Introduction

Breeding superior oat (Avena sativa L.) varieties involves selection for multiple qualitatively and quantitatively inherited disease, agronomic, and quality traits. Knowledge of genomic regions affecting these traits can be used to breed superior varieties through the use of marker-assisted selection. Oat varieties developed in Canada and the northern United States are typically of spring habit (McMullen and Patterson 1992). However, most quantitative trait loci (QTLs) in oat have been identified using recombinant inbred (RI) populations developed from crosses between the facultative winter-type varieties ‘Kanota’ or ‘TAM O-301’ and the spring-type varieties ‘Ogle’ or ‘Marion’. QTLs affecting four agronomic and two kernel quality traits (Siripoonwiwat et al. 1996) and vernalization response to heading date and plant height (Holland et al. 1997) were detected in the ‘Kanota’ × ‘Ogle’ (KO) population. Vernalization and photoperiod response QTLs were detected in the ‘Ogle’ × ‘TAM O-301’ (OT) population (Holland et al. 2002). QTLs affecting chemical composition and kernel characteristics were found in both the KO and ‘Kanota’ × ‘Marion’ (KM) RI populations (Kianian et al. 1999, 2000; Groh et al. 2001). It is uncertain whether QTL linkages identified in populations segregating for vernalization requirements can be transferred to more adapted oat germplasm. For example, only 5 of 40 restriction fragment length polymorphism (RFLP) allele-trait associations were found to be similar between a set of 64 North American spring and winter-type oat accessions and the KO population (Beer et al. 1997). Molecular marker associations have been reported for QTLs affecting barley yellow dwarf virus tolerance (Jin et al. 1998) and hull percentage (Ronald et al. 1997) in spring oat populations. Knowledge of other QTLs affecting important agronomic and quality traits in spring oat germplasm may ultimately be useful for developing genetic tools for marker-assisted selection.

Genetic diversity remains an important factor in selecting the parents of a mapping population but can be problematic for subsequent QTL analysis. Greater diversity favours the identification of increased marker polymorphism and QTLs with larger effects. Selecting parents from divergent germplasm groups can increase genetic diversity. QTL studies have involved populations derived from parents with different end uses, different inflorescence types, and different seed characteristics. Examples of divergent intraspecific barley mapping populations include: “feed” × “malt” quality (Hayes et al. 1993; Tinker et al. 1996), two-rowed × six-rowed (Kjaer and Jensen 1996; Marquez-Cedillo et al. 1999), and covered × hulless (Heun 1992; Pecchioni et al. 1999). Major genes may affect many quantitative traits (Powell et al. 1985a, 1985b, 1990). If these effects are large, they may interfere with the ability to detect other QTLs with more moderate effects. These populations can be divided into sub-populations corresponding to the different alleles at the major locus, such as was done for the vrs1 locus in barley (Marquez-Cedillo et al. 1999); however, the smaller sub-population sizes will reduce the chances of detecting QTLs. As an alternative to sub-dividing these populations, composite interval mapping (Jansen and Stam 1994; Zeng 1994) and two-locus additive models can be used to incorporate the covariate effects of a major locus as a background marker in the QTL analysis. Epistatic interactions between a major locus and other genomic regions can also be assessed using some QTL analysis software packages, such as MQTL (Tinker and Mather 1995).

In oat, hulless varieties possess a multiflorous spikelet with elongated rachillae and thin lemmas that detach readily from the caryopses during threshing. A review by Marshall and Shaner (1992) indicated that a single, incompletely dominant gene (N1) interacting with modifying genes controls the hulless grain characteristic in oat. The expression of the hulless trait is affected by environmental conditions and management practices (Boland and Lawes 1973; Lawes and Boland 1974; Marshall et al. 1992). Hulless oat varieties have unique end-use advantages compared to covered varieties (Burrows et al. 1993); however, numerous limitations, including low grain yield, have constrained the use of these varieties (Barr et al. 1992; Peltonen-Sainio 1994). Comparatively little breeding effort has been undertaken to improve hulless relative to covered types (Brown and Patterson 1992). For this reason, most hulless by covered oat crosses are made to improve the hulless types; however, these crosses may also transfer desirable alleles to the covered oat germplasm pool.

An RI population derived from a cross between a Western Canadian hulless spring variety and an eastern Canadian, covered spring variety was utilized to develop a molecular linkage map using amplified fragment length polymorphism (AFLP), random amplified polymorphic DNA (RAPD), sequence characterized amplified region (SCAR), and RFLP markers. In this paper, we present this linkage map and summarize the phenotypic and QTL analyses of agronomic and quality traits evaluated on this population in multiple environments in Canada and the United States. The objectives of this study were (1) to develop a new linkage map in oat that can further our understanding of the oat genome, (2) to identify genomic regions affecting agronomic and quality traits in a hulless by covered oat population, (3) to identify the genomic location of the hulless allele (N1) and assess its affect on quantitatively inherited traits, and (4) to compare QTL detection methods that account for the variation caused by the N1 locus.

Materials and methods

Plant material

A ‘Terra’ × ‘Marion’ (TM) F5:6 RI population was generated by single-seed descent from random F2 lines of a cross between the hulless oat variety ‘Terra’ (McKenzie et al. 1981) developed in Manitoba, Canada and the elite covered seeded variety ‘Marion’ developed in Quebec, Canada. Both parents have a spring growth habit. Marion possesses a relatively high groat β-glucan content compared to other Canadian oat varieties (Miller et al. 1993; Humphreys and Mather 1996). A subset of 101 RI lines, which included 59 covered and 42 hulless lines, was used in the construction of the map and for quantitative trait analyses.

Phenotypic data

The 101 RI lines were grown in 13 environments from 1992 to 1997. The testing sites included one location in the United States (Aberdeen, Idaho) and two locations in Canada (Ottawa, ON and Saskatoon, SK). The local variety testing methods and agricultural practices at each collaborating site were used for field experiments. One to four randomized complete blocks were grown in each environment. Sixteen agronomic, kernel, and chemical composition traits (Table 1) were evaluated in 2 to 13 of the environments (Table 2).

Table 1 Agronomic, kernel, and chemical composition traits measured in the ‘Terra’ × ‘Marion’ oat recombinant inbred (RI) population
Full size table
Table 2 Listing of environments with number of replicates grown and traits measured for the ‘Terra’ × ‘Marion’ oat RI population
Full size table

Marker data and linkage mapping

Tissue preparation, DNA purification, and marker analyses were based on methods described by Wight et al. (2003). Clones for RFLP mapping were chosen from among those mapped previously in the KO population (O’Donoughue et al. 1995; Wight et al. 2003). Marker sources are described in Wight et al. (2003).

AFLP markers were detected using reagents and procedures from the AFLP Analysis System I kit (Invitrogen). Primers were labelled using [γ 33]P ATP. Modifications to the standard protocol included adding 50% more Taq DNA polymerase to the selective amplification reaction mixture and increasing the number of cycles after the touch-down phase during selective amplification to 35. Bands were separated by electrophoresis through Long Ranger polyacrylamide gels (BioWhittaker Molecular Applications) and visualized by exposing the dried gels to Kodak BioMax MR-1 film (Eastman Kodak). Marker names, e.g., acg_cac127, reflect the choice of selective primer pair [3 selective bases (EcoRI primer) _3 selective bases (MseI primer)] and location of the band (usually distance run in millimetres) in the first gel analyzed for each primer pair.

The sb978 SCAR marker was derived from the barley cDNA clone bcd978 (Heun et al. 1991). The clone was end-sequenced using an ABI 373 automated DNA sequencer, then a pair of PCR primers (sb978f1 = AAGGTCAGCCAGGTCAAATG and sb978r1 = CGAGATAACCTCCCTCCCTC) was designed and synthesized. SCAR amplification of genomic DNA was performed using a Temptronic thermocycler (Barnstead/Thermolyne) with a standard PCR program of 5 min at 94°C; 33 cycles of 1 min at 94°C, 1 min at 49°C, 1 min 20 s at 72°C; and 5 min at 72°C. PCR reactions were conducted in a volume of 25 µl containing 1–2 mM genomic DNA, 200 nM of each primer, 1.5 mM MgCl2, 200 μM dNTP, 2.5 U Taq DNA polymerase, and 1× PCR buffer. PCR products were separated on 1.8% agarose gels, stained with ethidium bromide, and visualized with UV light.

Linkage mapping was performed using both MAPMAKER, version 3.0 for PC (Lander et al. 1987) and G-MENDEL Win32, version 0.8b (Holloway and Knapp 1993) software. A robust framework map was developed by choosing marker subsets and determining orders that were consistent between both programs. Detailed descriptions of these procedures are provided by Wight et al. (2003). Linkage group numbers are preceded by the designation TM to distinguish them from previously reported oat linkage groups. The linkage group TM_4+16 arose when TM_4 and TM_16 were merged following analysis of additional markers.

Statistical analysis

The phenotypic data for most traits measured in the TM population were first analyzed for the whole RI population, then separately for the covered and hulless sub-populations. Tertiary kernel data were collected only for the covered lines. Analysis of variance for each trait in each environment and over environments was performed using the SAS general linear model procedure (SAS Institute 1985). Due to the varying number of replicates, means for each entry were used to compute the analysis of variance across environments. Both environments and genotypes were considered random. The variance components associated with genetic variance (σ2 g) and error variance (σ2 e) were determined using the expected mean squares from the analysis of variance. Heritability (h 2) on an entry mean basis was estimated from the genetic and error variance components using the formula: h 2=σ2 g/[σ2 g+(σ2 e/r)], where r is the number of environments. Factor analysis was conducted on overall entry means using the SAS factor procedure to investigate the relationships among the 15 traits evaluated in the whole population. Extracted factors with eigenvalues greater than 1 were rotated by the varimax method (Kaiser 1958) to allow biologically relevant interpretation of the results. Pairwise comparisons between covered and hulless sub-population means were conducted using t-tests.

Three approaches for QTL detection in the TM population were evaluated using the MQTL software package (Tinker and Mather 1995). These methods of QTL identification included: (1) simple interval mapping (SIM) in the whole population, (2) SIM in the covered and hulless subsets of the population, and (3) testing two-locus additive and epistatic models with N1 as a covariate. Tests for QTL main effects and QTL-by-environment (QTL×E) interactions were conducted using SIM. The two-locus models were used to evaluate QTL main effects only. Permutations (n=5,000) were performed to determine threshold test statistics for declaring significant (P<0.05) experiment-wide QTL main effects and QTL×E interactions, as described by Tinker et al. (1996). Parameters of QTL main effects and QTL×E interactions were estimated using multi-locus linear models in the full population and within each sub-population. Allelic effects for a given locus were estimated by the average difference between homozygous classes. The proportions of phenotypic variation explained by models with QTL main effects and both QTL main effects and QTL×E interactions were estimated from the coefficients of determination (R 2).

Results and discussion

Parental identity

The hulless parent of this population was originally identified as ‘Tibor’. Later, molecular testing (data not shown) confirmed that this parent was actually ‘Terra’, the only other hulless variety in the crossing block. Supplementary field tests containing ‘Terra’, ‘Tibor’, ‘Marion’, and two RI progeny were conducted (data not shown), and these data supported the phenotypic identity of ‘Terra’ as a parent. The fact that this information was obtained only after considerable data were collected has several implications for this study. The absence of orthogonal data from ‘Terra’ relative to the full set of RI lines restricts our ability to draw conclusions regarding transgressive segregation relative to ‘Terra’. Preliminary marker surveys including ‘Tibor’ were not useful for identifying markers that segregated in the RI lines. Surveys including ‘Terra’ were more accurate, but the seed of the original ‘Terra’ parent was not available and some inconsistency between alleles in these newly grown ‘Terra’ plants and those segregating in the RI lines was also observed. This is not unusual, since we have observed considerable marker heterogeneity in different seed sources from identical oat varieties (unpublished). Some differences between alleles in the ‘Marion’ parent and those in the RI lines were also apparent, so the most reliable surveys were those including a subset of 15 RI lines. This information is important for anyone choosing to conduct further studies using this population.

Molecular marker mapping

A total of 430 polymorphic markers were scored in the RI mapping population. These included 242 AFLP markers, 160 RFLP markers, 26 RAPD markers, and one SCAR marker. The hulless character was also scored and mapped as a phenotypic marker. These markers formed 35 linkage groups at LOD 6. A robust subset of 124 ordered markers was selected from these, favouring RFLP markers where possible. After the construction of a framework map (Fig. 1), the remaining markers were either placed between the two framework markers giving the highest chi-square statistics, or left unplaced when the chi-square statistic was less than 6. The chi-square statistic (testing deviation from independent assortment) was used for computational simplicity, as it is a close approximation of the LOD score. Placement was based on a simple division of the framework distance by the ratio of distances between the placed marker and the two closest framework markers. The total map distance represented by this framework map is 727 cM. However, this does not take into account the linked regions on either end of each linkage group and the regions flanking each of the unlinked markers. Comparison to the KO linkage map (Wight et al. 2003) indicates that this represents approximately 38% of the oat genome, based on the ratio of total map lengths. The mapping of AFLP markers was an attempt to obtain map coverage in regions unrepresented by RFLP markers. A few novel regions were identified by this method (TM_12, TM_32, TM_34, TM_35, and TM_39), but most AFLP markers fell within clusters, such as those on groups TM_1 and TM_15, or remained unlinked. This may be a result of inadequate polymorphism in some regions of the parental genomes. Detailed comparisons of this map to other oat maps will be the topic of a future study.

Fig. 1a, b
figure 1figure 1

Molecular marker linkage map from the ‘Terra’ × ‘Marion’ F6 recombinant inbred oat population. Markers from the framework map (a) are indexed by cM positions in the complete marker list (b). Placed markers are represented by the placement bars shown to the left of each group in the framework map and by non-indexed marker names in the complete marker list. The placed marker order in b corresponds to a right-to-left reading of the placement bars in a. Note that the N1 locus maps to TM_5. RFLP suffixes reflect the names of the same marker bands in the ‘Kanota’ × ‘Ogle’ reference population map or are designated as x, y, or z if the bands were not polymorphic in KO. The suffix p indicates the designation is “probable”

Full size image

Quantitative traits

Combined environment analysis indicated that there were differences among both lines and environments (P<0.01) for all traits. Heritability for the traits ranged from 0.38 for lodging to 1.00 for milling yield (Table 3). Some of the high heritability values can be attributed to the segregation of the hulless allele (N1) within this population. Sub-population means were different (P<0.05) for 11 of the traits. The covered sub-population had higher lodging, yield, kernel weight, milling difficulty, plump kernels, and FIA β-glucan content and lower visual score, milling yield, test weight, thin kernels, and oil content versus the hulless sub-population. Differences in lodging and oil content would not be expected to be due to differences in kernel type alone. Despite the fact that the hulless parent ‘Terra’ was not evaluated in this study, useful transgressive segregation was observed relative to ‘Marion’ for reduced tertiary and thin kernels and increased yield, kernel weight, and plump kernels (Table 3). Within the covered sub-population, there were lines with higher milling yield and test weight than Marion (data not shown).

Table 3 Agronomic, kernel, and chemical composition trait data for 101 RI progeny from the ‘Terra’ × ‘Marion’ cross and for the hulless and covered sub-populations
Full size table

Factor analysis was conducted to group correlated traits by their association with extracted factors or principal components. Four factors with eigenvalues greater than 1.0 explained 77% of the total variance in the population (Table 4). Eight traits, each showing a difference between the covered and hulless sub-population means, had high (>0.5) loadings for the first factor. The four chemical composition traits had a high factor loading for the second factor. Lodging and milling difficulty had high loadings for the third factor, and heading date and height had high loadings for the fourth factor. Only oil had a high loading factor for more than one factor (1 and 2). High milling yield, high test weight, high percentage of thin kernels, high oil content and high visual score were associated with low yield, low kernel weight, and low percentage of plump kernels. High β-glucan (FIA and NIR) and high oil were associated with low protein. There were positive associations between lodging and milling difficulty and between heading date and height. The distribution of factor scores for the 101 TM RI progeny on the first two rotated axes is shown in Fig. 2. There is a separation of most covered and hulless lines based on the factor 1 scores. These results suggest that the N1 locus is the underlying biological component influencing factor 1. The effect of major genes on quantitative traits has also been demonstrated in barley, where genes controlling height and biomass production were linked to the V locus (Powell et al. 1990). It is plausible that the traits with high loadings in the other factors also have a common genetic basis.

Table 4 Loading of fifteen traits on four factors obtained from factor analysis with a varimax rotation of axes, and the percentage of total variance accounted for by the four factors
Full size table
Fig. 2
figure 2

Distribution of 59 covered and 42 hulless lines from the ‘Terra’ × ‘Marion’ RI population on the first two rotated factor axes

Full size image

Quantitative trait loci: main effects

For each trait, four QTL main effect tests were conducted: (1) using the whole population, (2) using N1 as a covariate, (3) using only covered lines, and (4) using only hulless lines. Figure 3a–d shows the result of these analyses for grain yield. Regions of the genome showing significant test statistics for the other traits are summarized in Fig. 4. In total, 34 main effect QTLs were detected using one or more QTL detection methods. These QTLs were located in 19 regions of the genome. Markers linked to some of these QTL are undergoing further study in validation and marker-assisted selection projects.

Fig. 3
figure 3

Genome-wide scans of test statistic for simple interval mapping (SIM) of main effects for grain yield using the whole population (a), N1 as a covariate in the model (b), covered lines only (c), and hulless lines only (d). Horizontal lines represent threshold value corresponding to the analysis used, based on 5,000 permutations. Peak marker locations are labelled using trait (Table 1) and linkage group designations

Full size image
Fig. 4
figure 4figure 4

Linkage groups containing significant marker-trait associations based on simple interval mapping (solid bar), two-locus model containing N1 locus (horizontally striped bar), SIM QTL × environment interactions (Q×E) (diagonal striped bar), and epistatic interactions with N1 (white bar). Peak marker locations are labelled using trait (Table 1), sub-population (A all, C covered, H hulless) and ‘Terra’ × ‘Marion’ linkage group designations. LOD scores are given on the line below the peak marker designation followed by the additive effects and parental donor of the high allele (T ‘Terra’, M ‘Marion’) at this genomic location

Full size image

Whole population

Eighteen QTLs were found in the whole population using SIM. Main effect QTLs affecting lodging, visual score, yield, kernel weight, milling yield, plump kernels, test weight, thin kernels, and oil content mapped at or near the N1 locus on linkage group TM_5. Other traits mapping to common locations in the genome included oil and protein (unlinked marker bcd1829y), and milling yield and test weight (unlinked marker cdo212x). QTLs affecting heading date, plant height, oil, β-glucan (NIR), and lodging were also detected using the whole population.

N1 as a covariate

Fourteen QTLs were found in the population when N1 was used as a covariate in the additive genetic model. Multiple-trait QTLs were detected in four regions: linkage groups TM_2 (lodging and yield), TM_5 (test weight and plant height), and TM_35 (lodging, kernel weight, and plump kernels), and with the unlinked marker bcd1829y (oil and protein). QTLs affecting heading date (TM_9), oil (TM_4+16), β-glucan (NIR) (TM_18), and plump kernels (TM_4+16 and unlinked marker bcd1872a) were also detected when N1 was included as a covariate in the model. Seven of the QTLs detected by using N1 as a covariate were not found in the previous whole population analysis.

Covered sub-population

Nine QTLs were found in the covered sub-population using SIM. Lodging and yield QTLs mapped to linkage group TM_2 and kernel weight and plump kernel QTLs mapped to linkage group TM_35. Additional QTLs affecting test weight (TM_1), plump kernels (TM_4+16), β-glucan (FIA) (TM_29 and unlinked marker act_cat244), and thin kernels (unlinked marker bcd1872a) were also identified using the covered lines. Four of the QTLs found in the covered sub-population were not identified in the previous analyses.

Hulless sub-population

Eight QTLs were found in the hulless sub-population using SIM. Oil and protein QTLs mapped to the same linkage group (TM_15). QTLs affecting thin kernels (TM_1), test weight (TM_5), oil (TM_4+16), β-glucan (NIR) (TM_21), and yield (TM_22) were identified in the hulless sub-population. Five of the QTLs detected in the hulless sub-population were not identified in the previous analyses.

Localization of the N1 locus

The hulless condition in oat was previously found to be controlled by a single, incompletely dominant gene (N1) interacting with modifying genes (Boland and Lawes 1973). In this study, we have localized the N1 gene to linkage group TM_5 (homologous to KO_24_26_34). Using an indirect measure of the hulless phenotype, milling yield or groat percentage, we have identified another genomic region (unlinked marker cdo212x) affecting this trait. Further study will be required to confirm that this region contains a gene or genes that interact with the N1 locus to affect the hulless condition.

QTL-by-environment interactions

For each trait, three QTL-by-environment (QTL×E) interaction tests were conducted using the whole population, the covered lines, and the hulless lines. Figure 5a–c shows the result of these analyses for grain yield. Regions of the genome showing significant test statistics for the other traits are summarized in Fig. 4. In total, 24 QTL×E interactions were detected in ten regions of the genome using one or more detection methods. Thirteen of these QTL×E interactions corresponded to main effect QTLs affecting the same traits. Eleven of the interactions were localized to the region near the N1 locus.

Fig. 5
figure 5

Genome-wide scans of test statistic for SIM of QTL×E interactions for grain yield using the whole population (a), covered lines (b), and hulless lines (c). Horizontal lines represent threshold value corresponding to the analysis used, based on 5,000 permutations. Peak marker locations are labelled using trait (Table 1) and linkage group designations

Full size image

Epistatic interactions with N1

Tests for epistasis revealed three regions where QTLs interacted with N1 to affect thin kernels. These are shown in Fig. 4 and are located on linkage groups TM_1, TM_18, and at the unlinked marker bcd1829y. A main effect QTL affecting thin kernels in the hulless sub-population and test weight in the covered sub-population was located at the same location as the epistatic interaction for thin kernels on linkage group TM_1. QTLs affecting chemical composition traits in the whole population were also localized to the same regions showing significant epistasis with N1, including β-glucan (NIR) on linkage group TM_18 and oil and protein at bcd1829y. These observations suggest that there may be a relationship between thin kernels and other kernel or chemical composition traits. Epistatic interactions between marker loci were also reported in the KO RI population (Holland et al. 1997) and the OT RI population (Holland et al. 2002).

Multiple locus models

Multiple locus models were constructed using markers identified as being linked to QTLs or QTL×E interactions for each of the traits studied (Table 5). The main effect models accounted for 7–95% of the phenotypic variance for the various traits in the whole, covered, or hulless sub-populations. The main effect plus QTL×E interaction models accounted for 11–97% of the phenotypic variance. These values are similar to those obtained for heading date, test weight (covered population), β-glucan, and oil content in other oat QTL studies (Siripoonwiwat et al. 1996; Kianian et al. 1999, 2000). For example, MQTL multi-locus models for oil content accounted for 22% and 48% of phenotypic variance in the KM and KO RI populations, respectively (Kianian et al. 1999), compared to 33–39% obtained in our study. For other traits influenced by the N1 locus such as yield, milling yield (or groat percentage), and test weight in the whole population, our estimates were considerably higher than other studies (Siripoonwiwat et al. 1996; Groh et al. 2001). Estimates for yield (covered population) and plant height were lower that those obtained by Siripoonwiwat et al. (1996). This could be related to the reduced opportunities for QTL identification resulting from partial genome coverage in the TM map. Inclusion of the epistatic interactions with N1 for thin kernels increased the proportion of variance explained by the model from 63% (Table 5) to 86%. Multiple locus models for traits associated with vernalization response in KO were also improved by adding epistatic interactions (Holland et al. 1997).

Table 5 Percentage of phenotypic variance explained by QTLs affecting 15 traits measured on all, covered, or hulless lines from the ‘Terra’ × ‘Marion’ RI oat population. ME Model including significant main effect QTL only, ME+QE model including both main effect QTLs and QTL-by-environment interactions
Full size table

Trait relationships revealed by QTL analysis

Nine genomic regions identified in this study contained QTLs affecting multiple traits. The trait relationships identified by QTL analysis generally agreed with the factor analysis discussed previously in this report. The most important of these regions was on linkage group TM_5 near the N1 locus. Fourteen traits showed significant QTLs or QTL×E interactions in this region. A study by Ougham et al. (1996) on the biochemical differences between hulless and covered oat lines lead the authors to propose that the N1 locus encodes a regulatory molecule, as do other homeotic genes previously characterized in plants. Our data support this hypothesis by illustrating the effect this region has on multiple traits, some of which are not kernel characteristics.

The N1 allele has been shown to affect lignin deposition in developing spikelets (Ougham et al. 1996). This phenomenon could also be responsible for the effect of the hulless gene on lodging if there were differential lignin distribution patterns in the stems of hulless versus covered lines. Lodging resistance is an important goal of most cereal breeding programs, due to its effect on grain yield and quality (Keller et al. 1999). Lignin plays an important role in plant stem rigidity. Plants with altered lignin composition, such as those containing the brown midrib mutation in maize, have lower stem strength and are more susceptible to lodging (Weller et al. 1985). Understanding the role of the N1 locus in stem lignin deposition may lead to improved strategies for overcoming this important problem in oat.

The effects of the N1 allele on agronomic and chemical composition traits may also be due, in part, to agricultural management practices. In our study, the hulless and covered lines were planted, grown, and harvested in the same field experiment with similar treatment given to both types. However, there are slightly different requirements for seeding, threshing, and storage of the two seed types (Burrows et al. 1993), which could have affected trait expression in some environments. Furthermore, identical management practices may have different impacts on the more fragile hulless seeds (Marshall et al. 1992).

Other genomic regions affecting multiple traits were identified from main effect QTLs and QTL×E interaction analyses of the TM population. These include lodging with plant height (TM_2 and 5), grain yield (TM_2), kernel size (TM_35), and test weight (TM_5); test weight with thin kernels (TM_1) and milling yield (unlinked marker cdo212x); oil with protein (TM_15 and unlinked marker bcd1829y); and the kernel size parameters kernel weight and plump or thin kernels with each other (TM_35 and unlinked marker bcd1872a). Lodging QTLs have also been shown to map to regions affecting plant height in a wheat × spelt population (Keller et al. 1999) and regions affecting plant height, grain yield, kernel weight, and test weight in barley (Tinker et al. 1996). Several kernel size parameters measured by digital image analysis mapped to the same regions in both KO and KM (Groh et al. 2001). No previous QTL study has reported the oil-protein relationship in oat; however, there have been other reports of a strong negative correlation between these traits (Brown et al. 1966).

Comparison to other QTLs in oat

Accurate comparison of QTLs across populations requires linkage maps with common markers. In hexaploid oat, this process is confounded by the multi-allelic nature of many RFLP markers and the prevalence of inter-varietal chromosome structural rearrangements and duplicated loci (O’Donoughue et al. 1995; Portyanko et al. 2001; Wight et al. 2003). This may be further complicated by the possibility of gene loss during the evolution of allopolyploids, as reported in wheat (Kashkush et al. 2002).

When possible, TM linkage groups were assigned corresponding KO (Wight et al. 2003) or KM (Groh et al. 2001) linkage group numbers to facilitate inter-population map comparison. QTLs for test weight, plant height, grain yield, oil content, and plump kernels (or kernel width) were detected in the same genomic regions in different oat populations (Table 6). When the locations of different RFLP markers associated with identical cDNA clones were analyzed, QTLs were also detected in homoeologous regions in the different oat populations (Table 6). These include heading date, oil content, and β-glucan. Also, the region containing a β-glucan QTL on TM_18 is homoeologous to a region near the oat β-glucanase gene located on KO_24_26_34. The β-glucanase enzyme catalyzes the hydrolysis of 1,4-β-d-glycosidic linkages in β-d-glucans containing 1,3- and 1,4-bonds (Yun et al. 1993). In addition, a plant height QTL×E interaction on TM_25 maps near umn145 and a second homeologous umn145 locus was previously found to be associated with the dwarfing gene Dw6 in an ‘OT207’ × ‘Kanota’ population (Milach et al. 1997). To our knowledge, this is the first formal report of QTLs affecting lodging, kernel weight, and protein content in oat, so no comparison amongst populations was possible for these traits.

Table 6 Locations of common QTLs across oat mapping populations: ‘Terra’ × ‘Marion’, ‘Kanota’ × ‘Ogle’, ‘Kanota’ × ‘Marion’, and a recurrent selection population for seven agronomic and quality traits. TM group ‘Terra’ × ‘Marion’, KO group ‘Kanota’ × ‘Ogle’, KM group ‘Kanota’ × ‘Marion’, RS group University of Minnesota recurrent selection population
Full size table

Comparison to QTLs identified in other populations may also provide evidence for determining trait relationships at the genetic level. Three QTLs affecting kernel size (area, length, and width) were detected in KO and KM in regions that correspond to grain yield QTLs in TM. QTLs affecting kernel area and length on KM_11 (homologous to TM_2) and a QTL affecting kernel width on KO_5 (homologous to TM_22) were identified in the KM and KO populations, respectively (Groh et al. 2001). Kernel area and width, measured using digital image analysis, were correlated with kernel weight in a previous study (De Koeyer et al. 1993) and kernel weight is one of three components of grain yield in oat (Grafius 1956). The KO_11 (homologous to KM_11 and TM_2) linkage group is an interesting region of the oat genome, as it has been shown to be the location of QTLs affecting oil (Kianian et al. 1999) and β-glucan (Kianian et al. 2000) in KO and KM and various agronomic characters in KO (Siripoonwiwat et al. 1996), a recurrent selection population (De Koeyer et al. 2001), and TM (current study).

A QTL affecting β-glucan on TM_18 not only corresponds to the genomic region containing the β-glucanase gene on KO_24, but also to the homoeologous location of a heading date QTL on the same linkage group (Siripoonwiwat et al. 1996). Heading date QTLs were also associated with β-glucan QTLs in three genomic locations in previous reports (Kianian et al. 2000; Groh et al. 2001). The common location of heading date and β-glucan QTL could indicate a pleiotropic effect caused when later maturing lines fail to complete the optimum grain filling period, which results in shrunken kernels with less endosperm and enhanced relative β-glucan content. The unlinked marker bcd1872a, associated with thin and plump kernels in TM, maps to a region on KO_27 where a test-weight QTL in KO is located (Siripoonwiwat et al. 1996). This cross-population trait association is the same as one detected on TM_1 in our study. As more QTL studies in oat are completed, comparison of QTLs across populations will give insight into the genetic basis of traits in the various populations.

QTL detection methods

One-half (9 of 18) of the QTLs identified in the whole population mapped to the region containing the N1 locus on linkage group TM_5. Additional QTLs were localized using N1 as a covariate in the additive genetic model and in the two sub-populations. QTLs detected only within the hulless and covered sub-populations had lower LOD scores, which was likely due to the smaller sample sizes. Analysis of the whole population, combined with the use of N1 as a covariate in the analyses, was the best approach for identifying the most important genomic regions affecting agronomic and quality traits in the TM population. Population size is a critical factor in determining the probability of detecting a QTL within a mapping population (van Ooijen 1992). Therefore, statistical approaches that maintain the full population size are more powerful than approaches that use separate sub-populations. Further studies within larger covered and hulless sub-populations may validate the robustness of the QTLs identified in this study.