn-H1 and Vrn-H2 alleles were present, although the number of lines in the different categories was unbalanced (Table 6). Vernalization requirement was maximum in genotypes carrying the repressor allele at Vrn-H2 (1400 bp), and the 5150 bp allele in Vrn-H1, under long photoperiod. This allelic combination is present in the winter cultivars (Table 1). When the repressor Vrn-H2 was present, the response to vernalization decreased as the size of the intron in Vrn-H1 diminished (1200, 1190 and 150 bp), and was reduced to a minimum in genotypes lacking any amplification product. When Vrn-H2 was absent, the effect of vernalization was rather similar on all Vrn-H1 alleles (Table 6).
Table 6

Interaction between markers at the vernalization genes. Number of lines and average leaf number for all allele combinations of the vernalization genes Vrn-H1 and Vrn-H2. Means separation as in Table 4

HvBM5

HvZCCT

NV_LP

V_LP

VER

(Vrn-H1)

(Vrn-H2)

No.

Mean

No.

Mean

No.

Mean

Allele size (bp)

5150

1400

100

18.4 a

115

10.2 a

112

7.2 a

1200

1400

27

12.5 b

24

7.2 bc

22

4.6 b

1190

1400

15

9.8 cd

14

6.9 bcd

14

2.8 c

150

1400

4

11.0 bc

4

8.0 b

2

2.3 cde

0

1400

11

8.4 ef

15

6.6 cd

13

1.2 de

5150

0

22

9.0 de

23

6.8 bcd

19

1.5 d

1200

0

3

7.3 f

2

6.4 bcde

2

0.1 de

1190

0

4

8.4 ef

4

5.9 de

2

2.0 cde

150

0

24

8.0 f

23

6.5 cd

23

0.5 e

0

0

45

7.5 f

45

5.7 e

43

0.8 de

The 1200 bp allele comes from the Spanish facultative cultivars Albacete and Pané (both also have Vrn-H2). The 1190 bp allele was present only in the cultivar Orria, selected in Spain from a multicross line of CIMMYT materials. The 0 and 150 bp come from spring cultivars. Allelic combinations for each of the tested cultivars are shown in Supplementary Table 1.

Photoperiod response QTL

Markers aMWG518 and Bmag382 (chromosome 1H, bin 10–11) showed a significant effect on heading date in both autumn-sown experiments. aMWG518 also showed a significant effect in the V_LP treatment (Table 4, Figs. 3 and 4). The allele that caused delay in heading time is usually present in the winter cultivars (coded as 2, Table 4).

The HvT SNP22, on the short arm of chromosome 2H, was the most important marker for heading date variation in winter-sown experiments. This marker showed an important effect also in the greenhouse trials (Table 4, Figs. 3 and 4). Spring cultivars usually carried the photoperiod insensitive allele (coded as 1, Table 4).

Earliness per se QTL

The centromeric region of chromosome 2H showed a significant effect in all field and greenhouse trials. Several markers were above the threshold in this region but, in all cases, the peak was located on the SSR marker Bmac132. Lines that presented the 189 bp allele flowered significantly later with respect to the other two alleles (Table 4).

Another QTL of notable effect was found in the centromeric region of chromosome 6H, with a peak on the SSR marker Bmag173. This marker was significant at all the field trials. Bmag173 presents five alleles in this study. Clearly, the band of 148 bp (from parents Beka and Orria) was associated with earlier heading. The other four bands also presented some differences. Although means separations were not clearcut, plants carrying bands of 158 and 152 bp were, on average, between 0.8 to 1.9 days later than those with bands of 124 and 150 bp, suggesting the possible presence of a total of three alleles for the linked QTL.

Other markers (Bmag120, WMC1E8, and MWG699) had a significant effect in at least two experiments. For Bmag120, there were at least two alleles at the linked QTL, represented by bands of 230, 236, 260 and 262 bp (always latest) versus the rest, though a third allele was suggested if this last group was split into band 232 (always intermediate) and bands 240 and 254 (always earliest). Lines carrying the A allele in MWG699, or 238 bp on WMC1E8 (Table 4) flowered later than lines carrying any other allele. Several other QTL were detected in only one experiment, and they were usually classified as secondary according to the criterion described in the M&M section. The effects of these markers are not reported here.

Regarding the effect of markers on chromosome 6H, that were used to characterize only spring × spring and winter × winter populations, only scssr05599 showed a significant effect, in the WINHU04 and NV_LP experiments and in the VER effect (data not shown).

Discussion

The aim of the study was to validate the effect of markers detected in biparental populations, so the estimates are reduced to the positions where the markers are located, and thus no interpolation between markers was made. QTL detection may be hampered in this kind of study by the possible presence of homoplasy (bands of similar size but with different sequence). Different studies have revealed its presence and relevance at the intra- or inter-specific level (reviewed by Estoup et al. 2002). Size homoplasy is a main concern when dealing with phylogenetic relationships among species (Estoup et al. 2002), but for studies involving mapping or gene discovery in biparental populations, looking at the diversity of accessions within a species (as in this study), or among closely related species, variation at the electromorph level (bands distinguished by electrophoretic mobility) provides sufficient resolution to be both efficient and useful (Chen et al. 2002).

Some theoretical studies have pointed out the possibility of occurrence of false QTL when analyzing complex mating designs (Verhoeven et al. 2006). The presence of genetic variance not explained by real QTL, and the presence of linkage disequilibrium generated by the mating design, may lead to the occurrence of spurious associations. Verhoeven et al. (2006) recommended accounting for family structure in these analyses, to prevent the occurrence of false QTL. This was not necessary in this study, as we could account for a great proportion of phenotypic variance (44–93%), thanks to the use of markers for all major regions described as relevant for the trait, and also to the use of a cofactor approach to reduce residual variance over the genome. It is important to get a good consensus linkage map of the markers, in order to obtain accurate results, and especially to determine which cofactors are located in the 10 cM vicinity window used in the analysis. The method used for the consensus map construction, with a maximum of only 20 individuals per population, was different to other methods used for this purpose, usually based on the construction of individual maps for each population (Karakousis et al. 2003; Wenzl et al. 2006). We can conclude that this is an appropriate method, since the outcome agrees well with other barley linkage maps based on single cross populations, and it uses all the information on recombination in the whole set of 281 lines. Another evidence of the accuracy of the map comes from the shape of the LRT score profiles around the QTL. For most cases, the peaks of significance showed a decreasing trend for tightly linked markers at both sides. An exception is the case of marker Hv2F4, in the Ppd-H1 region (2H), which presented a remarkable reduction in significance between two much more significant flanking markers. The reason for this is not a misplacement of the marker (its position fully agrees with other maps), but probably a lack of power in the test at this point caused by the unbalanced number of individuals present in each allele class, as most of the DH lines carried the same allele for marker Hv2F4.

Correlation coefficients between field experiments, and between greenhouse treatments were much larger than correlations between the two sets of experiments. The conditions at both sets of trials are similar only to some extent. The fact that the correlation coefficient between autumn-sown field trials (with sufficient vernalization) and the V_LP treatment were rather low suggests that the photoperiod, longer at the V_LP treatment than at the field, apparently plays a large role in determining flowering in this set of populations.

All markers significantly associated with heading date or number of leaves were previously reported as allele-specific or tightly linked to flowering time determining genes. This remarkable coherence with previous knowledge is a strong point supporting the consistency of the analysis done. The most significant markers in this study are discussed below.

HvT SNP22 (Ppd-H1 region)

Marker HvT SNP22, on the short arm of chromosome 2H, showed the largest effect on heading date in the winter-sown experiments (between 5 and 6 days of heading time difference between the two alleles, Table 4). This is the allele-specific marker for the long photoperiod response gene Ppd-H1 (Turner et al. 2005). Ppd-H1 is a pseudo-response regulator, a class of genes involved in circadian clock function and causes an increased expression of HvFT with photoperiods over 12 h (Turner et al. 2005). HvFT is the barley orthologue to the key flowering promoters FT in Arabidopsis, and Hd3a in rice (Turner et al. 2005; Yan et al. 2006). In plants grown under 16 h photoperiod, under similar photoperiod as for the greenhouse trials of this experiment, peak expression of HvFT occurred at the end of the light period in plants carrying the photoperiod response allele Ppd-H1, whereas reduced expression of HvFT remained constant over the light period for plants carrying the opposite ppd-H1 allele (Turner et al. 2005). As the greenhouse treatments were carried out under constant long photoperiod, Ppd-H1 showed a significant effect, also reported in other studies (Laurie et al. 1994, 1995).

aMWG518 (Ppd-H2 region)

The RFLP and the STS-derived marker aMWG518, and its linked SSR marker Bmag382, on the long arm of chromosome 1H, showed a significant effect on autumn-sown field experiments (Table 4). Both are linked to the photoperiod response gene Ppd-H2, which causes differences on heading date under short photoperiod conditions (Pan et al. 1994; Laurie et al. 1995; Boyd et al. 2003; Francia et al. 2004). Lines with the allele 2 at aMWG518, or the 105 bp allele at Bmag382, had a later heading date. This QTL is a major determinant of heading date under Northern Spanish autumn sowing conditions, in which most of the vegetative phase elapses with photoperiods around 10–11 h, decreasing in the first developmental stages (Cuesta-Marcos et al. submitted). However, a significant effect of these markers was also detected at the V_LP treatment, where plants were grown in the greenhouse under long photoperiods. We have found no previous report of an effect of Ppd-H2 under long photoperiod.

Bmac132 (Eam6 region)

The only significant marker in all field and greenhouse experiments was the SSR marker Bmac132, in bin 8 of chromosome 2H (Figs. 3 and 4, Table 4). This marker co-locates in bin 8 with Eam6 (Franckowiak and Konishi 2002), which confers early heading under both long and short photoperiod conditions (Horsley et al. 2006). In a similar position, Laurie et al. (1995) identified eps2, also with an effect independent of day length. Three Bmac132 alleles were detected in this study. In all experiments, the 189 bp allele showed a marked delay with respect to alleles 183 and 191 bp, which presented similar heading dates. Thus, we clearly found two alleles in the linked QTL in this germplasm pool. The Eam6 region has also been identified as a major determinant of heading date in the population Beka × Logan (Moralejo et al. 2004; Cuesta-Marcos et al. submitted) in autumn sowings under Northern Spanish conditions, and also in other materials under Australian Mediterranean conditions (Boyd et al. 2003). This study confirms the importance of the Eam6 gene under Mediterranean conditions, with an overall effect between three and four days. The peak was consistently present at Bmac132, which has not been widely used thus far, and is not present in the barley genotyping set proposed by Macaulay et al. (2001). Other tightly linked markers, such as Bmac093 and EBmac640, have been used in other studies. Because of the importance of the linked QTL, we propose the use of Bmac132 as a diagnostic marker for Eam6 in breeding programmes, until this gene is cloned.

Bmag173

The SSR marker Bmag173, in the centromeric region of chromosome 6H, was also significant in all field trials. It is located in the same region of the gene eam7, which confers earliness under short-photoperiod conditions (Franckowiak and Gallagher 1997; Stracke and Börner 1998). Although the studies in which this QTL was identified used mutants, in this set of DH lines there is, at least, a consistent effect associated with the allele of 148 bp, which is the earliest in all cases. Canci et al. (2004) described a QTL for heading date, in a similar position, in three mapping populations (Chevron/M69, Stander/MN93, and MS92-299/M81).

MWG699

The STS marker MWG699, on the long arm of chromosome 2H, is tightly linked to the vrs1 gene, which controls the development and fertility of the lateral spikelets (Komatsuda et al. 1999). This marker showed a significant effect in one of the winter-sown trials and in both autumn-sown ones. Kjaer et al. (1995) found a similar effect in a 2-row by 6-row population, but they did not have data to prove whether this effect was due to pleiotropy of the vrs1 gene, or linkage with the QTL now known as Eam6. Some populations of this study were 2-row by 6-row crosses, which were the origin of the polymorphism at MWG699. Thus, population structure might influence heading date QTL detection in a set of populations including 2-row by 6-row crosses, as in our study. But, as the model we have used places cofactors at all other significant loci, any carry over effect from an association between MWG699 and Vrn-H1 (HvBM5), or Eam6 (Bmac132), has been likely removed, and the effect of MWG699 seems a true one. However, it can not be ruled out that vrs1 is associated with a gene segregating between winter and spring barley and having an effect on flowering time which has not been tested in the populations (since genome coverage is not complete).

The SSR Bmag125 was also significant at all field trials, but with the present QTL analysis method, we could not discern whether it had a direct effect on heading date per se, or because of its relative proximity to Eam6 and vrs1.

Vernalization genes and their interaction

The system of the vernalization response genes VRN2 and VRN1, and their epistatic interaction under long photoperiod conditions, is the main factor controlling heading date in temperate grasses (Yan et al. 2003; von Zitzewitz et al. 2005). The epistatic model of the interaction was suggested by Yan et al. (2004), for the Triticeae family and was validated with the corresponding barley genes Vrn-H1 and Vrn-H2 by von Zitzewitz et al. (2005). A central role for photoperiod in this model has also been proposed (Dubcovsky et al. 2006; Trevaskis et al. 2006). According to this last model, Vrn-H2 encodes for a dominant flowering repressor that inhibits the expression of Vrn-H1, which is a central control point for the transition from vegetative to reproductive growth. Vernalization down-regulates the expression of Vrn-H2, allowing Vrn-H1 expression in the winter cultivars (genotype vrn-H1). If the repressor is absent (genotype vrn-H2), there would not exist any vernalization requirement. This model has been validated in the populations Dicktoo × Kompolti korai (Karsai et al. 2005), Hardy × Jubilant (Kóti et al. 2006) and also in Dicktoo × Oregon Wolfe Barley Dominant, Dicktoo × Calicuchima-sib and Calicuchima-sib × Oregon Wolfe Barley Dominant (Szűcs et al. 2007). von Zitzewitz et al (2005) identified a 436 bp region in the first intron of Vrn-H1, critical for the expression of the vernalization requirement.

Lines with all combinations of several distinct alleles at Vrn-H1, combined with presence/absence of the Vrn-H2 repressor were included in this study, though the number of lines in each class was unbalanced (Table 6).

We identified five different alleles in Vrn-H1, whose differences were based on different length of deletions in a region within the first intron. The 5150 bp allele, is usually considered as the winter allele (von Zitzewitz et al. 2005) and, in this study, presented the strongest vernalization response when the Vrn-H2 repressor was present. The 1200 bp allele of Albacete and Pané has a large deletion in the first intron, but conserves the critical region (von Zitzewitz et al. 2005). Its effect on vernalization requirement, as long as the Vrn-H2 repressor was present (1400 bp allele), was intermediate between the typical winter allele (5150 bp), and the alleles carried by spring cultivars (0 and 150 bp), as shown in Table 6. These results confirm the findings by Szűcs et al. (2007), who found that the 436 bp critical region is necessary but not sufficient to allow the presence of full vernalization requirement. Therefore, there could be other regulatory regions in Vrn-H1. The vernalization effect of lines carrying the 1190 bp allele was similar to (Table 4) or intermediate between the 1200 and 150 bp alleles (Table 6), though data were not conclusive enough to declare whether it corresponds to a functionally distinct allele, or it is the same as the 1200 bp allele. Finally, 0 and 150 bp alleles, that correspond with larger deletions including the critical region, showed the lowest vernalization requirements (either with or without the Vrn-H2 repressor). Thus, an allelic series at Vrn-H1 is suggested, with strength of VER effect related to size of the intron on Vrn-H1, when the repressor Vrn-H2 was present, as in Szűcs et al. (2007).

We have determined that it is possible to carry out QTL detection in a complex germplasm set, representative of the materials used in an active breeding programme. QTL position and allelic effects were consistent with estimates found in the literature. In most cases, apparently the markers presented more diversity than linked QTL. Correspondence between marker and QTL alleles was straightforward when marker allele number was low. Most QTL detected had presumably two alleles, with a few exceptions, such as HvBM5 (Vrn-H1). The set of markers used in this study was previously selected to represent regions with heading date QTL. The coverage of the genome achieved, though not complete, was good enough to find QTL that explained a very large proportion of the phenotypic variance. These results support the use of this kind of approach for the validation of QTL found in single cross studies, or to survey allelic diversity in plant breeding sets of materials.

Acknowledgements

This work was supported by the Spanish Ministry of Education and Research (Projects AGL2001-2289, including a scholarship granted to Mr. Alfonso Cuesta-Marcos, and AGL2004-05311) and by the European Regional Development Fund. The authors appreciate the critical reading of the manuscript by Prof. Steve Ullrich, and the valuable suggestions of two anonymous reviewers.

Supplementary material

11032_2007_9139_MOESM1_ESM.doc (37 kb)
(DOC 36.5 kb)

Copyright information

© Springer Science+Business Media B.V. 2007