Suitability of the hydroponic system for early selection of favourable S42ILs
In 2008 and 2011, a set of 42 introgression lines was tested in six hydroponic experiments under two different N treatments. The fixation of the plants in Eppendorf tubes during growth in the hydroponic system allowed us to analyse shoots and roots simultaneously. Huang et al.
, who used a similar plant fixation in their hydroponic system, also highlighted the flexibility of those systems. Also other studies reported that plant responses to low N availability vary among genotypes, developmental stages and plant organs
. Regarding the N treatments, we did not observe any symptoms of starvation on the plants under full supply condition, which indicates that the macro and micro nutrients were sufficiently supplied throughout the experiment
. In contrast, the reduction of N supply caused a severe reaction in most traits studied. For example, root length was increased while plant height was decreased under N deficiency (Table
Taking the parameter heritability as a quality measure of genetic to phenotypic variation, we found that the phenotypic data collected in our hydroponic system predominantly revealed strong and medium heritabilities with h
2 > 70% for five traits and 70% > h
2 > 30% for two traits, respectively. Shoot dry weight (ALL and N100) and chlorophyll content (ALL, N100 and N25) have to be considered as not heritable with h
2 > 10% (Table
3). In this regard, the mixed model analysis of variance was not carried out for the respective treatments of those traits.
For the other traits, the phenotypic data collected from the hydroponic experiments were subsequently analysed in a line by trait association study. In total, 12 out of 58 QTL effects (21%) that we found with juvenile S42ILs in the hydroponic system corresponded to QTL for comparable traits that were detected in previous field and greenhouse studies with adult plants of a similar S42IL set or with the original S42 population (Table
5). The corresponding QTL predominantly controlled the traits plant height (8 x) and, to a lesser extent, tiller number (1 x) and root length (1 x). Both findings, the medium to strong heritabilities observed and the high amount of corresponding QTL, support the idea to use the hydroponic system for early mapping of QTL and early selection of the QTL-bearing genotypes in plant physiology experiments and in plant breeding.
In the following paragraphs, we discuss the detection of juvenile plant growth QTL in wild barley ILs in general and in regard to N deficiency. Finally, we will draw connections between our QTL results, candidate genes and previously described QTL.
Detection of juvenile plant development QTL
The major advantage of growing juvenile barley plants is the fast collection of phenotypic data, for instance within 14 days in our hydroponic system. This time-, cost- and effort-reducing aspects are of major importance for plant breeding where the reduction of time for phenotypic assessment can substantially improve the breeding process because new varieties may be introduced earlier and at lower selection costs to the market. Furthermore, the hydroponic system allows to simultaneously study the development of barley root and shoot traits
. Mattsson et al.
, for instance, could show that in barley root relative growth rate increased while root dry weight decreased at the same time. In addition, the reaction of nutrient deficiencies on barley growth has often been carried out with juvenile plants
[24, 27, 35]. We also conducted a juvenile hydroponic experiment to map QTL that control barley root and shoot traits in regard to N supply.
With 126 significant (P < 0.01) line by trait associations for nine traits under study, the number of QTL detected in the hydroponics experiment was higher than in any other study that was carried out with adult plants of the S42IL population. For example, Schmalenbach et al.
 found 65 associations for seven field agronomical traits in the S42IL population. The increase in the number of associations may be attributed to the different screening methods (hydroponic vs. field testing), the different developmental stages (juvenile vs. adult plants), different traits (shoot and root vs. agronomic traits) and the varying N levels, as already reported for the S42 population by Saal et al.
The highest number of QTL was found for the length parameters leaf length, plant height and root length. The three traits showed high heritabilities of 75.4%, 80.8% and 85.2% across both treatments, respectively. Trait heritability is known to influence the number and the probability of detected QTL. For traits with heritabilities of h
2 < 10%, Ellis et al.
 did not detect any QTL.
The QTL we mapped were scattered across the whole genome (Figure
1). We detected the highest number of QTL effects on chromosome arm 7HL. Here, QTL were found for all traits except root dry weight and shoot to root ratio (Figure
1). In the same chromosomal region QTL effects for plant height and shoot dry weight where already described in studies of the S42IL and S42 population (Table
5). In this regard, genes located on chromosome arm 7HL may be particularly important for regulation of plant growth.
At 75% of all detected QTL, the S42ILs showed increasing Hsp effects (Table
5). In contrast, when studying agronomic traits with S42ILs in the field, only 47% of the detected associations showed increasing Hsp effects
. Since most of the traits measured are related to a fast growth during the juvenile phase, this finding was expected. It may be explained by the fact that the donor of the S42ILs, the Hsp accession ‘ISR42-8’, is well adapted to survival under low nutrient supply. Those S42ILs that reveal QTL effects compared to the elite barley cultivar ‘Scarlett’ may have, thus, contributed exotic Hsp alleles that speed up or increase juvenile growth. Under high N treatment an increase of the traits we evaluated is generally desirable.
A fast juvenile plant development is of great importance, especially when plants are grown under field conditions. Baethgen et al.
 reported that the application of a high dosage of N early in the growing season stimulated tiller formation of malting barley. However, many of these tillers did not develop fertile ears and, thus, the extra dosage of N fertilizer was wasted. In addition, for malting barley one has to consider that a high N supply during grain filling may be conflicting with grain quality
. On the other hand, if plants are cultivated under low N supply, many tillers may also be undesirable since plants may not be able to produce ears and fill the grains for all of the tillers. Multivariate analyse methods like the principle component analysis, are important statistical tools for primary assessment of data structure and, in addition, may support to identify the most important traits for phenotypic selection. However, our primary aim was to locate QTL early on under control and N starvation conditions in a hydroponic system. Thus, the difference in trait performance of the S42ILs compared to ‘Scarlett’ was of greatest interest to us. Due to space and time limitations, a thorough multivariate analysis will be delayed to a follow up study where the most promising traits and, in addition, favourable QTL alleles, will be used for genotype selection. Subsequently, the potential selection gain will be measured within the selected offspring generation.
QTL and N deficiency
Twenty-five of the 42 associations we found under N25 (Model 2) were simultaneously present across treatments (Model 1, see Table
5). Jana and Wilen
 recommended that it is optimal, when lines that are good under non stress conditions, also perform well under stress conditions. This may be of great importance for breeders, who do not have to conduct two parallel breeding programs, and for farmers, who do not have to select different varieties based on expected N stress conditions. On the other hand, selection of breeding lines in stress environments may result in genetic gains by using adapted germplasm, especially in regard to originally low-input crops like malting barley
. However, our results indicated that a number of QTL effects were not in common between both N treatments. Sixteen and 17 QTL were solely detected under N100 or N25, respectively. In one case, the direction of the exotic QTL effect was actually opposed, decreasing leaf length under high N supply while increasing leaf length under low N supply (see Table
5, QLl.S42IL-3H.a and QLl.S42IL-3H.b, present in S42IL-113 and-115, respectively). We, thus, conclude that it may be worth to select barley cultivars for N stress tolerance separately from experiments under low N and high N fertilization.
Associations to tiller number and leaf number were detected across treatments, under N100 and under N25. Similar to our results, Andersen
 also found an increase in the number of leaves under low N treatment in juvenile and adult plants. For leaf length, two QTL where found in the same region of chromosome 3H. The decreasing effect at QLl.S42IL-3H.a and the increasing effect at QLl.S42IL-3H.b were detected under N100 and N25, respectively. Our results suggest that the gene effect on reducing leaf length is reversed under N deficiency. Shoot dry weight of the S42ILs was increased by up to 58.2% under N25 (Table
3), which was already observed by Marshall and Ellis
Regarding plant development, root characteristics are of great importance, which is especially shown when plants are grown under stress conditions
. Most root length QTL showed an increase of trait performance under N25, which may indicate that the S42ILs try to react to N starvation by increasing their capacity to take up N from the solution. In contrast, root dry weight showed a mean reduction of approximately 30% compared to the control treatment, which was also shown in various other studies
[28, 40], and only two increasing QTL were found for this trait. In addition, Karley et al.
 demonstrated in barley that trait differences between the N treatments increased with advancing growth stages, except for root parameters. The authors reported a lack of genotype by N supply interaction for root traits and assumed a limited potential for exploiting genetic variation to improve barley root performance
. However, this finding is in contrast to our study, where we detected a substantial number of root-related QTL effects. Also Naz et al.
 reported a severe Hsp QTL effect on chromosome 5H that is present in a S42IL and caused an substantial increase in root biomass, both, under drought and controlled water conditions. Unfortunately, this particular line, S42IL-176, was not included in the set of S42ILs that we studied under hydroponics.
We did not detect leaf to root ratio or shoot to root ratio QTL solely under low N supply. Also Bahrman et al.
 did not detect variety by N level effects in the shoot to root ratio of eight week old wheat plants. We conclude that in order to detect a change in shoot to root ratio, plants may have to be under stress for a longer time
Because of the missing heritability for chlorophyll content we did not conduct the analysis of variance for this trait. In our experiments genetic effects on chlorophyll content were, thus, not stable across or within treatments. This finding is in accordance with Cartelat et al.
, who reported that the change in chlorophyll content was reflecting the N nutrition status, but showed no direct association to genotype, growth stage or environment, respectively.
Comparison of our results with known candidate genes and QTL
As highlighted before, we revealed more QTL than in any other QTL study conducted with the S42IL or S42 populations, indicating that new genes may have been expressed in our hydroponic study. Furthermore, comparing trait performances of coinciding S42ILs, analysed simultaneously in our study and in field and greenhouse studies
[14, 23], revealed low correlations, for example between juvenile and adult plant height (r < 0.08, data not shown) as well as between juvenile and adult number of tillers and ears, respectively (r < 0.12, data not shown). However, in Table
5 altogether 12 QTL for tiller number, plant height, shoot dry weight and root length under hydroponics corresponded to previously detected field and greenhouse QTL for number of ears, plant height, biomass and root length
[14, 16, 18, 22, 23, 43, 44]. It remains open if the same genes in the respective S42ILs caused the QTL effects under hydroponic and field or greenhouse cultivation. Alternatively, it may also be possible that linked Hsp alleles that are present on the same introgression may have caused the QTL effects in the independent studies. To further elaborate on this question, interesting QTL effects should be narrowed down and validated in high-resolution offspring lines that are derived from the original S42ILs and segregate for the detected QTL effects
In the following paragraphs, we attempt to draw connections between 12 significant Hsp effects and already described candidate genes and QTL effects.
 reported that juvenile tiller number is highly correlated with adult plant ear number, which is also an indicator for yield potential. We mapped one tiller number QTL on chromosome 4H that was consistent with two previously detected QTL for number of ears in population S42
[22, 45]. At QTn.S42IL-7H, S42IL-135 revealed a strong increase in tiller number. Ellis et al.
 also found a strong increasing effect on tiller number in the same region of 7HL with barley seedlings, too.
The number of leaves influences the photosynthetic capacity of the plant. Thus, growth, development and yield may be co-regulated through the control of leaf number. Cuesta-Marcos et al.
 estimated the number of leaves until heading in barley lines and found a significant QTL effect nearby the earliness per se locus Eam6 [Franckowiak and Konishi 1995, cited in 46]. We also found a QTL in this chromosomal region with S42IL-109 under N25 at QLn.S42IL-2H. In addition, the Flt
2L locus, controlling flowering time and plant height, maps to the same region
 and may be present in S42IL-110. Furthermore, VRN
H2 and VRN
H3, two major genes determining the requirement for vernalisation, map to the same chromosomal regions where we found leaf number QTL in the S42IL population
[18, 48–50]. Their influences on growth related traits
[18, 51] may be due to primary effects on juvenile leaf number as it was already shown for the final leaf number in wheat
We detected one decreasing and one increasing Hsp effect for leaf length on chromosome 3H under N100 and N25, respectively. Similar to this, Gregory et al.
 described a reduction in barley shoot length under control conditions and associated the effect with the candidate gene for dwarfism on chromosome 3HL: sdw1[54, 55]. Saal et al.
 also detected two Hsp effects on total plant height under low and high N treatment on 3HL, but there was no difference between the directions of both effects. Thus, we conclude that sdw1 controlled juvenile shoot development in our study; however, plant reaction depends on the status of N supply. On chromosomes 5H and 7H two additional dwarfing genes are located: ari
e.GP and brh1. At both loci, we also detected QTL for leaf length. Chloupek et al.
 already described that ari
e.GP has pleiotropic effects on plant growth in general, including shoot development. However, the ari
e.GP allele is a very rare allele in the elite barley gene pool, making it unlikely that ‘Scarlett’ carries this dwarfing gene. In addition, brh1 was associated with strong effects on plant height and leaf length in a barley backcross population
. S42IL-133 and -134, probably carrying brh1, showed Hsp effects of the same direction on leaf length.
We verified 33 QTL, previously described for adult plant height in studies with the S42 or S42IL population, with twelve lines associated to juvenile plant height (Table
5). In addition, we found candidate genes influencing barley shoot development, which may explain the Hsp effects. On chromosome 1H for example, S42IL-143 caused the strongest decreasing effect on plant height. QTL of the same direction were described with the S42 population
[22, 45] and the decreasing Hsp effect on adult plant height at the HvFT3 locus
 on chromosome 1HL was already described
. Besides, the three dwarfing genes on chromosomes 3H, 5H and 7H revealed strong effects on plant height in our experiments. The effects are similar to those described for leaf length and are discussed in detail in the previous paragraph. Additionally, for each of the three dwarfing genes we refer to Hsp effects described in previous studies carried out with adult plants of the S42IL or S42 population.
On chromosome 4H, an increasing Hsp effect on straw weight was described by Schnaithmann and Pillen
 which is consistent with our effect QSdw.S42IL-4H. Von Korff et al.
 also detected a biomass effect on chromosome 4H.
Semi-dwarf genes are known to have a major impact on aerial plant organs
[54, 55] but their effects on root traits was also affirmed in the literature
. In this regard, Wojciechowski et al.
 found differences in wheat root length due to the dwarfing-gene Rht. With our data, we found two contrasting root length QTL in the chromosomal regions of the barley semi-dwarfing genes sdw1 and ari
e.GP on chromosomes 3H and 5H, respectively. At both loci, contrasting effects for several shoot and root parameters like stabile isotope discrimination for C and N were reported
[35, 56]. We assume that the Hsp alleles at the dwarfing-gene loci may also have caused the root length QTL.
The number of QTL effects detected for the leaf to root ratio was relatively low compared to the traits it was derived from. Wang et al.
 demonstrated that QTL for mathematically derived traits are harder to detect than the component traits they are derived from.