Morphological characterization and Genetic diversity analysis of a Tunisian durum wheat (Triticum turgidum var. durum) collection

Background: Tunisia is a center of genetic diversity of durum wheat and has a large number of abandoned old local landraces. An accurate investigation and characterization of the morphological and genetic features of these landraces would allow their rehabilitation and use for practical and benecial purposes. In this context, a collection of 304 local accessions of durum wheat, collected from ve regions and three climatic zones of central and southern Tunisia, was studied. Results: Morphological characterization was carried out using 12 spike-related traits and rendered a mean Shannon-Weaver Index (H') of 0.80 indicating the presence of a high level of polymorphism among accessions. Based on these traits 11 local landraces, namely Mahmoudi, Azizi, Jneh Khotifa, Mekki, Biskri, Taganrog, Biada, Badri, Richi, Roussia and Souri were identied. Spike length (H’=0.98) and shape (H’=0.86) with grains size (H’=0.94), form (H’=0.87) and color (H’=0.86) were the most polymorphic morphological traits. The genetic diversity was assessed using 10 SSR markers, with a polymorphic information content (PIC) of 0.69. Levels of genetic diversity were generally high, with a Shannon's Information Index (I) of 0.62 and a gene diversity (He) of 0.35. In addition, population structure analysis distinguished 11 genetic subpopulations signicantly correlated with the morphological identication. Analysis of molecular variance (AMOVA) showed high genetic variations within regions (81%) and within wheat subpopulations (41%) reecting a considerable amount of admixture between landraces. The moderate (19%) and high (59%) genetic variations among regions and among wheat subpopulations observed highlighted farmers selection practices . Furthermore, Mahmoudi landrace showed spike densities signicantly different between the center to the south of Tunisia; notably loose spikes with open glumes in the south and compact ones in the center, which may represent an adaptation form for tolerance to high temperature. Conclusion: Overall, this study underlined the genetic richness of local resources for better in situ or ex situ conservation and for their subsequent use in plant The morphological diversity was higher than the previously described for Tunisian durum wheat germplasm (H’=0.53) of 930 accessions collected from a reduced number of sites from southern Tunisia, using twenty-two qualitative and three quantitative traits [16]. Lower phenotypic diversity was also observed for Moroccan durum wheat populations composed of 101 landraces (H’=0.62) using nine agro-morphological traits [24] and of 59 traditional durum wheat (H’=0.78) using nine agro-morphological traits [25]. Ethiopian durum wheat populations composed of 32 landraces had an H’ index of 0.71 using 8 qualitative traits [26], while Oman populations composed of 161 accessions showed H’ index of 0.52 and 0.66 using 15 qualitative and 17 quantitative characters respectively [26]. to a protocol established by Euron (https://www.eurons.fr). DNA amplication was performed by preheating the DNA at 95°C for 5 mn, followed by 35 cycles of 95°C for 30 s, 60°C for 90 s and 72°C for 30 s, with a nal extension step of 60°C for 30 mn. PCR products were checked for amplication on a 2% agarose gel and fragments were separated according to their size on an ABI Prism Genetic Analyzer (Applied Biosystems). Data was checked again using Peak scanner software version 1.0. Two accessions had missing data for all used SSR and were discarded from our study.


Morphological characterization of the Tunisian durum wheat accessions
Phenotypic diversity and morphological characterization The Shannon-Weaver index (H') revealed a high morphological diversity among the durum wheat accessions with an overall H' of 0.80 (Table 1). The most polymorphic characters were the spike length (H'=0.98), the grain size (H'=0.94), grain forms (H'=0.87), the grain color (H'=0.86) and the spike shape (H'=0.86), while the spike color showed the least polymorphic level (H'=0.53).  The 304 durum wheat accessions were grouped into eleven landraces namely Azizi, Jneh Khotifa, Taganrog, Mekki, Richi, Souri, Roussia, Badri, Biskri, Biada and Mahmoudi, according to the catalog of old durum wheat landraces and are part of the 40 durum wheat landraces recorded in Tunisia [10]. These landraces were characterized by the 12 speci c morphological traits based on IPGRI [20] and UPOV [21] (Table S1, Table S2). A multitude of spike characteristics has been observed between the durum wheat landraces, whereas these characteristics were homogeneous between accessions of the same landrace. In fact, the Shannon-Weaver index (H') calculated for each landrace were relatively low, ranging between H'=0.00 for Badri and Jneh Khotifa landraces and H'=0. 23 for Richi landrace with an overall mean H' of 0.11 (Table S3). For instance, the variety Mahmoudi accessions had particularly large spikes with sub-pyramidal shape, very long awns and big grain size; whereas rectangular and very at spikes characterized Azizi accessions. Biskri accessions had fusiform and big size spikes. The Spike color, length and shape were variable among the studied accessions and varied from dark to light, and from short to long spikes For example, Badri spike was very short and thick with a greyish color, while Biada was characterized by very light (white) spikes and awns color. Souri and Roussia were both characterized by tight and red colored spikes with distinct spike shape characterized as rectangular for Souri and cylindrical for Roussia. The former varieties were also characterized by a distinct orange grain color. Interestingly, Richi accessions had a unique feathery spike, while Mekki was characterized by short and dense spikes with parallel edges. Finally, Taganrog accessions are characterized by white colored spikes washed with black, while Jneh Khotifa accessions had a very dark (black to purple) long and dense spike and awn colors.

Principle Coordinates Analysis (PCoA)
The PCoA performed on the 12 spike morphological traits of the 304 durum wheat accessions showed that axes 1 and 2 accounted for 25.73% and 22.34% of the total genetic variation, respectively ( Figure 1). The rst axis was mostly associated with Spike Shape (SS), Spike Length (SL), Number of Spikelets/spike (NS), Grain Color (GC) and Awns Length (AL). While the second axis was mainly de ned by Grain Form (GF), Size (GS) and Number/spikelet (GN) (Figure 1, a).
Color-coding of the accessions in the 2-dimensional PCoA plot (axis 1 vs. 2) showed a good correspondence between the morphological grouping and the landraces denomination (Figure 1, b). In fact, accessions belonging to the same landrace were included in the same PCoA subgroup. Biskri, Jneh Khotifa and Taganrog accessions were agglomerated and positively correlated to both axes and shared similar spike characteristics such as spike length (mostly long spikes) with a high number of grains per spikelet (>3) and similar black awns longer than the spike. Azizi accessions were grouped into a distinct subgroup mainly characterized by rectangular medium-sized spikes with a tan color. Mahmoudi accessions also formed a distinct subgroup that was mainly characterized by unique pyramid shaped spikes. Accessions of Souri and Roussia formed almost a single subgroup characterized by red colored loose and long spikes, as well as red colored glumes and awns. Landraces Badri and Mekki formed distinct subgroups negatively correlated to axes 1 and 2 and were both mainly characterized by short spikes with a low to intermediate number of grains per spikelet. Biada and Richi accessions were grouped mainly in the center of the plot and were particularly characterized by white colored spikes, glumes and awns (Table S2). Overall, all landraces were morphologically distinguished using the two axes based on the 12 spike characteristics, except for landraces Roussia and Souri and for landraces Biskri, Jneh Khotifa and Taganrog that were not distinct from each other in respect to their spike size and color. Thus, additional morphological traits were considered to classify the latter landraces into distinct subgroups such as glumes form (Table S2).

SSR polymorphism
Ten SSR markers were used in this study and were mapped on eight different chromosomes and considered therefore largely independent (   Overall, each genetic group corresponds to a landrace. The genetic groups G2, G3, G4, G5, G7, G9, G10 and G11 corresponded to Jneh Khotifa, Taganrog, Mekki, Richi, Badri, Beskri, Biada and Mahmoudi, respectively. Moreover, a signi cant correlation between the genetic distance matrix and morphological distance matrix was obtained (P=0.01; R xy =0.435). However, a discrepancy between the genetic distance matrix and the morphological distance matrix was observed for the landraces Azizi, Souri and Roussia. In fact, Azizi was clustered by STRUCTURE into two different genetic groups G1 and G8, and the two landraces Souri and Roussia were clustered in one genetic group G6 despite their distinct morphological characters.
Forty-one admixed individuals were observed in the collection. The majority of the admix is composed by G6 (Roussia and Souri) and G10 (Biada) representing 14.6 % of the admixed genotypes, followed by G1 (Azizi) and G9 (Beskri) representing 9.7 %.

Analysis of diversity indices and molecular variance
The eleven clusters de ned by the STRUCTURE analysis presented different levels of genetic diversity (  (Table S5). G10 and G4 had both two diagnostic alleles, while G3, G5 and G7 had one diagnostic allele with a frequency > 70%. The xation index (Fis) ranged from 0.698 for G4 to 1 for G7 where Ho was 0.100 and null, respectively. Furthermore, the analysis of variance showed that 59% of the total genetic diversity was observed between the distinct genetic groups, while 41% of the genetic diversity was explained by differences within each group (Table 4).

Network analysis
The genetic relatedness between genotypes was tested using the mininum spanning network (MSN) based on Bruvo's distance. MSN separated all the accessions into two main clusters ( Figure 3). The rst cluster named C1 grouped accessions belonging to Azizi G1 and G8, Jneh Khotifa G2, Richi G5, Souri and Rousia G6, Badri G7 and Biskri G9, while the second cluster named C2 grouped accessions belonging to Taganrog G3, Mekki G4, Biada G10 and Mahmoudi G11. In addition, the pairwise Nei's genetic distances calculated between the 11 genetic groups were also in agreement with the accession clustering by the MSN (Table S6). The highest Nei's genetic distance value (2.416) was recorded between G10 and G5, followed by the genetic distance value (2.319) recorded between G10 and G7. The lowest genetic distance was registered between G1 and G8 (0.421), between G11 and G3 (0.630), and between G3 and G4 (0.630); indicating that G1 and G8, as well as G11, G3 and G4 were the most genetically related groups respectively. In addition, a morphological comparison between the network groupings revealed a signi cant difference (p-values< 0.05) between C1 and C2 for spike shape, spike length, awn length, grain color, grain form, the number of spikelet/spike and for awns and glumes colors ( Table 5). The cluster C1 had a higher gene diversity (He=0.740) and phenotypic diversity (H'=0.77) than cluster C2 with He=0.425 and H'=0.61 (Table S7 and S8). The C1 cluster presented higher spike shape, and spike length values than C2; while C2 had signi cantly higher awns length and grain size traits (Table 5).  Morphological diversity analysis by regions and climatic stages Shannon-Weaver index (H') was assessed based on 12 spike's morphological traits at the ve regions (Sousse, Mahdia, Kairouan, Gabes and Medenine) and the three climatic stages (low semi-arid, high-arid and mid-arid climates) of the designated sites (  (Table S9).

Genetic diversity analysis by regions and climatic stages
The analysis of variance showed that 19% and 10% of the total genetic diversity were observed among regions and among climatic stages, respectively, while 81% and 90% of the genetic variabilities were explained by differences within regions and within climatic stages, respectively (Table 4).
Genetic diversity by region showed a number of effective alleles ranging from 1.366 for Sousse to 3.031 for Gabes (Table 3). Overall and among all investigated regions, Sousse region has shown the lowest genetic diversity indexes, in contrast to the outstanding indexes registered at Gabes. In fact, Gabes region had the highest number of MLG (31) and the highest Shannon's diversity index with 1.296, while Sousse and Medenine had the lowest number of MLG (7), and the lowest Shannon's diversity index registered at Sousse (0.305). Moreover, the percentage of polymorphism was 100% for all regions except for Sousse which was 50%. Gabes had also the highest number of private alleles (17), while Sousse and Medenine had the lowest number of private alleles (1).
The xation index was above 0.800 in each region except for Sousse which was 0.691. Interestingly, solely the diagnostic allele and heterozygosity index were high in Sousse compared to the other regions. In fact, three diagnostic alleles with a frequency that exceeded 70% and a Ho of 0.100 were registered at Sousse, compared to only one at Gabes.
The SSR data analysis by climatic stages revealed that the mid-arid climate was outstanding among the studied climatic stages and had the highest number of effective alleles (3.174), the highest Shannon's diversity index (1.318) and the highest number of private alleles (19). Contrary to the mid-arid climate, the high-arid climatic stage showed the lowest number of effective alleles (2.707), the lowest Shannon's diversity index (1.050) and the lowest number of private alleles (2). However, the xation index was similar among all studied climatic stages and recorded an index above 0.800 for all climatic regions (Table 3).

Correlations between genetic distance and geographic distance
The Mantel test showed a signi cant correlation at (P=0.010; R xy =0.286) between genetic and geographic distances among durum wheat accessions,  Table 5).

Discussion
In the present study, we investigated the genetic diversity of 302 Tunisian durum wheat accessions using ten SSR markers which were able to reach the maximal differentiation among the multi-locus genotypes suggesting that these markers have a good resolution power (data not shown). Overall, the studied collection is characterized by a high genetic diversity level with an overall number of alleles per locus Na of 9.9, a Polymorphic Information Content PIC of 0.690 and a gene diversity He of 0.346. Similar levels of polymorphism (Na=8; PIC=0.68) was previously reported on a Tunisian durum wheat collection composed by 7 modern cultivars and 27 old cultivars ngerprinted with 15 SSR markers [7]. More recently, Slim et al. [19] analyzed the genetic diversity of Tunisian durum wheat germplasm composed of 41 traditional varieties and 13 cultivars using 16 SSR markers, showing a mean PIC value of 0.57 and a He varying from 0.28 to 0.82, with a number of alleles ranging from 2 to 13. A higher level of polymorphism (Na=10; He=0.71) was reported in a wider geographical collection of 172 durum wheat landraces collected from 21 Mediterranean countries and 20 modern cultivars genotyped by 44 SSR markers [22].
However, lower genetic diversity was observed in 33 Anatolian, 136 south Italian and 40 North-West African durum wheat landraces using 14, 44 and 29 SSR markers, respectively [2,12,23]. Low genetic diversity (PIC=0.1; He=0.25) was also observed in 196 Tunisian durum wheat accessions by Robbana et al. [18], due i) to the use of bi-allelic DArTseq markers with lower informativeness level than multi-allelic SSR markers and ii) to a germplasm collection limited to 5 landraces. This variability between results suggests that capturing the maximum genetic diversity would depend essentially on the type of deployed markers, the number of landraces, the origin and geographical distribution (genetic backgrounds) of the studied collection.
Based on 12 morphological traits, the levels of phenotypic diversity detected in our study were consistent with those observed for genetic diversity, with a Shannon-Weaver index H' of 0.80. The morphological diversity was higher than the previously described for Tunisian durum wheat germplasm (H'=0.53) of 930 accessions collected from a reduced number of sites from southern Tunisia, using twenty-two qualitative and three quantitative traits [16]. Lower phenotypic diversity was also observed for Moroccan durum wheat populations composed of 101 landraces (H'=0.62) using nine agro-morphological traits [24] and of 59 traditional durum wheat (H'=0.78) using nine agro-morphological traits [25]. Ethiopian durum wheat populations composed of 32 landraces had an H' index of 0.71 using 8 qualitative traits [26], while Oman populations composed of 161 accessions showed H' index of 0.52 and 0.66 using 15 qualitative and 17 quantitative characters respectively [26]. Population structure, network analysis and relationships between genetic and morphological data In addition to farmer's selection pressure for speci c types of landraces, natural selection was observed morphologically within a single landrace, Mahmoudi. Mahmoudi accessions collected from southern Tunisia showed signi cantly looser spikes than Mahmoudi accessions collected from central Tunisia, characterized by compact spikes. We might suggest that the relaxed spike characterized by an open glume in the Mahmoudi accessions originated from the south could provide tolerance to high temperature by maintaining fertility as it has been observed in rice germplasms [36]. This differentiation between Mahmoudi accessions offers potential tools i) to use relaxed spike trait in breeding programs for heat stress tolerance, and ii) to identify genes and mechanisms involved in ower development useful for improving wheat adaptation to arid and marginal environments.
MSN analysis grouped the accessions into two major clusters C1 and C2. However, C1 and C2 do not correlate with the landraces' geographical origins.
Notably Mahmoudi and Biskri were both introduced from Algeria, while landraces Jneh Khotifa, Azizi, Mekki, Biada and Roussia were considered as local landraces that were cultivated mainly in the north and the center of Tunisia. Nevertheless, landraces Azizi and Mekki had various reported origins, however, no origin has been attributed to landraces Richi and Taganrog that were reported as very old landraces but not local [9,10]. According to Deghais et al. [10], the landrace Jneh Khotifa was also called Jneh Zarzoura and/or Kahla; the denomination of the landrace Souri was extended, from 1915, to Sarebouza wheat received from Armenia. Soriano et al. [22], using 44 SSR, reported that Tunisian durum wheat landraces have four geographical origins, namely East Mediterranean, East Balkan and Turkey, West Balkan and Egypt, and West Mediterranean, with dominance (at more than 50%) of the West Mediterranean genetic group. In addition, Sorriano et al. [22] demonstrated that western Mediterranean landraces were characterized by the heaviest grain weight compared to the three other genetic groups. In our study, grain size did not signi cantly differ between C1 and C2 suggesting that both clusters have indeed western Mediterranean origin. Moreover, Robbana et al. [18] mentioned that most of Tunisian landraces were introduced from the early Carthage trade maritime activity in the Mediterranean Sea, through pathways from Lebanon, Greece and Italy.

Conclusions
Tunisian old durum wheat, characterized here by both high genetic and morphological diversity, represents an important and valuable genetic resource that should be included in breeding and well-established conservation programs. This study showed that Tunisian old durum wheat is structured into landraces revealing the effect of selection pressure directed by farmers for speci c wheat types and agro-morphologies. Nevertheless morpho-geographical spike density trait revealed speci cally in Mahmoudi accessions suggests that environmental selection may occur. Thus, our results provide an interesting venue to improve wheat adaptation to extreme or uctuating Mediterranean conditions. Further physiological and agronomic analysis should be conducted to ascertain whether this trait could be exploited in durum wheat breeding programs for tolerance to heat and drought.
Local durum wheat collection and multiplication A collection of 304 old durum wheat accessions provided by the National Gene Bank of Tunisia (BNG) was used for this study. Accessions were collected from ve regions that belong to three distinct climatic zones -central Tunisia, which is characterized by a low semi-arid climate at Sousse and Mahdia regions and by a high-arid climate at Kairouan region, and southern Tunisia, which is characterized by a mid-arid climate at Gabes and Mednine regions. Global Positioning System (GPS) coordinates of 163 out of the 304 accessions were registered (Table S10). Each accession was sown and puri ed from a single spike-derived lineage by the BNG team and a BNG code has been assigned to each accession. All accessions were further multiplied for spike characterization. The collected set of accessions, used in this study, is preserved at the BNG of Tunisia and is available upon request.

DNA extraction and SSR genotyping
Five seeds from one spike of each accession were germinated and grown under controlled conditions with a photoperiod of 16h/24h, a hygrometry (RH) of 70% and a temperature of 20°C/16 °C (day/night rhythm) at Bioger research unit, INRAE France. At the seedling growth stage (Zadock scale [13][14], one leaf of each accession was sampled and placed in extraction plates. The plates were placed at (-80°C) for 12h before DNA extraction. DNA extraction for each of the 304 accessions was carried out using the DNeasy PowerPlant Pro HTP 96 Kit (Qiagen, France). DNA concentrations were quanti ed using a Nanodrop spectrophotometer (ND-1000) and stored at (-20°C) for subsequent processing. For each accession, DNA was adjusted to 15 ng.μl −1 and genotyped using 10 SSR markers (Table S4). The 10 SSR markers used in this study were selected among 15 SSR markers used in Sahri et al. (2014) [24] study. The forward primers were labeled with uorescent dyes and SSR markers were multiplexed as described by Gautier et al. [37]. For each multiplex, PCRs and electrophoresis were accomplished according to a protocol established by Euro n (https://www.euro ns.fr). DNA ampli cation was performed by preheating the DNA at 95°C for 5 mn, followed by 35 cycles of 95°C for 30 s, 60°C for 90 s and 72°C for 30 s, with a nal extension step of 60°C for 30 mn. PCR products were checked for ampli cation on a 2% agarose gel and fragments were separated according to their size on an ABI Prism Genetic Analyzer (Applied Biosystems). Data was checked again using Peak scanner software version 1.0. Two accessions had missing data for all used SSR and were discarded from our study.

Morphological characterization
The morphological characterization was carried out on ve spikes per accession. Overall, a total of 1520 spikes were characterized among the entire studied collection. Accessions were evaluated using 12 quantitative and qualitative spike morphological traits. Spike evaluation was based on durum wheat descriptor standards of the International Plant Genetic Resources Institute [20] and the International Union for the Protection of New Varieties of Plants [21] (Table S11).
Spike morphological traits, de ned by distinct phenotypic classes, were visually and numerically estimated. Visual phenotypic estimates were attributed to the spike (SC), glumes (GlC), awns (AC) and grains (GC) colors, the density (SD) and the shape (SS) of the spike, the grains form (GF) and size (GS), and awns length compared to the spike. However, the grain size (GS), spike length (SL), the number of spikelets per spike (NS) and the number of grains per spikelet (GN) were measured for each accession and then converted into codes.
Subsequently, accessions were named based on the catalog of cereal varieties cultivated in Tunisia [10]. In fact, the catalog represents a reference reporting and describing typical varietal characteristics of more than 40 old local durum wheat landraces cultivated in Tunisia. The rst author of this paper, PhD student, has undertook the formal identi cation and morphological characterization of the 304 accessions used in the present study

Data analysis
Polymorphism of SSR markers using polymorphic information content PIC To measure the informativeness of the markers, the average polymorphic information content (PIC) was calculated for each SSR by determining the frequency of alleles per locus according to the formula given by Powell et al. [38]: Polymorphism of morphological traits using Shannon-Weaver index H' Frequencies of the different phenotypic classes were calculated for each of the 12 spike's morphological traits in the whole collection, by regions, by climatic stages (Table S9) and by landraces (Table S1). Based on these frequencies, the Shannon-Weaver index (H) was calculated for each trait using Past software [39]. H was estimated for the entire durum wheat collection, regions, climatic stages and for each landrace. Based on SSR data generated for 302 accessions, 188 multilocus genotypes (MLG) were identi ed with GIMLET software version 1.3.2 [41] .A population genetic structure analysis was conducted on the 188 MLG, using the program STRUCTURE version 2.3.4 [42]. The run was conducted with K-values varying from 1 to 20 in an admixture ancestry model applying 10 independent runs for each of the different K values. A burn-in phase of 100,000 iterations and 100,000 Markov Chain Monte Carlo (MCMC) iterations were performed. The run with maximum likelihood was used to assign individual genotypes into genetic groups. Genotypes with a liation probabilities (inferred ancestry) > 75% were assigned to a distinct genetic group and those with < 75% were treated as admixed. Plot of mean posterior probability (ln P(D)) values per clusters (K), and delta-K method of ln P(D) STRUCTURE harvester version 0.6.94 were used to determine the optimal number of genetic groups [43] .
In addition, a minimum spanning network (MSN) based on Bruvo's distance [44] (Bruvo et al., 2004) using "poppr" and "adegenet" packages was generated under R 3.3.2 [45], in order to classify the 302 accessions according to their genetic relationship . Furthermore, a mean of each of the 12 spike's morphological traits was calculated for accessions belonging to the different clusters de ned by the MSN analysis, as follows: Where N is the number of genotypes per genetic cluster as de ned by the MSN analysis, n is the number of individuals per phenotypic class and C i is the i th phenotypic class per morphological trait.
Based on the calculated means, an analysis of variance ANOVA was carried out using R 3.3.2 [46] to test for signi cant differences between genetic clusters for each morphological trait.
Population genetics and data analysis by regions and climatic stages GenAlEx version 6.501 [40] was used to calculate the number of alleles (Na), the number of effective alleles (Ne), the number of private alleles (PA : alleles speci c to a single population), the Shannon's information index (I), the expected (He) and observed (Ho) heterozygosity, the xation index (F), the percentage of polymorphic loci (P), and the diagnostic alleles (DA is a rare allele with a frequence >70% for a genetic group or region and <30% for the others) within each genetic group, region and climatic stage.
In addition, the correlation between the genetic distance and the log (1+geographic distance) transformed geographic distance of accessions was analyzed using a Mantel test [46] for the entire collection, under GenAlEx version 6.501. Correlations between the genetic distance matrix and morphological distance matrix were also assessed using a Mantel test.
Furthermore, an analysis of molecular variance (AMOVA) was performed under GenAlEx version 6.501 to investigate the signi cance of genetic group differentiation as de ned by STRUCTURE and the genetic variability explained by regions and climatic stages.
Moreover, a mean of the 12 spike's morphological traits was estimated for Azizi and Mahmoudi accessions, both existing in different climatic zones of central and southern Tunisia, as described above. For each morphological trait, an analysis of variance ANOVA was carried out using R 3.3.2 [45] to test for potential regional effects on the morphological traits. Availability of data and materials The data sets supporting the results of this article are included in this manuscript and its additional information les.

Competing interests
The authors declare that they have no competing interests.  302 Tunisian durum wheat accessions genotyped with 10 SSR markers inferred from STRUCTURE [42]. Each MLG is represented by a vertical line and they are ordered by color-coded genetic subpopulation (G1 to G11). For each genetic subpopulation, corresponding durum wheat landrace is mentioned. * Azizi landrace was divided into two genetic subpopulations G1 and G8. Minimum spanning network using Bruvo's distance of 302 durum wheat accessions genotyped with 10 SSR markers, performed under R software. Each node represents a multilocus genotype (MLG) and the size of the node is proportional to the number of accessions representing the MLG. MLGs were color-coded according to their membership to a genetic subpopulation (G1 to G11) as de ned by STRUCTURE at K=11. Admixed individuals were color-coded in gray. Edge widths represent relatedness.