SNP-based Genetic Structure of Tunisian Durum Wheat Landraces and Distribution of the Resistance to Septoria Tritici Blotch

Background: Septoria tritici blotch (STB) has marked durum wheat production worldwide. This fungal disease is until today a challenge for farmers, researchers and breeders all united in the aim of reducing its damage and improving wheat resistance. Tunisian durum wheat landraces were reported to be valuable genetic resources for resistance to biotic and abiotic stresses and are therefore prominently deployed in breeding programs to develop new varieties adapted to fungal diseases as STB and to climate change constraints overall. Results: A total of 366 local durum wheat accessions were assessed for resistance to two virulent Tunisian isolates of Zymoseptoria tritici Tun06 and TM220 under eld conditions. Population structure analysis of the durum wheat accessions, performed with 286 polymorphic SNPs (PIC >0.3) covering the entire genome, identied three genetic subpopulations (GS1, GS2 and GS3) with 22% of admixed genotypes. Interestingly, all of the resistant genotypes were among GS2 or admixed with GS2. Conclusions: This study revealed the population structure and the genetic distribution of the resistance to Z. tritici in the Tunisian durum wheat landraces. The grouping pattern of accessions appear to be associated, to some extent, with the geographical pattern of the landraces. We suggested that GS2 accessions were mostly introduced from eastern Mediterranean populations, unlike GS1 and GS3 that originated from the west. Resistant GS2 accessions belonged to landraces Taganrog, Sbei glabre, Richi, Mekki, Badri, Jneh Khotifa and Azizi. Furthermore, we suggested that admixture contributed to transmit STB resistance from GS2 resistant landraces to initially susceptible landraces such as Mahmoudi (GS1), but also resulted in the loss of resistance in the case of GS2 suscpetible Azizi and Jneh Khotifa accessions.

durum landraces favored their divergence into genetically distinct groups. In fact, Moragues et al [9] classi ed a collection of 63 durum wheat landraces from the Mediterranean basin in a north dispersal group (including south European and south-western Asian landraces) and a south dispersal group (including landraces from North Africa and the Iberian Peninsula). Soriano et al [11] also structured a collection of 172 durum wheat landraces from 21 Mediterranean countries into four genetic populations related to their geographical origin, namely eastern Mediterranean, eastern Balkans and Turkey, western Balkans and Egypt, and western Mediterranean.
Durum wheat cultivation history in North Africa and particularly Tunisia, involved the intervention of Phoenicians importing wheat from Lebanon to Carthage, along with the development of Carthage trade maritime activity in the Mediterranean Sea favoring seed exchanges between Tunisia and the Mediterranean countries. North African landraces were also introduced and diffused by Romans, who greatly in uenced durum wheat cultivation in this area by setting up modernized irrigation systems [5,12,13]. Recently, Robanna et al [13] studied the structure of six Tunisian durum wheat landraces, and reported their genetic similarity with landraces from North African countries and landraces from Greece, Italy, and Lebanon. Durum wheat was prevalent and well established in North Africa in the classical times [5]. Accordingly, North Africa and Abyssinian regions are considered as secondary centers of diversity for durum wheat [14,15]. Ren et al [16] also suggested that North Africa should be considered as a microcenter of wheat diversity. Tunisia, being part of the secondary center of diversity for durum wheat, has a rich collection of old local durum wheat landraces [14]. To date, around 40 old durum wheat landraces are known in Tunisia that were morphologically characterized and classi ed by Boeuf [14] and Dghais et al [17]. Several studies emphasised high levels of genetic and agro-morphological diversity [13,[18][19][20][21][22], phenological features [23][24][25][26] and resistance to biotic and abiotic stresses i.e. drought, heat and fungal diseases [21,[27][28][29] within the Tunisian old durum wheat germplasm. Recently, Ben Krima et al [30] showed a complex structure of 14 Tunisian durum wheat populations that was not entirely related to their geographic origin and variety name. However, Ouaja et al [22] identi ed a strong correlation between the genetic structure of 11 Tunisian durum wheat landraces and their morphological characterisation and nomenclature. Therefore, various interacting factors were reported that have in uenced the structure and evolutionary dynamics of durum wheat in Tunisia and the Maghreb region overall, among which complex selection trajectory, the signi cance of variety names, the occurrence of heterogeneous mixtures within populations, local adaptation, local and regional exchanges between farmers and, loss and misidenti cation [22,30,31]. After the Green revolution, old durum wheat landraces were mainly grown and managed by smallholder farmers under low-input traditional agrosystems in the marginal areas of Mediterranean region, notably in southern Europe and North Africa [4,32] as they were progressively abandoned from the early 1970s and replaced by improved genetically uniform modern varieties/cultivars [33,34]. In Tunisia, durum wheat landraces are still cultivated by low-input farmers, in northern and central mountainous areas, under traditional farming systems. These landraces, transmitted by farmers from one generation to the next, are designated by a variety name linked to a historical origin, regional location and speci c phenotypic characteristics [22,30]. Septoria Tritici Blotch (STB) caused by the fungus Zymoseptoria tritici (Z. tritici) (Desm.) (Formerly Mycosphaerella graminicola) is currently considered among the most damaging and worldwide distributed fungal disease on cultivated wheat [35,36]. The appearance of Z. tritici as a host-specialized wheat pathogen occurred about 10,500 years ago via host tracking throughout the time of wheat domestication [37][38][39]. Comparative genomic analysis between Z. tritici and its close relatives highlighted strong adaptive evolution of Z. tritici in relation to specialization on wheat [40]. Howbeit, Stukenbrock et al [37] demonstrated that wheat-adapted Z. tritici was derived from an ancestral population infecting wild grasses in the Middle East and that the domestication of an agricultural crop was concomitantly accompanied by the domestication of a fungal pathogen. Accordingly, the Fertile Crescent is considered a hotspot of Z. tritici genetic diversity [37,41]. Therefore, wheat landraces and their wild relatives from the Fertile Crescent, having co-evolved for a long time with Z. tritici, must harbor the greatest diversity for resistance to STB [42]. Moreover, a host species specialization was highlighted in Z. tritici populations to either bread or durum wheat [43][44][45] making bread wheat-derived Z. tritici strains not suitable to decipher Stb genes in durum wheat, as studies into the genetic basis of STB resistance were entirely based on the Z. tritici-bread wheat pathosystem [46][47][48][49][50][51].
In Tunisia and under suitable environmental conditions for infection, STB causes considerable yield losses up to 50-60% [52,53]. The introduction of the modern cultivar 'Karim' in 1980 displaced the cultivation of landraces, resulting in a reduction of the genetic diversity (genetic erosion) and therefore, enhancing the susceptibility to STB [54,55]. Durum wheat landraces, characterized by a substantial level of genetic diversity [20,22,32], represent the main sources of resistance to Z. tritici to be incorporated into breeding programs for a sustainable STB disease control and management [21,56,57]. The present study consisted on a) analyzing the genetic structure of 366 Tunisian durum wheat accessions belonging to 13 landraces using 286 single nucloetoide polymorphic markers (SNPs), b) evaluating the resistance of these accessions to two durum-wheat derived Z. tritici isolates (Tun06 and TM220) and, c) relating genotyping and disease screening data of the Tunisian durum wheat accessions to describe the life history of STB resistance in local durum wheat landraces.

Wheat accessions
A set of 375 durum wheat accessions was used in this study. Accessions were collected by the National Gene Bank (BNG) from four regions in Central (the Sahel and Kairoun) and southern (Gabes and Medenine) Tunisia (Table S3). Accessions were morphologically characterized, identi ed and classi ed into 13 landrace-populations namely Mahmoudi, Biada, Bidi, Biskri, Azizi, Badri, Mekki, Jneh Khotifa, Sbei glabre, Taganrog, Richi, Souri and Roussia as described by Ouaja et al [22]. Information about the origin and year of introduction and selection of these landraces are depicted in Table S4.
Field trials and Z. tritici isolates Durum wheat landraces were screened for their resistance to Z. tritici at the adult plant stage in the experimental station of CRRGC Beja in northwest Tunisia. Inoculation assays were performed using two durum wheat-derived Z. tritici isolates on separate eld experimental plots; notably the well-characterized and virulent reference isolate Tun06 (also reported as TunBz-1) sampled from the Bizerte region of Tunisia in 2006 and the isolate TM220 collected from Lansarine region of Tunisia in 2014 [21,56,83].
Tun06-and TM220 Field trials were realized as reported by Ouaja et al [21] and followed an Augmented Randomized Complete Block Design (ARCBD), including 6 blocks spaced apart of 1 m. Each block was composed of 70 accessions and 7 checks. The checks included six susceptible to moderately resistant modern durum wheat cultivars "Karim, Khiar, Om Rabia, Salim, Maali and Nasr" [53,54] . Durum wheat landraces were evaluated for resistance to Tun06 isolate during two consecutive growing seasons (2015-2016 and 2016-2017), and for resistance to isolate TM220 on the growing season of 2016-2017.
Inoculation and screening for resistance to Z. tritici Tun06 and TM220 isolates were maintained frozen at -80°C. A preculture of the inoculum was prepared by shaking Z. tritici isolates (at 100 rpm/ 25°C) for six to seven days, in 100 ml yeast glucose liquid medium (30g glucose, 10g yeast per liter demineralized water). The produced spore suspensions of Tun06 and TM220 were subsequently transferred to 500 ml yeast glucose liquid media and were incubated under the aforementioned conditions to provide su cient inoculum for the eld trials. Spores of both isolates were collected after overnight settling in static cultures, concentrated by decanting the supernatant medium, and were adjusted to 10 6 spore/ml. Accessions in all experimental plots were inoculated twice, at the three-leaf stage (approximately GS21) and at the stem elongation stage (approximately GS37) [84] as described by Ouaja et al [21].
The disease severity was evaluated by estimating pycnidia coverage percentages. The same leaf layer (F3-F4) of each accession in all experimental plots was assessed for STB resistance, at three time points for isolate Tun06 and at two time points for isolate TM220. Tun06-disease scores were assessed at 18, Three classes of response to Z. tritici at the adult plant stage, de ned by Ouaja et al [21], were considered in this study. The resistant class (R) include accessions having pycnidia score <25%, while the susceptible class (S) include accessions showing >47% of pycnidia on infected leaves. Accessions with pycnidia score ranging between 25% and 46% constitute the class of intermediate (I). About 4% and 22.4% of accessions had missing data for Tun06 and TM220, respectively.
Genotyping and data analysis A total of 20,279 polymorphic SNPs were generated by Illumina sequencing 375 Tunisian durum wheat accessions using a High-density 90K wheat SNP array (iSelect, San Diego, USA) [60]. SNPs having < 5% of missing data and a minor allele frequency of 5% were ltered. Map position of the markers was derived from the consensus durum wheat genetic map [85]. From these markers, 286 SNPs covering the entire genome with 12 to 23 SNPs per chromosome of 10 cM and with a PIC > 0.3, were selected to perform the analysis in the present study (Table S5). PIC values were calculated by determining the frequency of alleles per locus as Powell et al [86] (Table S6). Genotypes with > 10% of missing data out of the 286 SNPs were eliminated, and 366 genotypes were included in the rest of the analysis.
The genotypic data based on the 286 selected SNP markers were used to study the population structure of the Tunisian durum wheat accessions with the model-based clustering algorithm STRUCTURE 2.3.4 [58]. STRUCTURE program was run with K values between 1 and 20, by applying 10 independent runs for each value of K, 100,000 burnins and 100,000 Markov Chain Monte Carlo (MCMC) repetitions. The optimal number of genetic subpopulations was determined using the mean posterior probability (ln P(D)) value per cluster (K) and the delta-K method of ln P(D) under STRUCTURE harvester 0.6.9.4 [59]. Population structure was investigated at the optimal K as well as at K ranging from 13 to 20 in order to assertain the genetic grouping according to landrace denomination. Individuals presenting a membership coe cient less than 0.7 to any genetic subpopulation were considered as admixed genotypes. To study the phylogenetic relationships between accessions/landraces, an Unweighted Pair Group Method with Arithmetic Average (UPGMA) tree was created using Nei standard genetic distance [87]. A bootstrap analysis was performed with the program Populations 1.2.32 [88] and branch support values were estimated using 1000 bootstrap randomizations. A consensus tree was generated and viewed using TreeView 1.6.6 [89].
In addition, a pairwise Fst and gene ow (Nm) coe cients were calculated between the genetic subpopulations with GenAlEx 6.501 [90]. An analysis of molecular variance (AMOVA) was also carried out using GenAlEx 6.501 [90] in order to investigate the signi cance of genetic differentiation within and among genetic subpopulations (as de ned by STRUCTURE). A Principal Components Analysis (PCAs) on the genotypic data was performed under GenAlEx 6.501 [90]. This multivariate analysis was completed to visualize the structure of the 366 durum wheat accessions sorted by genetic subpopulations (as de ned by STRUCTURE), by landrace denomination [22] and by STB resistance class to Tun06 (R, I and S).
An analyses of variance (ANOVA) were carried out under R 3.3.2 [91] to determine the effect of genetic subpopulations and isolates on RAUDPC scores. Genetic subpopulations x isolates interaction were also investigated.
PCAs were performed using the 286 SNPs on the 366 genotypes ( Figure 2). Axes 1 and 2 of the PCAs accounted for 24.33% and 16.54% of the total genetic variation, respectively. Figure 2a showed that PCA grouping corroborated the genetic structure inferred by STRUCTURE, pointing a clear differentiation between GS1, GS2 and GS3. Admixed genotypes were essentially distributed between GS1 and GS2 and between GS2 and GS3, re ecting ongoing hybridization and allele exchanges between these groups. Pairwise Fst values also showed considerable genetic differentiations between GS1, GS2 and GS3 ( Table 1). The highest Fst value (0.751) was observed between GS1 and GS3. Fst values between GS1 and GS2 and between GS2 and GS3 were both aroud 0.400. Similarly, Nm indices between GS1 and GS2 and between GS2 and GS3 were close; while Nm between GS1 and GS3 was the lowest (0.078) re ecting an almost absence of genetic exchange between these two populations. The AMOVA ( Table 2) was consistent with pairwise Fst and Nm analysis revealing that the genetic variation between subpopulations (61%) was higher than the variation within subpopulation (39%).  Table S1).
In addition, population structure was investigated from K=13 (corresponding to the number of landraces studied) to K=20 (Table S2). At K=13, landraces Azizi, Badri, Biada, Bidi, Jneh Khotifa, Mekki, Richi and Sbei glabre were assigned to separate genetic subpopulations. However, landraces Souri and Roussia were grouped in the same genetic subpopulation. Landraces Mahmoudi and Biskri were both divided into two genetic subpopulations. Taganrog landrace was entirely composed of admixed genotypes (between Sbei glabre, Jneh Khotifa, Azizi and Badri). The STRUCTURE assignment that matched exactly the landraces denomination was obtained for K=15. Indeed, at K=15, all 13 landraces were discriminated and assigned to genetically distinct subpopulations, except for Mahmoudi and Biskri, which were both divided into two genetic subpopulations as for K=13. In fact, Mahmoudi subpopulations correponded to two morphological types of Mahmoudi namely Mahmoudi-122 and Mahmoudi-986. Biskri subpopulations corresponded to two morphological types of Biskri namely Biskri-Ac1 and Biskri-glabre ( Figure S1).

Distribution of the resistance to Zymoseptoria tritici among subpopulations
The 366 durum wheat accessions were evaluated for their resistance to two Tunisian Z. tritici isolates Tun06 and TM220. Based on disease scoring, the landraces were categorized into three classes of response to Z. tritici (R, I and S) as de ned by Ouaja et al [21]. Overall, 27%, 15% and 55% of the collection were R, I and S genotypes to isolate Tun06, respectively. In addition, 28%, 7% and 43% of the collection were R, I and S genotypes to isolate TM220, respectively.
The ANOVA (Table 3) revealed that the genetic subpopulation effect is highly signi cant, indicating that the variation in the responses to infection by Z. tritici isolates relied on the population genetic structure. However, a non-signi cant effect of isolates on the RAUDPC scores was revealed. Genetic subpopulations x isolates interaction was also non-signi cant. Accordingly, isolates Tun06 and TM220 did not signi cantly varied in their severity (RAUDPC scores) towards wheat accessions. Therefore, the rest of the analysis were persued with solely Tun06 isolate, a reference isolate with the lowest missing data.
The distribution of R, I and S classes by genetic subpopulation was uneven (Figure 2.c, Table S1). GS1 and GS3 were mainly formed by accessions of class S. In fact, 93% and 85% of the accessions were susceptible to Tun06 within GS1 and GS3, respectively. Interestingly, 44% and 31% of resistant accessions to Tun06 belonged to GS2 and admixed genotypes with GS2, respectively. Genotypes with intermediate responses were also mainly among GS2 or admixed with GS2. While GS2 comprised 35% of susceptible accessions. About 44% of admixed genotypes (mostly G1-G2 and G2-G3) were susceptible to Tun06.

Discussion
Old durum wheat germplasm represents a precious genetic heritage. Understanding the genetic and phenotypic structure of old local landraces will help retracing their life history of struggle for resistance and survival over time. Durum wheat population structure study also help deciphering new sources of resistance to cope with challenging abiotic and biotic stresses, notably STB, one of the most devastating fungal disease on durum wheat crop. In the present study, we genotyped 366 Tunisian durum wheat accessions belonging to 13 old Tunisian landraces [22], collected from three central and two southern regions in Tunisia, using 286 SNPs derived from a High-density 90K wheat SNP array [60]. This study revealed the population structure and the genetic distribution of the resistance to Z. tritici in the Tunisian durum wheat landraces.
Identi cation of a population structure in the Tunisian durum wheat landraces related to their potential introduction pathways The genetic structure of 366 durum wheat accessions was investigated using 286 SNPs. Three major genetic supbpopulations (K=3: GS1, GS2 and GS3) were obtained under STRUCTURE [58]. At K= 15, we were able to attribute each landrace to a distinct genetic group, with the exception of Mahmoudi and Biskri landraces that were both divided into two groups. This result is in agreement with our previous study [22], where 8 out of 11 Tunisian durum wheat landraces corresponded to distinct genetic groups using 10 SSR markers. AMOVA analysis showed a high genetic variability (61%) between subpopulations, suggesting that the three subpopulations were derived from different gene pools. This is consistent with previous studies of tetraploid wheats [2,61] and barley [62] showing that local landraces were derived from multiple ancestral populations and had reticulated phylogenetic relationships. However, Soriano et al [32] detected only 13% of genetic varibility between four genetic subpopulations that were tracing distinct geographical pattern of the Mediterranean durum wheat germplasm. The formation of distinct domesticated gene pools were also reported for several other crops such as common bean, which underwent parallel evolution and spread further through the development of landraces with distinct characteristics and speci c adaptations [63-66].
Informations about putative origins and years of introduction in Tunisia of the 13 herein studied landraces, along with their agro-morphological characteristics as described by Ouaja et al [22], were analyzed to explain their inferred genetic structure. In fact, the grouping pattern of accessions appear to be associated, to some extent, with the geographical pattern of the landraces. In this study, the subpopulation GS2 mainly included landraces of North African and East Mediterranean origins. GS2 includes landrace Mekki from Morocco, landrace Taganrog from Cyprus, landraces Azizi, Jneh Khotifa and Sbei glabre that were considered as local populations and Richi which was reported as foreign [17,34,67]. Robbana et al [13] also reported that Tunisian landraces were genetically associated to North African landraces. However, using a set of DArtSeq markers to describe the genetic diversity of Tunisian landraces, Robbana et al [13] reported that Jneh Zarzoura, a close relative to Jneh Khotifa [17] clustered distinctly with accessions from Jordan. Moreover, local landraces names have been traditionally selected by natives, generally according to morphological features or locality, and are often consciously used by farmers for management, selection or exchanges [14,68]. Sahri et al [68] particularly highlighted the signi cance of variety name, which have largely in uenced the structure and evolutionary dynamics of durum wheat in Morocco. Therefore, nominal analogies were frequently reported between landraces of different Mediterranean regions, probably re ecting trade and migration of the same landraces around the Mediterranean basin. Xynias et al [6] reported that the Italian cultivar « Senatore Capelli » was selected in 1915 from the local North African landrace « Jean Retifah » which was very prominent and marked the cultivation of durum wheat in Italy. The landrace « Jean Retifah » must be the known Jneh Khotifa in Tunisia [14,17]. In addition, Soriano et al [32] studied the structure of durum wheat landraces from 21 Mediterranean countries, using SSR markers, and reported two Italian landraces named « Hymera » and « Aziziah » that were associated to eastern Mediterranean genetic group. Knowing that Tunisia was the former bread basket of the Roman Empire [69,70], these latters could be the landraces known Hmira and Azizi in Tunisia [14,17,20,22], suggesting though that Azizi landrace may also have an eastern origin. Nevertheless, Boeuf [14] mentioned that the landraces of North Africa had dominant characters, speci c to Abyssinian wheats, such as red, purple or black spikes, pubescent glumes and red or dark-colored grains. These characters were totally absent and unknown in Europe. Herein it was noted that landraces of GS2, with the exception of Biskri, shared features of spikes and grains [22] corresponding to the Abyssinian wheats as Boeuf [14] reported.
Furthermore, according to Boeuf [14], the geographical expansion and domestication of wheats from Abyssinia enhanced the accumulation of recessive characters, in particular white spikes, hairless glumes and light-colored grains widely adopted in Europe. During crop domestication process, several changes were induced for major morphological, structural and functional traits associated with adaptation and cultivation in order to meet human needs, as reported for the common bean (Phaseolus vulgaris) [66]. Boeuf [14] mentioned that the whiteness of the spike and light-colored grains were among the most sought after and preferred criteria in wheat by European farmers during commercial trade in North Africa. So far, these traits are characteristics of the landraces Mahmoudi (GS1), Biskri (GS2), Biada and Bidi (GS3) [22]. Landraces Biskri, Biada and Bidi were introduced in Tunisia from Algeria, while Mahmoudi was considered as a local landrace population with various reported origins including Algeria and Italy. Another example re ecting nominal analogy between Mediterranean landraces is the Tunisian landrace Biada and the Spanish landraces « Blancal » and « Blanco de Baleares » meaning, among others, the white wheats [17,22,71], thus indicating that such phenotypic characteristic was probably derived from western Mediterranean. Based on these ndings/statements, although the majority of Tunisian landraces are North Africa, we suggest that subpopulations GS1 and GS3 were introduced to North Africa and particularly to Tunisia from Europe/western Mediterranean, unlike subpopulation GS2 which may be originated from the Middle East. Ben Krima et al [30] also agree that the combination of both genetic and agro-morphological approches are essential for retracing the history, origin and dynamic lifestory of Tunisian durum wheat landraces. In the same context, Moragues et al [9] highlighted two dispersal pathways of the Mediterranean durum wheat landraces which had contributed to the divergence of these landraces into distinct genetic groups following their adaptation to different local environments. A rst pathway through the North-East of the Mediterranean basin to Europe, and a second pathway through the South of the Mediterranean basin to North Africa reaching the Iberian Peninsula.
Genetic distribution of the resistance of Z. tritici in the Tunisian durum wheat populations Tunisian durum wheat landraces have been reported to exhibit valuable sources of resistance to STB, useful to include in breeding programs and to develop varieties with durable and broad spectrum of resistance [21,56,57]. In the present study, the 366 genotyped durum wheat accessions were also screened for Z. tritici resistance, under eld conditions, using two Z. tritici isolates Tun06 and TM220 collected from two Tunisian regions, Bizerte and Manouba, respectively. The analysis of variance showed a non-signi cant variation between isolates Tun06 and TM220 towards the durum wheat accessions, suggesting that they likely have similar virulence pro l. This result agrees with Ferjaoui et al [56] ndings, detecting only two virulence pro ls among 55 Tunisian Z. tritici isolates screened at seedling stage.
Overall, 60% of the accessions were susceptible, showing that Tun06 and TM220 isolates were virulent on the majority of the accessions, which re ect the adaptation of Z. tritici virulence to durum wheat landraces in Tunisia. Likewise, Ouaja et al [21] suggested that the Tunisian Z. tritici isolate Tun06 still preserve virulences against old durum wheat landraces even though they are currently marginally grown in wheat production areas in Tunisia. In addition, several studies reported that Z. tritici undergoes frequent sexual reproduction on durum wheat in Tunisia [72,73]. In fact, sexual reproduction plays a key role in the evolution of pathogenicity traits, including virulence and aggressiveness [74], enabling the fungus to quickly evolve and circumvent the resistance genes by creating new combinations of alleles and, in combination with the asexual reproduction allowing frequent generation of the new genotypes [42,73]. Nevertheless, about 27% of the collection was resistant, suggesting that Tunisian durum wheat landraces still carry effective STB resistance genes.
The analysis of variance showed a large and signi cant variation between GS1, GS2 and GS3 regarding the resistance to isolate Tun06. Subpopulations GS1 (composed of Mahmoudi accessions) and GS3 (Biada accessions) showed higher frequencies of susceptible responses than GS2. The distribution of the resistance within each landrace observed in the UPGMA tree, indicate that the resistance relies on the landrace instead of the genetic structure (K=3), as GS2 grouped both resistant and susceptible genotypes. These results also indicate that although landraces of GS2 formed an invidivualized genetic group, they may harbor combination of resistance genes that differ in nature, number, structure, chromosomal localisations and type of interaction that still need to be depicted and revealed by a genome wide association study (GWAS).
The susceptibility found within GS1, GS3 and Biskri and Azizi of GS2 genotypes is due to large-scale use and commercial share of these landraces, which were extensively cultivated especially in northern Tunisia where they have lost resistance [14,17]. In fact, Z. tritici is mainly prevalent on durum wheat in northern and northwestern regions of Tunisia, where sub-humid zones represent hot spots of STB disease with high infection levels [52,54,75] . The landraces that were widely grown in northern zones ended up being susceptible as a consequence of the rapid adaptation of Z. tritici isolates [42,49,51,76]. Moreover, rapid adaptation of Z. tritici to landraces from western mediterannen origin (GS1 and GS3) could be facilitated by domestication process mainly accompanied by a strong reduction in genetic diversity and/or high levels of gene loss compared to wild ancestors or wild gene pool [77][78][79]. These events reduced the adaptation of cultivated wheat to erratic environmental variations, where wild traits show much greater tness over domesticated ones [66]. This senario might be consistent with the signi cant variability of the resistance to Z. tritici observed between subpopulations in the present study, suggesting that resistance (R) genes of the three subpopulations may have evolved divergently and crucially under a combination of environmental and human pressures. Therefore, we hypothesize that major genes conferring speci c resistance (as dominant characters) have undergone modi cations and alterations by mutations during the geographic expansion and with the wide exchange network of durum wheat landraces among Mediterranean regions, which resulted in the loss of dominant R genes/alleles and the spread of susceptibility as observed within landraces of subpopulations GS1 and GS3. Alternatively, the results suggest that landraces of GS1 and GS3 were initially susceptible to Z. tritici when introduced to Tunisia. Nevertheless, unlike GS1 and GS3, most landraces of GS2 (except for Biskri and Azizi landraces) were resistant because they had recourse to different geographical pattern/pathway as they were local or directly derived from an eastern origin and did not pass via Europe. Western Europe farmers were among the rst to create and adopt modern methods of plant breeding and exert wheat genetic improvement, involving direct selection for homogeneous material, thus, indirectly reducing the variability of the genetic sources of resistance or even unintentionally selecting for recessive genes [14,80,81]. In contrast, smallholder farmers in North Africa have been preserving the local durum wheat diversity with on-farm conservation practices over generations [14,17,22,68].
In this study, 22% of the accessions were admixed, among which 51% of admixed genotypes between GS2 and GS1 and 24% admixed between GS2 and GS3. Admixtures occurs mainly by gene ows, through the frequent introduction of new genotypes into elds and seed exchange network within and between farmer communities [2,16,32]. In fact, Nm between GS1 and GS2 and between GS2 and GS3 were both aroud 0.4. All the resistant accessions of Mahmoudi were admixed between GS1 and GS2 and all the resistant accessions of Biada, Souri and Roussia were admixed between GS3 and GS2; suggesting that resistant landraces in GS2 were probably the sources transmitting resistance to Z. tritici via admixture. Indeed, resistant Mahmoudi genotypes may have acquiered their resistance from Jneh Khotifa, as they appeared phylogenetically close in the UPGMA tree. On the other hand, admixture may have also caused loss of resistance as some admixed Jneh Khotifa and Richi accessions were susceptible to Tun06. This result indicated that admixture between genetically distinct landraces/populations and frequently recurring gene exchanges (or gametic association between gene loci) [82], may have elicited susceptibility within initially resistant accessions throughout an alteration or loss of the resistance genes/alleles.

Conclusion
The present study revealed that Tunisian durum wheat life history of resistance to STB involve the interaction of miscellaneous factors, including the landrace genetic structure and introductory pathways, the local commercial share de ning the geographic and regional distribution of the landraces and the occurrence of admixtures within these landraces. In fact, landraces were subjected to genetic differentiation in time and space during their introductory pathways in the Mediterranean area, their adaptation to speci c environments and the Human selection pressure/domestication, contributing to their divergence in their resistance spectra. In addition, the regional distribution of the landraces across Tunisia in uenced the evolutionary history of resistance genes under different climatic constraints in relation to hotspots of STB disease. Admixtures also represent one of the main driver factors of the resistance to Z. tritici involving old or ongoing exchanges that contributed to the introgression and/or the loss of resistance genes. Our results suggest the presence of diverse resistance sources towards two Z. tritici isolates Tun06 and TM220 in the Tunisian durum wheat landraces, which implies the need for more in-depth research to investigate and characterize the related resistance genes. Continuous identi cation of new sources of resistance to STB is required for the development of wheat cultivars with sustainable eld resistance. Therefore, old local durum wheat landraces can play an important role as donor of resistance genes in breeding programs. In addition, further investigation and study of virulence patterns in Z. tritici populations in Tunisia will provide new insights and understanding of the Z. tritici-durum wheat interactions and trace their co-evolution.       Table S1 and are color-coded by resistance class to Z. tritici isolate 'Tun06'. Branches are color-coded according to STRUCTURE results at K=3. Three subclusters were identi ed and labeled SC-I, SC-II and SC-III.