Introduction

Hybrid maize breeding programs were started in Central Europe in the 1950s. As a promising heterotic pattern, high yielding US dent lines were crossed with adapted European flint lines (Schnell 1992). During the initial phase, dent inbreds from the US were primarily selected for earliness. The steady influx of dent germplasm from North America to Europe has continued over the past 50 years. In contrast, the parental flint inbreds were developed by selfing from a few European open-pollinated varieties (OPVs) such as Lacaune, Lizagaraute, Gelber Badischer Landmais, and Rheintaler (Messmer et al. 1992). Therefore, it can be conjectured that (1) a bottleneck occurred in the flint pool during the transition from OPVs to hybrids and (2) OPVs, which did not serve as germplasm source for the original flint inbreds, contain untapped allelic variation useful for future breeding progress. Detailed information about a reduction in genetic diversity could help to emphasize the importance of identifying germplasm sources for broadening the elite breeding pools.

In the second phase of hybrid breeding, new lines were primarily developed by second cycle breeding, i.e., from crosses among elite inbreds within heterotic groups. Moreover, outstanding elite lines were shared as parents of different commercial hybrids. In combination with intensive selection, this is expected to result in a reduced genetic diversity in the breeding pools but even more seriously in the cultivars grown by farmers. However, the risk of genetic erosion does not only depend on plant breeding practices but also on the system that delivers the final products of plant breeding to the market. In the European Community, this includes the regulations to register new varieties and the marketing of registered varieties. Statutory testing of new varieties is required to register them on the national lists (UPOV 2002). Afterwards, their acceptance by farmers depends on the amount and quality of the marketing effort of breeding companies but also on further series of voluntarily recommended lists based on regional trials. Consequently, only a few of the registered varieties are grown on a large scale.

Monitoring the genetic diversity available to farmers is important, because plant breeding practices, the registration procedures, and the marketing of new varieties could have caused a potential genetic erosion and, consequently, a potential increased genetic vulnerability of cultivated varieties. Snap-shots of the diversity present in maize breeding programs were reported (e.g., Messmer et al. 1992). In addition, the temporal trend of genetic diversity was investigated for a single US breeding program (Duvick et al. 2004a) as well as for important public US lines (Lu and Bernardo 2001). However, no information is available on the temporal trends in genetic diversity of important Central European maize varieties cultivated by farmers.

The main goal of our study was to monitor the temporal trends in genetic diversity over the past five decades among maize cultivars with the largest acreage in Central Europe. Our objectives were to (1) investigate how much of the genetic diversity present in important Central European OPVs has been captured in the elite flint germplasm pool, (2) examine changes in the genetic diversity among the most important commercial hybrids as well as in their dent and flint parents, (3) analyze temporal changes in allele frequencies between the dent and flint parental inbreds, and (4) investigate linkage disequilibrium (LD) trends between pairs of loci within the set of parental dent and flint lines.

Materials and methods

Plant materials

A set of 85 maize hybrids (Table 1) representing the most important cultivars grown in Germany over the last five decades was chosen based on acreage and seed sale figures. Estimates about the relative importance of these cultivars were obtained from (1) the Bundessortenamt, (2) the Deutsches Maiskomitee, and (3) maize experts from sales, breeding, production, and research. For 67 of the 85 hybrids, all parental inbred lines or parental populations were available. For 11 single-cross hybrids only one parent and for seven single-cross hybrids none of the parents was available. In a companion study (Reif et al. 2005), 30 individuals of each of five prominent flint populations from Central Europe (Gelber Badischer Landmais, Maleksberger, Mahndorfer, Strenzfelder, and Rheintaler) were examined. These populations had played an important role in the pre-hybrid breeding era in Germany and were grown in 1950 on more than 93% of the acreage (Oettler et al. 1976).

Table 1 Description of the hybrids grouped into four time periods: single cross (2W), three-way cross (3W), double cross (4W), and top-cross hybrids (TO) and open-pollinated varieties (OPVs) from the pre-hybrid breeding era (PH)
Full size table

According to the year of release, the five OPVs and 85 hybrids, as well as their parents, were grouped into pre-hybrid breeding era (PH) (<1951), Period A (1951–1975), Period B (1976–1985), Period C (1986–1995), and Period D (1996–2001). The decreasing trend in the length of the time periods reflects the decreasing number of years during which the cultivars played a significant role on the market.

SSR analyses

DNA was extracted from the leaf material of parental inbred lines employing a modified CTAB procedure (Saghai-Maroof et al. 1984). For the 11 single-cross hybrids, where only one parent was available, DNA was also extracted from the hybrids using the above mentioned procedure. For the seven single-cross hybrids, where none of the parents was available, DNA was isolated from leaf material and maternal pericarp tissue of the hybrids, using the GenElute Plant Genomic DNA miniprep Kit (Sigma, Cat Nr. G2N350). A total of 55 SSR markers uniformly distributed across the maize genome was used as described by Reif et al. (2005). Details about SSR amplification, detection, and allele calling procedure were described in our companion paper (Reif et al. 2005).

Statistical analyses

Unknown marker genotypes of parental inbreds of single-cross hybrids were deduced by “subtracting” the genotype of one of the parental inbreds or the maternal pericarp tissue from the genotype of the hybrid (Stojsin et al. 1996; Wang et al. 2002). Thus, in total genotypes of 148 parental inbred lines were available. These lines were assigned to the dent and flint heterotic groups (1) according to pedigree information or (2) by applying the K-means clustering algorithm (Hartigan and Wong 1979) assuming two groups, if no pedigree information was available. This resulted unambiguously in 66 flint and 82 dent lines. Most hybrids consisted of dent × flint crosses except for five hybrids that had only dent parents. Allele frequencies of the hybrids were determined from their parental components. Associations among the hybrids were analyzed by applying principal coordinate analysis (PCoA) (Gower 1966) based on the modified Rogers’ distance (Wright 1978), which was calculated from the allele frequencies of the hybrids. Associations among parental dent and flint inbreds were determined by PCoA based on modified Rogers’ distance calculated from the allele frequencies of the groups of parental dent and flint lines from each of the four time periods. In addition, the number of alleles (n A ), the number of new alleles (n N ), and the number of lost alleles disappearing from one period to the next (n L ) were determined.

Rogers’ distances (RD) (Roger 1972) were calculated among dent lines, among flint parental components, and among dent versus flint parental components. In addition, RD were determined between individual genotypes within OPVs and between individual genotypes of different OPVs. These RD estimates were averaged within each time period.

Consider two maize hybrids X and Y. The average RD between two randomly selected individuals from two different maize fields planted either with hybrid X or Y reflects the diversity between cultivated varieties. Thus, all pairwise comparison of hybrids of a certain time period represents a measure of the diversity present at that time between cultivated varieties at an individual level. The average expected RD between two randomly chosen individuals of hybrid X with its two parental components π′ and π′′ and hybrid Y with its two parental components π′′′ and π′′′′, was calculated according to Melchinger (unpublished data) as:

$$ \begin{aligned} E{\left( {\text {RD} {\left( {X,Y} \right)}} \right)} & = \frac{1} {m}{\sum\limits_{i = 1}^m {\left[ {{\sqrt {\frac{1} {2}} }} \right. + {\left( {\frac{1} {2} - {\sqrt {\frac{1} {2}} }} \right)}} } \\ & \quad {\left( {H^{{{\text{I}} \times {\text{III}}}}_{0} + H^{{{\text{I}} \times {\text{IIII}}}}_{0} + H^{{{\text{II}} \times {\text{III}}}}_{0} + H^{{{\text{II}} \times {\text{IIII}}}}_{0} } \right)} \\ & \quad + {\left( {\frac{1} {2}{\sqrt 3 } - {\sqrt {\frac{1} {2}} }} \right)}{\left( {H^{{{\text{I}} \times {\text{II}}}}_{0} + H^{{{\text{III}} \times {\text{IIII}}}}_{0} } \right)} + {\left( {{\sqrt 2 } - \frac{1} {2}{\sqrt 3 } - \frac{1} {2}} \right)} \\ & \quad {\left( {H^{{{\text{I}} \times {\text{II}} \times {\text{III}}}}_{0} + H^{{{\text{I}} \times {\text{II}} \times {\text{IIII}}}}_{0} + H^{{{\text{I}} \times {\text{III}} \times {\text{IIII}}}}_{0} + H^{{{\text{II}} \times {\text{III}} \times {\text{IIII}}}}_{0} } \right)} \\ & \quad + {\left( {{\sqrt {\frac{1} {2}} } - 1} \right)}{\left( {H^{{{\text{I}} \times {\text{III}}}}_{0} H^{{{\text{II}} \times {\text{IIII}}}}_{0} + H^{{{\text{I}} \times {\text{IIII}}}}_{0} H^{{{\text{II}} \times {\text{III}}}}_{0} } \right)} \\ & \quad + {\left( {{\sqrt {\frac{1} {2}} } + 1 - {\sqrt 3 }} \right)}H^{{{\text{I}} \times {\text{II}}}}_{0} H^{{{\text{III}} \times {\text{IIII}}}}_{0} \\ & \quad \left. { + {\left( {2{\sqrt 3 } + 1 - 3{\sqrt 2 }} \right)}H^{{{\text{I}} \times {\text{II}} \times {\text{III}} \times {\text{IIII}}}}_{0} } \right] \\ \end{aligned} $$

Here, \( H^{{I \times II}}_{0} = {\sum\nolimits_j^{n_{i} } }\,{p}\ifmmode{'}\else$'$\fi_{j} {p}\ifmmode{''}\else$''$\fi_{j} \) \( H^{{I \times II \times III}}_{0} = {\sum\nolimits_j^{n_{i} } }\,{p}\ifmmode{'}\else$'$\fi_{j} {p}\ifmmode{''}\else$''$\fi_{j} {p}\ifmmode{'''}\else$'''$\fi_{j} \) and \( H^{{I \times II \times III \times IIII}}_{0} = {\sum\nolimits_j^{n_{i} } }\,{p}\ifmmode{'}\else$'$\fi_{j} {p}\ifmmode{''}\else$''$\fi_{j} {p}\ifmmode{'''}\else$'''$\fi_{j} {{p}\ifmmode{''''}\else$''''$\fi_{j}} \) where p′ j , p′′ j , p′′′ j , and p′′′′ j are the frequencies of allele j at locus i in parental component π′, π′′, π′′′, and π′′′′, respectively, whereas m refers to the number of loci and n i to the number of alleles at locus i. \( E{\left( {\overline{{RD}} {\left( {X,Y} \right)}} \right)} \) ranges from 0 (both hybrids are identical single cross hybrids) to 1 (both hybrids share no alleles in common). The average RD between individual genotypes of one hybrid reflects the diversity present in a certain field. The average RD between individual genotypes within hybrid X with parental components π′ and π′′ was also calculated with the above formula using π′=π′′′ and π′′=π′′′′. The average of these values within time periods is therefore a measure of the diversity present at that time within cultivated varieties at an individual level. The maximum average RD between individual genotypes ranges from 0 for single-cross hybrids to 0.43 for double cross hybrids. The Wilcoxon rank sum test was applied to examine the significance between pairwise comparisons of average RDs (Hollander and Wolfe 1973).

Linkage disequilibrium between all pairs of loci was tested within the dent and flint groups with an exact test described by Weir (1996). The power for detection significant LD between pairs of markers depends on the sample size of the germplasm groups, which varied in our study. Therefore, a re-sampling strategy was used to obtain comparable estimates: a random sample of 15 individuals from each time period for the set of dent and flint lines was chosen for the above analyses; sampling was repeated 50 times, and the results were averaged. All analyses were performed with the software Plabsoft (Maurer et al. 2004), which is implemented as an extension to the statistical software R (Ihaka and Gentleman 1996).

Results

A total of 404, 323, 328, and 337 alleles were observed in the hybrids, OPVs, as well as in dent and flint parental components, respectively (Table 2). In total, 28% of the alleles present in the five OPVs were not recovered in the flint lines. Most of them (97%) had an allele frequency below 0.1 in the OPVs. Population Mahndorfer had the highest number of unique alleles (35) and Rheintaler the lowest number of unique alleles (20) not present in the flint lines.

Table 2 Number of open-pollinated varieties (OPVs), hybrids, or parental components (N), number of different parental components (N D ), number of alleles (n A ), number of new alleles (n N ) and number of lost alleles disappearing (n L ) from one period to the next in the varieties of Period PH, A, B, C, and D in marker assays with 55 SSR markers
Full size table

The PCoA of the 85 hybrids revealed a shift in allele frequencies from Period A to D (Fig. 1). This shift was also observed in the PCoA of the parental dent and flint lines grouped into the four time periods with respect to the second principal coordinate (PC2) (Fig. 2). In addition, PC1 revealed that for each time period, the group of dent lines was clearly separated from the group of flint lines.

Fig. 1
figure 1

Principal coordinate analysis based on modified Rogers’ distances between the 85 hybrids of Period A (filled square), B (open triangle), C (asterisk), and D (filled triangle). PC1 and PC2 are the first and second principal coordinate, respectively

Full size image
Fig. 2
figure 2

Principal coordinate analysis based on modified Rogers’ distances of the 82 dent and 66 flint lines grouped into four time periods. PC1 and PC2 are the first and second principal coordinate, respectively

Full size image

The number of alleles (n A ) in the hybrids decreased consistently from Period A to D (Table 2). Period B showed both the highest number of new alleles (n N ) and lost alleles (n L ). While n A decreased for the flint parental components from Period A to D, with the highest loss of alleles from Period A to B, n A increased for the dent lines from Period A to B but decreased thereafter monotonically to Period D. Across all four periods, the dent and flint lines had 65% of the alleles in common. The dent lines contained 22% and the flint lines 13% unique alleles not present in the opposite heterotic pool.

The average RD between individual genotypes was higher within OPVs (0.37) than within hybrids of Period A (0.18) (Table 3). The average RD between individual genotypes within hybrids decreased almost linearly from Period A to D. In addition, the average RD between individual genotypes of different varieties was higher for the OPVs (0.51) than for the hybrids of Period A (0.45). The average RD between individual genotypes of different hybrids increased from Period A to B but decreased from Period C to D.

Table 3 Average pairwise Rogers’ distances based on 55 SSRs between individual genotypes within hybrids (W-HY), between individual genotypes of different hybrids (A-HY), among 82 dent lines (DE), among 68 flint parental components (FL), and between dent and flint parental components (DE × FL) grouped into the four time periods (A, B, C, and D).
Full size table

For all time periods, the average RD among lines was greater for dent (0.62) than for flint inbreds (0.54). The RD between dent lines increased from Period A to B and decreased substantially from Period C to D (Table 3). The RD between flint parental components also increased from Period A to B but decreased slightly to Period D. The RD between dent and flint parental components from the same time periods increased from Period A to B, with a minor decrease to Period D.

Linkage disequilibrium was significant at a probability level P=0.01 for 20.2% of the SSR marker pairs in the 82 dent lines, and for 17.2% in the 66 flint lines (Table 4). We observed almost as much inter-chromosomal as intra-chromosomal LD. The percentage of SSR loci pairs in LD in a random sample of 15 lines decreased in the dent and flint inbreds from Period A to C with a slight increase from Period C to D.

Table 4 Percentage of SSR loci pairs in linkage disequilibrium (LD%) and percentage of SSR loci pairs in linkage disequilibrium of a random sample of 15 individuals (S-LD%) at probability level P=0.01 for unrelated dent and flint lines based on 55 SSRs. Test results were separated into LD between loci on the same chromosome (Intra) and LD between loci on different chromosomes (Inter). NT refers to the number of tests
Full size table

Discussion

Two events in the recent history of maize breeding had a major impact on the genetic diversity among and within cultivated varieties in Central Europe (Schnell 1992). First, the transition from OPVs to hybrids occurred in the 1950s. Second, during the 1980s there was a shift in cultivation of top- or double-cross hybrids to three-way- or single-cross hybrids. Production of the latter became economic due to an improved per se performance of inbred lines and was driven by (1) the major breeding goal to develop higher yielding hybrids and (2) farmer demands for increased uniformity of varieties (Hallauer et al. 1988). Other important breeding goals included selection for earliness and tolerance to high plant densities as a consequence of changing agricultural practices. In the starting phase of hybrid breeding in Central Europe, a few prominent public lines dominated as parents of hybrids (Messmer et al. 1992). Subsequently, the public lines were replaced by proprietary inbreds. Furthermore, at the beginning, the main focus was on grain usage while silage use became predominant later on. The aim of this study was to investigate the impact of these trends in plant breeding on the level and structure of genetic diversity within and among maize varieties of economic importance in Central Europe.

Genetic diversity loss during the transition from OPVs to hybrids

The significantly lower average RD between individual genotypes within hybrids than within OPVs reflects the reduced diversity within hybrid cultivars. This can be explained by the small number of individual genotypes used to generate hybrids (Messmer et al. 1992) but also by directional selection. With increasing genetic homogeneity of varieties, a higher yield is expected but also an increased genetic vulnerability due to a reduced population buffering (Allard and Bradshaw 1964; Hallauer et al. 1988). Considering the total population of cultivated cultivars during a certain period, the reduction in genetic diversity within varieties could have been counterbalanced by an increased diversity between hybrids. However, the RD between genotypes of different varieties decreased from OPVs to hybrids. In addition, the restricted number of five OPVs analyzed in this study presumably led to an underestimation of the diversity present in OPVs, and, thus, a considerable loss of diversity was observed within but also between varieties during the transition from OPVs to hybrids.

The large proportion of alleles of the OPVs (28%), that were not recovered in the flint lines clearly indicates a reduction in the allelic diversity during the establishment of elite flint inbreds used in hybrid breeding. Apart from exotic germplasm, adapted OPVs are a promising source to broaden the genetic base of the elite flint breeding pool. Population Mahndorfer had the highest number of unique alleles (35) among the OPVs, which were absent in the flint lines. This suggests that the germplasm contribution of Mahndorfer to the elite flint pool was low. Consequently, this OPV is a very promising source for untapped allelic variation.

Genetic diversity trends in hybrids and their parental dent and flint components

The PCoA of the hybrids revealed a clear shift in allele frequencies from Period A to D (Fig. 1). This was corroborated by the PCoA of the dent and flint lines (Fig. 2). The dent and flint lines evolved in a parallel way with respect to PC2, reflecting that dent and flint inbreds have undergone similar selection pressures. Drift and introgression of novel plant material could also have caused a shift in allele frequencies, but then a parallel evolution of both sets of lines is unexpected.

We observed a monotonic decrease in genetic heterogeneity within hybrids from Period A to D (Table 3). This reflects the change in the types of hybrid (Table 1). The transition in types of hybrid is also most likely the major cause for the decrease of n A from Period A to D (Table 2) and is in close agreement with results reported by Duvick et al. (2004a), who investigated 51 US hybrids with 100 SSRs. The trend in n A is of interest particularly for Periods A and B, because it decreased drastically for the flints but increased for the dents. The decrease in n A for the flints can be mainly explained by the use of OPVs as parental component of the top-cross hybrids Hymador and Perdux in Period A, and the disappearance of top-cross hybrids in Period B (Table 1). The increase in n A for the dent lines suggests an augmented introgression of novel germplasm from the US in the 1970s, as indicated also by n N .

The average RD between individual genotypes of different hybrids is influenced by (1) the RD between dent lines, (2) the RD among flint parental components, and (3) the RD between dent and flint parental components. The average RD of all the three groups increased from Period A to B and, consequently, caused an increase in the average RD between individual genotypes of different hybrids (Table 3). The higher RD among dents in Period B than A can be explained by the reduced use of identical parental components in different hybrids (Table 2). In addition, the enhancement can be due to a broadening of the genetic base of dent lines by adapting new germplasm from the US, as indicated by the high value for n N found in Period B (Table 2). The higher average RD among flints of Period B than A can be explained by the dominance of a few public inbred lines at the beginning of hybrid breeding (Messmer et al. 1992). Especially, the cross of the two French lines F2× F7 served in 30% of the hybrids of Period A as parental component (data not shown).

The average RD between individual genotypes of different hybrids decreased from Period C to D, which was mainly caused by a reduction of the diversity in the dent pool (Table 3). On the one hand, the decrease of diversity in the dent pool can be explained by an increased use of certain individual lines as parental component of various hybrids by breeders in Period D (Table 2). On the other hand, one could speculate that intensive selection and focusing on a small number of key parents for the development of new lines have caused a reduction in the diversity of the dent lines. To verify whether a reduction in the diversity of the dent pool is due to a multiple use of (1) inbreds as parents of several hybrids or (2) certain inbreds as key parents for the development of new inbreds, further investigations are required with fingerprinting of germplasm from the individual breeding programs and examining the pedigree relationships among the inbred lines.

We found a significantly higher average RD between dent than between flint lines, which is in accordance with previous studies with germplasm representing the diversity within breeding pools (Messmer et al. 1992; Lübberstedt et al. 2000). The higher diversity in the dent than in the flint pool can be explained by the genealogy of the germplasm. The Central European dent germplasm encompasses a wide range of materials of North American origin with genetic background of different heterotic groups such as SEPS Stalk and Non-SEPS Stalk. By comparison, the Central European flint germplasm is genetically narrower due to (1) a bottleneck during the introduction of maize into Europe and/or (2) a loss of genetic diversity during the selective adaptation to European conditions (Rebourg et al. 2003).

Genetic diversity trends between dents and flints

The high number of unique alleles, which are exclusively present in either the dent or flint lines (35%), reflects the large divergence between the allelic profiles of both heterotic groups. The differences between the dent and flint heterotic groups existed already at the beginning of hybrid breeding in Germany (Fig. 2; Table 3). This can be explained by the long isolation between the parental germplasm sources used to establish these heterotic groups: the US dent and the European flint pool (Rebourg et al. 2003). The clear divergence of the heterotic groups at the beginning of hybrid breeding in Central Europe is in contrast to results for US germplasm. Duvick et al. (2004b) found that the US germplasm was at the starting phase of hybrid breeding rather unstructured but diverged with ongoing reciprocal recurrent selection.

Trends in linkage disequilibrium

The absence of significant LD in the OPVs (Reif et al. 2005) suggested that LD was either absent or the applied marker density was not high enough to detect LD caused by linkage within populations of the pre-hybrid era, which served as the germplasm base for the elite flint pool. In contrast to the OPVs, LD between SSR marker pairs was observed in Period A in the dents and flints (Table 4). This LD can be caused by admixture of inbred lines originating from populations with different gene frequencies and/or from drift in small populations. In addition, this LD can also be caused by epistatic interactions or suppressed recombination. For the dent and flint pool, both causes are very likely because (1) the dent pool was established with limited number of US dent lines from distinct heterotic groups (Messmer et al. 1992) and (2) the flint pool was developed with a limited number of inbreds originating from selfing in European OPVs with different allele frequencies (Reif et al. 2005).

The standardized fraction of loci pairs in LD decreased for the dents and flints from Period A to C (Table 4). This can be explained by the successive generations of inter-mating since the establishment of both germplasm groups. The slight increase in the standardized fraction of loci pairs in LD from Period C to D for the dents and flints can be due to (1) a high selection intensity for favorable epistatic interactions of alleles from different loci (Falconer and Mackay 1996) and/or (2) drift in breeding populations with a small effective population size. Consequently, one could speculate that the number of different allele combinations increased during the first three time periods but decreased in the last period in Central European maize germplasm. On one hand, this increased extent of LD could be beneficial for plant breeding as it would allow to use additive × additive effects more efficiently. On the other hand a reduced number of different allele combinations could also cause a reduction in selection response (Mackay and Gibson 1993).

Conclusions

The genetic diversity within varieties decreased monotonically during the past 50 years. Consequently, the diversity among varieties is of increased importance, but decreased slightly during the last time period. The major cause for this decrease is the multiple use of elite lines as parents for various hybrids. While the multiple use of lines enables breeders to optimally exploit their elite germplasm, it leads inevitably to a reduced diversity among hybrids.

One important criterion for the eligibility of newly bred varieties is that they are in at least one trait (disease resistance, yield, quality) better than the already registered varieties. Adding a criterion ‘positive contribution toward an increased genetic diversity of the available varieties’, which could be measured with molecular markers, would encourage innovative breeding. Thus, the suggested modification of the registration procedure and changes in the intellectual property practices encouraging plant breeders to evaluate new exotic diversity, could help to prevent a further potential erosion in genetic variation among commercial hybrids.