Plant material

The plant material consisted of 15 Triticum spp. accessions, namely two cultivars and five landraces of T. aestivum, two cultivars, two landraces and two accessions of unknown status of T. durum, one landrace of T. polonicum and one landrace of T. dicoccum. The latter was collected in the early 20th century in Greece, it was conserved in USDA-GRIN (Code PI 94682) and repatriated recently. Landraces were collected from different regions in Greece and donated to the Plant Breeding and Biometry Laboratory of the Agricultural University of Athens (AUA) (Table 1 and Figure 1). The Institute of Plant Breeding and Genetic Resources (IPGRB) donated four cultivars, recommended for bread making (‘Yekora’ and ‘Elisavet’ belonging to T. aestivum) and pasta making (‘Mexicali 81’ and ‘Elpida’, of T. durum) which were used as controls.

Experimental design

The experiment was conducted at an experimental field of the Agricultural University of Athens (location: 37° 59' 06.8" N, 23° 42' 24.7" E, altitude 24m a.s.l.) during the winter growing season of 2019–2020. The meteorological data are presented in Figure 2. The values of mean air temperature and precipitation for the Athens region during the years 2015–2020 were provided by the Hellenic National Meteorological Service (HNMS, 2023). A randomized complete block design with three replications was used. Each replication consisted of 20 plants of each accession sown in rows spaced 0.2m apart. Sowing distances among plants in each row was 20cm. The sowing date was mid-November, a typical period for sowing in central Greece.

Manipulation of sowing time is one of the most effective techniques for the production of high-quality wheat kernels (Butkovskaya & Kozulina, 2021). In the present study, for comparison reasons, plant sowing was carried out simultaneously for all accessions, without considering the best sowing time for each accession.

Measurement of agromorphological traits

During vegetative and reproductive stages, time of ear emergence (days), total number of tillers, ear length excluding awns (cm), ear length including awns (cm), number of spikelets in the spike of the first tiller, plant length (measured from the base of the plant to the tip of the highest awn, cm), stem length (measured from plant base to ear base, cm) and whole plant weight (physically dried by sun, in g/plant) were measured. In addition, the qualitative variables ear colour, ear shape in profile, growth habit and awn colour, were scored. Agromorphological traits were evaluated on all the plants per plot to calculate the mean value for the quantitative traits and the median for the qualitative traits. The measurements of the following traits: plant length, growth habit, ear emergence, ear length, ear colour, ear shape, and awn colour were carried out using UPOV descriptors (UPOV, 2012; UPOV, 2017).

Physical properties of seeds

Before measurements, shrunken, broken kernels, and other impurities were removed. The weight of 1,000 randomly selected kernels of wheat was measured for thousand kernel weight (TKW). The same 1,000 selected kernels were then filled into a volumetric cylinder to determine their volume. Tap volume was measured after gently tapping the cylinder 100 times or until there was no further decrease in the sample level. Bulk density (ρ_b) and tap density (ρ_t) were calculated as weight of sample per volume and per tap volume of sample, respectively (g/ml) (Chaloulos, Bazanis, Georgiadou, Protonotariou, & Mandala, 2021).

Determination of flow properties is essential in processes such as transportation, mixing and storage (Fitzpatrick, Barringer, & Iqbal, 2004). The Carr index (CI) was used to calculate the flowability of the seeds (Carr, 1965):

CI (%) = (𝜌_𝑡 − 𝜌_𝑏) ⁄𝜌_𝑡 × 100

The firmness of seed (SF), defined as the force required to crush the seed, was estimated using an Instron (Universal Testing Machine, Model 3343, Nordwood, MA, USA) equipped with a 1kN load cell. Fifty intact seeds of each accession were compressed to 50% of their initial width with a 4cm diameter probe and a speed of 1mm/min.

Seed colour was measured using the colourimeters 3nh High-Quality Spectrophotometer NS800S (Shenzhen 3nh Technology, China), according to CIE-L*a*b* uniform colour space, where L* indicates lightness, a* indicates hue on a green (−) to red (+) axis, and b* indicates hue on a blue (−) to yellow (+) axis. White index (WI) was also calculated according to the following equation:

$WI\hspace{0.17em}= 100 - \sqrt{(100 - {L^{*}}^2) + {a^{*}}^2 + {a^{*}}^2}$  

Image analysis-shape factor measurements

One hundred seeds of each accession were scanned twice using a flatbed scanner (HP scan jet 4370, Hewlett Packard, USA). Images of the seeds were captured for further analysis using an image analysis software (Image Pro Plus 7, Media Cybernetics, USA). An indicative image of the scanned seeds is presented in Figure 3A. Area (mm²), aspect, average optical density (density/intensity mean), mean, maximum and minimum diameter (mm), perimeter (mm), roundness, size length (mm) and width (mm) were determined. Detailed descriptions and definitions of shape factors are presented in Figure 3B. The measurements were done in triplicates.

Statistical analysis

Statistical analysis of the data was performed with Statgraphics Centurion XV (Statgraphics, Rockville, MD, USA). For qualitative variables (plant growth habit, colour of ear, ear shape in profile and awn colour) frequency percentages (Supplemental Figure 1) and median values were calculated (Supplemental Table 1). Friedman Test was used to test the null hypothesis that the mean ranks of the samples groups are the same (Supplemental Table 1).

Quantitative trait data were subjected to analysis of variance after testing the assumptions of normality and homoscedasticity with Levene's and Bartlett's tests respectively (Supplemental Table 2). The comparisons of the means were performed using the Tukey HSD criteria with a level of significance of α = 0.05.

The correlation between quantitative parameters was assessed by Pearson's correlation test (significance level at α = 0.05). The correlation between qualitative parameters (nonparametric measures) was assessed by Spearman rank correlation test (significance level at α = 0.05). To correlate quantitative and qualitative variables the Spearman test was also used (Supplemental Table 3). Cluster analysis was used to group the quantitative observations with similar characteristics using nearest neighbour (single linkage) and distance metric of squared Euclidean on standardized data.

Agromorphological traits

Mean values of quantitative agromorphological traits for all studied wheat accessions are presented in Table 2. Time of ear emergence ranged from 107 to 138 days (Yekora - T. aestivum, cultivar, and Zoulitsa - T. aestivum, landrace, respectively). All cultivars, except Elisavet - T. aestivum, were earlier than landraces. Similarly,Frankin et al. (2021) reported cultivars to be on average 12% earlier than landraces. On the other hand, the landraces Zoulitsa (T. aestivum), Kopaida (T. dicoccum) and Leventis (T. polonicum) had the latest heading time of all accessions. The T. aestivum landraces, Asprositi Kozani (W12), Ntopio (W6) and Asprositi Kalavrita (W11) exhibited early maturation compared to other landraces. Consequently, these landraces have the potential to be incorporated into breeding programmes focused on selecting for early maturation traits (Javaid, Rabbani, & Rastina, 2005). Time of ear emergence was positively correlated with total number of tillers, plant length (from the base of the plant to the tip of the highest awn) and stem length (0.584, 0.674 and 0.710, P < 0.05, respectively) (Table 3). A positive correlation between days of heading and number of fertile tillers was also recorded in a study with 64 bread wheat genotypes in West Shewa and found that days to heading were positively correlated with the number of productive tillers (Mecha, Alamerew, Assefa, Assefa, & Dutamo, 2017). In the study of Bilgrami et al. (2020), where bread wheats were used, the total tiller number was also significantly correlated with days from sowing to heading (0.22, P < 0.05). Although landraces had more tillers than cultivars, the difference between landraces and specific cultivars was not statistically significant and the number of tillers ranged from 4.18 (Mavragani Lemnos - T. durum, unknown status, W13) to 10.38 (Grilos - T. aestivum, landrace, W8).

In most cases, the landraces had higher total plant weight than the cultivars. The weight of the landraces ranged from 28.97g/plant (Asprositi Kalavrita - T. aestivum, landrace, W11) to 73.01g/plant (Asprositi Kozani - T. durum, landrace, W12), while cultivars’ weight ranged from 17.55g/plant (Yekora - T. aestivum, cultivar, W1) to 25.33g/plant (Elpida - T. durum, cultivar, W3). This could be attributed to the higher biomass produced by landraces, a characteristic desirable by farmers who need straw for their animals (Thanopoulos et al., 2021). Plant weight was positively and moderately correlated with the total number of tillers and ear length including awns (0.570 and 0.526, P < 0.05, respectively) and highly correlated with the number of spikelets, plant length and stem length (0.644, 0.796 and 0.759, P < 0.01, respectively). The number of spikelets per spike of the first tiller did not differ significantly between accessions.

Cultivars were significantly shorter than the landraces, which is due to the introduction of Rht (Reduced height) genes in modern wheat varieties. These dwarfing genes have been used in wheat breeding to develop cultivars with short stature (Acquaah, 2012). Similar results were recorded in another study, where Italian landraces were found to have taller culms compared to cultivars (Preiti et al., 2022). The plant length for the cultivars was between 83.88cm (Yekora - T. aestivum, cultivar, W1) and 103.87cm (Mexicali 81 - T. durum, cultivar, W4), while for the landraces the plant length was between 118cm (Kopaida - T. dicoccum, landrace, W10) and 159.04cm (Aspratheri - T. durum landrace, W14). Mavragani accessions, W7 and W13, of unknown breeding status, were short, with 87.07cm and 93.18cm, respectively. These values did not differ significantly from cultivar plant length (W1 - W4) (Table 2). Low plant height is a desirable trait for commercial varieties, as it is correlated with good standing ability and resistance to lodging (Würschum, Langer, Longin, Tucker, & Leiser, 2017). The weak point of wheat landraces is the lodging, but under climatic change and high temperatures, taller plants could demonstrate higher yield than short ones (Jatayev et al., 2020).

Plant growth habit was significantly correlated with weight per plant, total number of tillers, plant length and stem length (base to ear) (0.525, 0.623, 0.538, 0.529, respectively P < 0.05) (Supplemental Table 3). A higher number of tillers leads to an increased inclination of the plant.

Cluster analysis based on the quantitative agromorphological traits grouped all cultivars with two Mavragani accessions of unknown status (W7 and W13) in cluster 1, while all landraces were included in a distinct cluster (cluster 2) (Figure 4). The similarity of Mavragani durum wheat accessions from Skyros and Lemnos, with wheat cultivars indicates that these accessions could be old cultivars grown and reproduced for years by farmers, who considered them of local origin (Thomas et al., 2013). Overall, the quantitative agromorphological traits based on UPOV descriptors were sufficient to separate wheat accessions by their breeding status, namely landraces or cultivars.

Physical properties of seeds

TKW is of great interest (Javaid et al., 2005) as it is an important index for the prediction of flour extraction rate (Posner, 2009). Higher values of TKW indicate a higher percentage of endosperm leading to higher flour yield (Wiersma, Busch, Fulcher, & Hareland, 2001). TKW values of the studied accessions are presented in Figure 5. TKW of W9 (T. polonicum) was significantly higher (80.45g) than all other accessions. W10 (T. dicoccum) presented the lowest value of TKW (26.92g), consistent with previous research, where the TKW of 38 emmer wheat accessions, collected in several European countries, ranged from 22.9g to 42.6g (Mondini, Grausgruber, & Pagnotta, 2013). T. dicoccum presented significantly lower values of TKW compared to T. durum, 32.02–36.12g and 45.17–46.31g, respectively (Shoormij, Mirlohi, Saeidi, & Mohammadi, 2022). TKW in the rest of the accessions, either cultivars or landraces, ranged from 34.08g (Elisavet - T. aestivum, cultivar, W2) to 67.16g (Mavragani Lemnos - T. durum, unknown status, W13) and were in accordance with Travlos, Karamanos, Economou, and Papastavrou (2012) who studied five Greek bread wheat landraces (30–50.7g). Specifically, in the present study, W5 (Zoulitsa) had an average TKW of 43.66g, while in Travlos et al. (2012), TKW in Zoulitsa ranged from 36g to 44g, indicating that climatic conditions can affect TKW but also the origin of the genetic material (Papadakis, 1929). Moreover, it is worth noting that in the present study, T. aestivum in most cases presented significantly lower values of TKW compared to T. durum accessions. Landraces had similar or higher values of TKW compared to cultivars of the same species, indicating that landraces could give high flour yield. However, results from previous research are contrasting as no significant difference between landraces’ and cultivars’ TKW was reported for Turkish (Cetiner et al., 2020) and Sicilian wheats (Ruisi et al., 2021). On the contrary, Spanish bread wheat landraces showed higher TKW compared to a reference set (López-Fernández et al., 2021). In another research, similar TKW values for T. aestivum landraces and cultivars were detected, while T. durum cultivars, compared to landraces, had significantly higher TKW values (Frankin et al., 2021).

Due to their high heritability grain perimeter, area and volume are important indirect indices for grain yield improvement (Abdipour et al., 2016). Seed volume ranged from 48ml to 113ml (Elisavet - T. aestivum, cultivar, W2 and Leventis - T. polonicum, landrace, W9, respectively) (Table 4). It is worth noting that W9 (T. polonicum) had significantly the highest values of TKW, volume and seed firmness (SF). Although Papadakis (1929) mentioned that T. polonicum did not provide any agronomic value, our findings show that this species has bread-making value.

There are many methods to determine the required force to evaluate grain mechanical properties. In the present study, a rapid measurement of the mechanical properties of a single grain was used, which provided valuable results (Shewry & Hey, 2016). SF ranged from 62.96N to 194.85N (W10, Kopaida - T. dicoccum, landrace and W9, T. polonicum, landrace, respectively) and was correlated positively with seed weight and volume (0.840 and 0.791, P < 0.05, respectively) (Table 5). High values of weight and hardness indicate healthy seeds. Seed texture determines, to a great extent, milling properties and end uses. Softer seeds require more energy to mill and produce finer flours suitable for biscuit production, while flour from harder seeds is more appropriate for bread making (Liu, 2008). As W10 (Kopaida - T. dicoccum, landrace) had the lowest SF values, it could be suitable for cake and biscuit production. A recent research concluded that in general old cultivars of T. aestivum were softer than modern cultivars (Cetiner et al., 2020). In the present research, this statement was not fully confirmed, as there was no clear trend. However, it should be mentioned that fewer old and modern varieties were used in the present work compared to the earlier study.

Bulk density could be used as a criterion for yield improvement, as it was correlated positively with grain yield (Karimizadeh, Sharifi, & Mohammadi, 2012). In the present study, density did not differ significantly between accessions except for W10 (Kopaida - T. dicoccum, landrace), having the lowest value (0.55g/ml).

Seed flowability for all accessions was very good as indicated by Carr indices (CI) with values lower than 15% (Carr, 1965). W11 (Asprositi Kalavrita - T. aestivum, landrace) had the lowest CI value, 1.79%, and W13 (Mavragani Lemnos - T. durum, unknown status) had the highest value, 6.53%.

In general, colour parameters did not differ significantly among accessions and were not correlated with the other physical properties (Table 5). The exceptions were the green/red hue (a*), which correlated negatively with volume (-0.516, P < 0.05), indicating that seeds were redder as their volume increased, and lightness (L*), blue/yellow hue (b*) and white index (WI), which were correlated positively with density (0.649, 0.719 and 0.565, P < 0.05, respectively). Increased density led to increased lightness, yellowness and whiteness of the seeds. W10 (Kopaida - T. dicoccum, landrace) was the only accession with a significantly lower value of b* among all other accessions. Colour determines to a great extent the overall consumer acceptance of food (Mandala & Protonotariou, 2021). Wheat kernel colour is related to different milling and baking attributes of the final products (Dowell, 1998). Bright yellow or amber colour of durum wheat results in high-quality pasta products (Cole, Johnson, Cole, & Stone, 1991), thus sample W10 (Kopaida - T. dicoccum, landrace) probably is the least suitable for pasta making among the studied accessions. Probably, W9 (Leventis - T. polonicum, landrace) and W13 (Mavragani Lemnos - T. durum, unknown status) could be the most suitable for pasta making considering their high values of b*, TKW and seed volume. However, further investigation is needed. Mechanical properties and sensory evaluation of produced pasta could verify our claim.

The determination of kernel morphology is of great importance for bread making, as kernel shape and uniformity may influence the milling quality (Campbell et al., 1999). Image analysis could play a key role in distinguishing different wheat varieties by defining the product traceability for wheat landraces (Grillo, Blangiforti, & Venora, 2017; Khoshroo, Arefi, Masoumiasl, & Jowkar, 2014). In the present research, mean seed diameter ranged between 4.17mm (W2, Elisavet - T. aestivum, cultivar) and 5.66mm (W9, Leventis - T. polonicum, landrace). Particle size distribution curves differed between accessions, indicating that some landraces had a more uneven seed population than cultivars (Figure 6). Curve of W12 (Asprositi Kozani - T. durum, landrace) presented an almost bimodal distribution, indicating the presence of two distinguished populations of seeds, with different sizes, which could be explained by the fact that sometimes landraces consisted of two or more populations with different morphological characteristics (Papadakis, 1929). This could be linked with seed heteromorphism, characteristic of wild populations. A similar curve was presented by W13 (Mavragani Lemnos - T. durum, unknown status), where the bimodal distribution was less evident, but also presented a shoulder on the left (Figure 6). In contrast, W3 (Elpida - T. durum, cultivar) presented the narrowest unimodal distribution curve, with the peak around 5.5mm. All the other accessions presented frequency curves similar to W2 (Elisavet - T. aestivum, cultivar) or W9 (Leventis - T. polonicum, landrace). The main characteristic of these curves is that distribution is broad, due to the uneven seed populations. W10 (Kopaida - T. dicoccum, landrace) presented a similar curve to W2 (Elisavet - T. aestivum, cultivar), despite the fact that they belonged to different species.

Bread wheat cultivars (W1, Yekora - T. aestivum and W2, Elisavet - T. aestivum) had significantly more roundness and shorter seeds compared to pasta wheat cultivars (W3, Elpida - T. durum, and W4, Mexicali 81 - T. durum), as referred to in the literature (Campbell et al., 1999). Perimeter was significantly lower for W2, Elisavet - T. aestivum, cultivar (15.71mm) and significantly higher for W9, Leventis - T. polonicum, landrace (23.62mm), compared to all other accessions (Supplemental Table 4). Similar results were observed for seed length, with W9 having the longest seeds (9.3mm) and W2 the shortest (5.96mm). T. polonicum (represented by W9) had characteristically long seed (Percival, 1921). Regarding the width of the seeds, the differences were not significant in all the cases and the values ranged from 2.72mm to 3.94mm (W10 and W13, respectively). Grain width could be used as a predictive index for mean grain weight determination (Haghshenas, Emam, & Jafarizadeh, 2022). However, in the present study, the most valuable parameter was the mean diameter.

The average grain area was 18.54mm² (from 13.85mm² to 26.08mm², for W2, Elisavet - T. aestivum, cultivar and W9, Leventis - T. polonicum, respectively), in accordance with other studies (Abdipour et al., 2016; Gegas et al., 2010; Okamoto, Nguyen, Yoshioka, Iehisa, & Takumi, 2013). Only Kondić et al. (2020) recorded significantly higher values of seed area, 53.5mm² and 40.17mm² in two experimental years in Bosnia and Herzegovina for seeds of T. aestivum.

Correlations among all parameters

In the present study, TKW was significantly positively correlated with area of seed, mean diameter, SF, perimeter, length, width and volume of the seeds (0.871, 0.887, 0840, 0.786, 0.677, 0.829, and 0.975, P < 0.05, respectively, Supplemental Table 5). There is previous research which also associated kernel weight with its width and length (Campbell et al., 1999; Kondić et al., 2020), but some of them could not associate these parameters when they studied only bread wheats (Schuler, Bacon, Finney, & Gbur, 1995). Moreover, Javaid et al. (2005) found that TKW was positively correlated with plant height in Pakistani bread wheat landraces, and Moghaddam et al. (1997) found a significant correlation between TKW and number of tillers in Irani bread wheat landraces, which was not observed in the present study, suggesting possibly different patterns.

There are several studies evaluating wheat kernels by compression tests (Ponce-García, Ramírez-Wong, Escalante-Aburto, Torres-Chávez, & Figueroa-Cárdenas, 2016). However, there is a research gap on the correlation between wheat seed mechanical properties with plant-agromorphological and seed-morphological traits. In the present study, SF was significantly correlated with plant weight, and seed area, mean diameter, perimeter, length, width, volume and TKW (0.626, 0.711, 0.709, 0.688, 0.615, 0.570, and 0.840, P < 0.05, respectively). Thus, SF could be used as an additional tool to predict all the above parameters. Fitting a simple linear regression model to describe the relationship between SF and TKW explained 70.58% of firmness variability (P < 0.05) (Supplemental Figure 2).

Grouping of accessions

Cluster analysis, using all quantitative data, allowed us to group the wheat accessions in accordance with their species and breeding status (Figure 7). W9 (Leventis - T. polonicum, landrace) and W10 (Kopaida - T. dicoccum, landrace) created a distinct clade each, as anticipated, considering that they were different species. The control cultivars of T. durum species (W3 and W4) created a distinct cluster, along with the T. durum accessions of unknown breeding status W7 and W13 (Mavragani - T. durum, unknown status from Skyros and Lemnos, respectively). Interestingly, except for W11 (Asprositi Kalavrita) all landraces, whether T. aestivum or T. durum, formed a separate cluster, related to the T. aestivum cultivars.

It is worth noting that with the addition of physical properties in the cluster analysis, the grouping of Triticum spp. accessions changed compared to Figure 4, where only observations according to agromorphological traits were used. The cluster of landraces was closer to T. aestivum cultivars, indicating that landraces – regardless of the species they belonged to – could be more suitable for bread making, considering their physical properties, but also the main preference and use (mainly for bread) by farmers in Greece (Douma et al., 2016).