The recent history of durum wheat has been characterized by the replacement of landraces at first with selected pure line varieties and after with modern cultivars obtained by crosses between lines that did also include lines from mutagenesis, exotic germplasm, and different wheat species. Luckily ex situ collections preserve precious genetic resources collected over time that include germplasm from different genepools. Since in most cases the passport data do not include genetic information, their effective use in agriculture and breeding programs is very low.
The identification of useful genotypes is largely improved if the genetic structure of the germplasm collection is known. In the present study, we combined molecular genotyping and plant phenotyping with the comparison to well-known pedigreed varieties to describe the genetic structure of durum wheat germplasm collected in Southern Italy and preserved ex situ.
4.1. Diversity of South Italian Durum Wheat Germplasm
Overall, gene diversity for SSR markers was 0.607, and within the range of previous studies in durum wheat [9
]. Gene diversity and allelic richness within ex situ accessions were higher than within pedigreed varieties. This indicates a substantial loss of genetic diversity over time [48
] that might involve alleles valuable for plant improvement and future demands of producers and consumers. Moreover, the gene diversity within accessions and within varieties is representative of different genepools. In fact, most modern varieties, that are the results of hybridization and also include newly introduced germplasm cluster in group C2 (n
= 16) or in C1 (n
= 2), while most accessions cluster in four different groups (C3, C4, C5, and C6) that include “landraces” or varieties from pure line selection largely representative of the Italian–Mediterranean genepool. In ex situ accessions a small increase of diversity was observed from Group 1 (n
= 37 accessions collected from 1947 to 1950) to Group 2 (n
= 77 accessions collected from 1973 to 1982). Group 2 accessions, which were included in all clusters, were “old” and “new” germplasm grown at the same time in the same area at the time when intense varietal substitution was in progress. As expected, the amount of collected diversity decreased in the following time (n
= 22 accessions collected from 1983 to 2003) when most obsolete varieties had been already replaced. Clustering also showed that the choice of collecting sites was very careful to avoid the most recently introduced varieties. Out of 136 accessions, only 10 did cluster in C2 that included most “modern” varieties, and 23 in C1 which includes varieties derived from hybridization.
It has been argued that the selection pressure applied in breeding programs may have reduced the level of genetic diversity in durum wheat germplasm [49
]; in our study modern varieties (1974–2007) did show a degree of genetic diversity lower than in the “old and intermediate” group (1915–1973).
Our findings on the elite germplasm agree with previous reports by Medini et al. [50
], Reif et al. [48
], and Figliuolo et al. [11
], however, Maccaferri et al. [10
] did show a progressive increase of the genetic basis in the elite durum wheat germplasm. Moreover, Martos et al. [19
] and Laidò et al. [13
] observed that the overall molecular diversity of durum wheat remained quite constant throughout genetic improvement occurring during the 20th century. The reduction of diversity might be explained by the “Green Revolution”, which was characterized by breeding semidwarf varieties with a higher yielding potential due to an increased harvest index and better lodging tolerance, particularly under high fertilizer and water inputs [6
]. These new high-yielding semidwarf modern varieties were based on a limited number of founder genotypes and rapidly dominated the wheat germplasm base [48
]. This assumption may suggest that a high proportion of SSRs markers used in this study may be associated with chromosomal regions selected during breeding programs. These could be chromosomal regions harboring some agronomic relevant trait loci (such as those determining semidwarf habit or yield) that resulted in more uniform and stable modern varieties.
The now available high-density consensus maps of durum wheat coupled with sequencing information provide an advanced tool [52
] for durum wheat fingerprinting. Indeed, the increasing knowledge of polymorphic marker distribution in the durum wheat genome will facilitate the selection of SSR makers useful to detect pattern of diversity associated to different gene pools and will improve the use of ex situ collections in breeding programs designed to exploit opportunities offered by the wider wheat gene pool.
4.2. Analysis of Population Structure
Discriminant Analysis of Principal Components and STRUCTURE analysis were very informative and complementary about the structure of the durum wheat ex situ germplasm collection and were useful to identify the most diverse accessions for utilization in low input agriculture and in breeding programs. For instance, accessions were included in all clusters, whereas modern varieties from hybridization, including in their pedigree the innovative semidwarf CIMMYT materials, were assigned to only two clusters (C1 and C2). These findings are consistent with the repeated use of a few founder genotypes that played a relevant role in the creation of the genetic basis of modern genetic pools, in turn becoming progenitors of new elite varieties and completely replacing traditional varieties or landraces [5
]. The continued use of these genotypes made the gene pool smaller for all of the durum wheat elite varieties and resulted in the loss of genetic diversity [13
]. Low levels of genetic diversity of Italian cultivars developed by Italian breeders were also observed by Kabbaj et al. [18
]. The low lewel of genetic diversity can be explain as the combined result of frequent hybridization of a reduced number of founders and the strong selection pressure for the same trait needed for Italian growing conditions and the requiraments of the pasta industry [18
New variability is needed to face the challenges of modern low input agriculture. Accessions from the ex situ collection that are genetically distant, as those included in different clusters, could be useful for this purpose. Pedigreed varieties were useful to identify the gene pool of origin of the accessions that did cluster in different groups. It should be noted that the varietal purity was higher for those released more recently, while genotypes associated with older releases might be different from accessions with the same name. In some cases, Saragolla, a brand new variety that has been released in 2004 with an old name, or variety and accessions with the same name as “Russello”, might be selected as pure lines from the same “landrace” (Russello 329-S.97-S.G.7) [21
], or the variety Senatore Cappelli, that now is very popular in organic agriculture, is genetically distinct from the two ex situ accessions with the same name, that cluster in the same group. The accession variety name must be also carefully verified: the Capeiti 8 variety is included in C1 as expected according to the pedigree, while the accession named “Capeiti”, which was probably misnamed, did cluster in C2 and is genetically admixed. Also, the Trinakria variety (C2) and “Trinakria” accession (C1) were included in different clusters, but in this case, the mixed genotype was the variety. The Tripolino that we analyzed was probably not the “old” variety since it did cluster in C2 with most modern varieties while the synonymous variety Azizia did cluster in C5. Moreover, the STRUCTURE analysis revealed that it is actually an admixed genotype, thus suggesting that seeds of “pedigreed” varieties of “true” sources must be carefully tested for varietal purity. Two accessions and one variety with name Senatore Cappelli cluster in C3. The two accessions look fairly similar genetically and are also similar to Bidi that originated from the same “landrace”, while the now grown Senatore Cappelli variety is genetically distinct and shows some degree of admixture with the old germplasm from Sicily and with C1 and C2 where the most advanced varieties are included. It seems that evolution through natural hybridization took place for Senatore Cappelli and that the pure line, recently reselected for distribution to farmers, is actually different from the old one. These results were also confirmed by phenotypic data that show Senatore Cappelli variety is different from both accessions and from Bidi: the plant is shorter and has later heading, longer head, less fertile spikelets, smaller kernels, and fewer kernels per spike.
In a recent study [18
], a panel of 370 durum wheat genotypes, including 35 Italian genotypes (29 elite variety and 6 landraces), were genotyped using SNP markers. The Italian genotypes were attributed to seven subclusters. Three “landraces” did cluster with elite varieties indicating that these accessions were probably not “landraces” but cultivars that were not properly identified during the collecting missions [18
]. For the other three Italian “landraces” no information about their origin was available in passport data so a direct comparison with the accessions used in this study was not possible. Out of twenty-nine Italian elite varieties studied by Kabbai et al. [18
] only five were common to our study (Capeiti 8, Creso, Applo, Claudio, and Svevo). Capeiti 8, an “old” Italian variety from a cross between Syriacum × Mediterraneum groups (Eiti 6 × Cappelli), did cluster with lines and cultivars from the International Center for Agricultural Research in the Dry Areas (ICARDA) breeding program that include the old cultivar “Om Rabi” in their pedigree [18
]. Also in our study was Capeiti 8, which did cluster with “old” genotypes selected from indigenous and exotic landraces. The remaining four varieties were included in three different clusters in Kabbai et al. [18
], while in our study they did cluster in the same group with the “modern” dwarf genotypes released after 1974 that were selected from crosses that also included CYMMIT breeding lines.
Discriminant analysis of principal components can be also used to identify the alleles with the largest contributions to the discriminant functions, as an approach to detect putative patterns among the genes determining the group differentiation [40
]. A plot of SSR allele contributions was used to identify alleles of major interest and, substantially, most of them were detected at loci associated to quantitative trait locus (QTLs) for phenological traits, as well as to resistance to wheat diseases and to grain qualitative parameters. Indeed, when analyzing the alleles’ contributions to the first principal component, the gene Vrn-B1
, affecting the vernalization response, was found to be tightly linked to the marker Xgwm408
]. Noteworthy, the genes of the Vrn
system are the most important to determine the rate of generative development in wheat and, thereby, its flowering time [55
]. Additionally, a QTL for Fusarium head blight (FHB) resistance was mapped proximal to marker Xgwm282
], whereas the locus Xgwm234
was found to be significantly associated with dough strength [59
] and the marker Xgwm285
was located at only 2.1 cM from the tsn2
gene, conditioning resistance to race 3 of the fungus Pyrenophora tritici-repentis
(Died.), causal agent of tan spot in durum wheat and inducing necrosis [61
]. Since the inspection of alleles for the abovementioned markers showed that C1 and C3 were highly differentiated from C4 and C5, this could be an effective criterion for selection.
Similarly, when considering the alleles’ contribution to the second principal component, we noticed that the marker Xgwm193
encompasses the peak of the QGpc.ndsu.6Bb
QTL for grain protein content. Thus, it could be useful to use this SSRs marker to select for the high-GPC allele in Marker-Assisted Selection programs [62
]. Finally, the marker Xgwm518
was found to be linked to the QLr.caas-6BS.2
QTL for resistance to leaf rust caused by Puccinia triticina
] and the marker Xgwm413
to be associated to a QTL for grain weight on the short arm of chromosome 1B [65
]. These three markers differentiated C3 and C4 from C5, allowing to highlight the potential of the DAPC method to go beyond mere group delimitation and to identify accessions potentially useful in breeding programs.
Genetic groups identified by DAPC were also differentiated on the basis of morphological traits, as confirmed by PCA. Moreover, accessions and varieties assigned to the same cluster, although genetically related on the base of molecular markers, exhibited different morphological traits. These results could be helpful to select, especially from C3 and C4, early-heading landraces well adapted to the Southern Italian semi-arid conditions, where earliness allows farmers to better tackle the drought season [16
]. Within each of the six DAPC clusters accessions plants were taller than varieties. Plant height could be a preventive measure to face weed competition, which remains one of the main problems in organic wheat crops since no herbicides are allowed [66
]. Indeed, durum wheat landraces, characterized by tall plants and early vigor, could provide early groundcover which is vital to compete for light, water or nutrients. This trait represents a competitive advantage over early emerging weeds and their suppression [67
]. Finally, higher values for yield components in ex situ accessions in C3, C4, and C5 could be due to higher N-use efficiency and their ability to produce high yields at low soil-N availability typical in organic and low-input agricultural systems [68
]. Therefore, since wheat landraces have evolved mostly in environments with low nutrient availability [71
], they represent a source of variation for selection of varieties adapted to cropping systems with low fertilizer input.