2.1. SSR Polymorphism
In order to evaluate the genetic diversity of the 187 Apulian almond genotypes, an in-depth molecular characterization was performed using 18 SSR markers, chosen on the basis of their dispersal map location. Capillary electrophoresis analysis produced clear genotyping profiles for all the examined loci, yielding good and reproducible amplification products within the expected allele-size ranges (
Table 1). A total number of 298 alleles (Na) was obtained, with an average value of 16.5 alleles per locus, ranging from a minimum of 5 for BPPCT014 to a maximum of 22 alleles for UDP96005 and BPPCT025. This mean value was slightly lower than that observed by [
35] (18.0) and by [
29,
36] (17.2 and 18.7, respectively) in their studies of almond genetic diversity, but it was perfectly in line with what was reported by [
37] (16.8) for the characterization of an Iranian almond collection.
The mean of the effective alleles (Ne) was 6.10, ranging from 2.23 for BPPCT014 to 11.00 for CPDCT045, slightly lower as compared to two other reports [
29,
37]. For each microsatellite, the PIC (Polymorphism Information Content) value, i.e., marker informativeness and richness in terms of number of allelic forms [
38], was calculated. In general, the molecular analysis revealed a high degree of polymorphism for all the loci considered, as the PIC values were always greater than 0.681, the mean being 0.783. The only exception was the BPPCT014 marker, whose PIC value was 0.453, thus resulting the least polymorphic microsatellite. In accordance with these findings, all SSRs except BPPCT014 harbored at least two private alleles (allelic frequency < 1%), the greatest number (9 private alleles) being observed for BPPCT025 and UDP96005. Comparable PIC values have been observed in other research works [
21,
37,
39], where the microsatellites used had mean discrimination powers of 0.79, 0.81 and 0.80, respectively. Interestingly, rather higher PIC values (0.85 and 0.92) were reported in the studies carried out by [
35,
40].
The observed Heterozygosity (Ho) ranged from 0.335 (UDP98409) to 0.935 (CPDCT045) with a mean value of 0.711, while the expected Heterozygosity (He) varied between 0.553 for BPPCT014 to 0.909 for CPDCT045, with an average measure of 0.807 (
Table 1). This difference determined positive fixation index (F) values (mean 0.116) for almost all markers, as a measure of genetic diversity; only two microsatellites, i.e., BPPCT014 and CPDCT045, were exceptions, since their F index value was negative (−0.006 and −0.028, respectively) (
Table 1). The average Ho value calculated in our study was similar to those obtained by [
29] (Ho = 0.72), [
35] (Ho = 0.71) and [
39] (Ho = 0.73), resulting quite a lot higher than the values reported by several other authors [
21,
34,
36,
37,
41]. In general, in all these research studies, the He values were higher than Ho, thus determining positive fixation indexes for most of the microsatellites used. The great heterozygosity of almond cultivars is in line with the mating system of this species that is normally self-sterile and out-crossing [
41]. Thus, the high number of detected alleles and the heterozygosity value reflected the ability of SSR markers to provide a unique genetic profile for each individual genotype. The presence of null alleles (Nu) was also detected in our study. A null allele frequency greater than 0.2 is usually considered the threshold above which a significant underestimation of the expected heterozygosity due to null alleles is possible [
42,
43]. Nu values higher than 0.20 were obtained for 4 of the 18 microsatellites used, namely the BPPCT001 (0.3486), CPSCT018 (0.2918), UDP98409 (0.3725) and UDP98412 (0.3803) loci (
Table 1). For this reason, these loci were not considered in further analyses.
Moreover, Shannon’s information Index (I) ranged from 0.926 (BPPCT014) to 2.608 (CPDCT045), with an average value of 2.052, indicating a high level of genetic diversity in the collection studied (
Table 1).
Table 1.
Genetic diversity indices estimated for the considered SSR loci in the Apulian almond collection. For each locus, the allele size range (basepair, bp), the number of detected alleles (Na), the effective number of alleles (Ne), the observed (Ho) and expected (He) heterozygosity, the fixation index (F), Shannon’s index (I), the PIC value and the frequency of null alleles (Nu) are reported. The presence of null alleles is highlighted in bold.
Table 1.
Genetic diversity indices estimated for the considered SSR loci in the Apulian almond collection. For each locus, the allele size range (basepair, bp), the number of detected alleles (Na), the effective number of alleles (Ne), the observed (Ho) and expected (He) heterozygosity, the fixation index (F), Shannon’s index (I), the PIC value and the frequency of null alleles (Nu) are reported. The presence of null alleles is highlighted in bold.
SSR Locus | Size | Na | Ne | Ho | He | F | I | PIC | Nu |
---|
BPPCT001 | 101–159 | 21.0 | 5.627 | 0.408 | 0.822 | 0.504 | 2.197 | 0.805 | 0.3486 |
BPPCT007 | 119–159 | 15.0 | 6.704 | 0.754 | 0.851 | 0.114 | 2.091 | 0.834 | 0.0573 |
BPPCT010 | 122–162 | 18.0 | 8.193 | 0.832 | 0.878 | 0.052 | 2.336 | 0.866 | 0.0229 |
BPPCT014 | 178–198 | 5.0 | 2.236 | 0.556 | 0.553 | −0.006 | 0.926 | 0.453 | −0.0060 |
BPPCT025 | 149–197 | 22.0 | 6.095 | 0.780 | 0.836 | 0.067 | 2.226 | 0.820 | 0.0338 |
CPDCT025 | 156–198 | 14.0 | 4.576 | 0.747 | 0.781 | 0.044 | 1.954 | 0.764 | 0.0192 |
CPDCT042 | 160–222 | 21.0 | 7.227 | 0.819 | 0.862 | 0.049 | 2.319 | 0.849 | 0.0262 |
CPDCT045 | 126–176 | 20.0 | 11.001 | 0.935 | 0.909 | −0.028 | 2.608 | 0.902 | −0.0154 |
CPPCT006 | 172–206 | 18.0 | 8.385 | 0.860 | 0.881 | 0.023 | 2.367 | 0.870 | 0.0108 |
CPPCT033 | 127–165 | 13.0 | 4.575 | 0.766 | 0.781 | 0.019 | 1.814 | 0.750 | 0.0069 |
CPSCT012 | 140–186 | 17.0 | 3.495 | 0.572 | 0.714 | 0.198 | 1.835 | 0.699 | 0.1181 |
CPSCT018 | 145–177 | 14.0 | 7.086 | 0.472 | 0.859 | 0.451 | 2.169 | 0.843 | 0.2918 |
EPPCU5176 | 106–134 | 16.0 | 4.168 | 0.746 | 0.760 | 0.019 | 1.838 | 0.730 | 0.0075 |
PCHGMS1 | 180–222 | 14.0 | 7.067 | 0.797 | 0.859 | 0.072 | 2.135 | 0.843 | 0.0364 |
UDP96003 | 87–137 | 14.0 | 3.488 | 0.713 | 0.713 | 0.000 | 1.678 | 0.681 | −0.0011 |
UDP96005 | 126–196 | 22.0 | 6.535 | 0.759 | 0.847 | 0.103 | 2.337 | 0.835 | 0.0581 |
UDP98409 | 120–170 | 19.0 | 3.553 | 0.335 | 0.719 | 0.534 | 1.930 | 0.706 | 0.3725 |
UDP98412 | 94–134 | 15.0 | 6.231 | 0.376 | 0.840 | 0.552 | 2.177 | 0.824 | 0.3803 |
Mean value | - | 16.5 | 6.105 | 0.711 | 0.807 | 0.116 | 2.052 | 0.783 | 0.0709 |
The allelic similarity among the Apulian almond genotypes analyzed was also calculated by means of LRM estimation (pairwise relatedness), setting 0.5 as maximum value for identical genetic profiles (cases of synonymy). A complete genetic identity was observed for the cultivars “Mollesca di Ruvo”, “Troito”, “Tuono” and “Stilla” (
Table 2). The genetic similarity of “Tuono” with “Troito”, as well as with other Italian and foreign cultivars such as “Supernova” [
44], “Moncajo” and “Laurenne” [
45], has been previously reported by other authors and is probably due to the extensive use of this cultivar in breeding programs, as they are good sources of self-compatibility [
24,
46,
47]. Therefore, as “Mollesca di Ruvo” and “Stilla” also showed a high genetic similarity with “Tuono”, we hypothesized that, together with “Troito”, they could be Tuono-related cultivars, probably assigned different names due to their morphological diversity.
Most of the examined genotypes had LRM values between 0.4 and 0.5, thus resulting unique individuals with no case of synonymy (
Table 2). High LMR values were found for the couples “Bianchetta-Lunghina”, “Del lago-Rachele tenera” and “Riviezzo Grosso-Riviezzo Piccolo” and for the groups “Bianchetta-Lunghina-Secolare Cotogni” and “Mollesca di Ruvo-Mollese grossa_2-Tuono-Troito-Stilla-Piangente”.
To our knowledge, the evaluation of the LRM parameter has never previously been reported in other studies about almond genetic diversity.
Table 2.
List of pairwise relatedness based on the LRM estimator.
Table 2.
List of pairwise relatedness based on the LRM estimator.
Genotypes with LRM = 0.5 |
---|
Mollesca di Ruvo | Stilla | 0.50 |
Mollesca di Ruvo | Troito | 0.50 |
Stilla | Troito | 0.50 |
Mollesca di Ruvo | Tuono | 0.50 |
Stilla | Tuono | 0.50 |
Troito | Tuono | 0.50 |
Genotypes with 0.4 < LRM < 0.5 |
Bianchetta | Lunghina | 0.49 |
Del lago | Rachele tenera | 0.48 |
Riviezzo Grosso | Riviezzo Piccolo | 0.47 |
San Giuseppe_2 | Troia | 0.46 |
Mollesca di Ruvo | Mollese grossa_2 | 0.45 |
Mollese grossa_2 | Stilla | 0.45 |
Mollese grossa_2 | Troito | 0.45 |
Mollese grossa_2 | Tuono | 0.45 |
Bianchetta | Secolare Cotogni | 0.45 |
Ficanera | Mollese di canneto | 0.44 |
Ciavea | San Michele strada | 0.44 |
Chino | Rachele tenera | 0.44 |
Don Carlo | San Michele strada | 0.44 |
Lunghina | Secolare Cotogni | 0.44 |
Montranese | Zin Zin | 0.43 |
Pilella | Scquicciarina | 0.43 |
Chino | Del lago | 0.43 |
Gianfreda | Riviezzo Grosso | 0.42 |
Mollesca di Ruvo | Piangente | 0.42 |
Piangente | Stilla | 0.42 |
Piangente | Troito | 0.42 |
Piangente | Tuono | 0.42 |
Pettolecchia | Strappasacco | 0.42 |
Gianfreda | Riviezzo Piccolo | 0.41 |
Portone Gioia | San Michele strada | 0.41 |
Ciavea | Zin Zin | 0.41 |
Ferrante | Sciddiata Nisi | 0.40 |
2.2. Genetic Characterization
The neighbor-joining analysis allowed us to obtain a phylogenetic tree and to subdivide the Apulian almond population into three main clusters: G-1, G-2, and G-3 (
Figure 1). The great majority of genotypes was included in cluster G-1, featuring 116 individuals, while 58 genotypes were grouped together in cluster G-2; cluster G-3 resulted the least abundant, including only 13 genotypes. Moreover, all these major clusters contained at least two recognizable subgroups, that we denominated G-1A and G-1B, G-2A and G-2B, and G-3A and G-3B (
Figure 1). All the commercial almond varieties used as references were included in the G-1 group, in three different specific branches of the subcluster G-1A. In particular, the two Spanish cultivars, “Masbovera” and “Marcona”, were found to be moderately distant from one another, instead resulting genetically closer to the French cultivar “Ferragnes” and to the American cultivars “Texas” and “Non Pareil”, respectively. These findings are in agreement with what was observed by [
35], since in that study, too, “Masbovera” skipped clustering with the American cultivars and was grouped together with some Sicilian genotypes. The third American cultivar “Ne plus ultra”, here used as reference, fell into a different branch of clade G-1A, moderately close to the other Americans, similarly to what was reported by [
48]. Some Apulian genotypes appeared to be genetically more closely related to the Spanish and the American commercial varieties than others, probably as a consequence of the gene flow between commercial varieties and local materials. In general, the Italian almond germplasm has proven to show a high level of mixed ancestry among western areas of the Mediterranean basin, mainly due to human migration along the reconstructed maps of ancestral trade routes [
49].
The phylogenetic clusterization of the Apulian genotypes was generally in line with the LRM values obtained, since cultivars with a high allelic similarity were found to belong to the same branch of the tree (
Figure 1). For example, the close genetic distance between “Mollesca di Ruvo” and “Stilla” confirmed that they share the same genetic profile, while “Troito” and “Tuono” seemed to exhibit some genetic differences despite their LRM = 0.5. Strong genetic relationships were also confirmed for “Bianchetta”, “Lunghina” and “Secolare Cotogni” and for “Rachele tenera” and “Del lago”. With the exception of “Rachelina”, that was in the same subgroup G-1B together with “Rachele tenera”, the other genotypes including the term “Rachele” in their name, i.e., “Rachele grande” and “Rachele piccola”, did not show such high genetic similarity, since they fell into two different branches of cluster G-2A. The cultivar “Riviezzo Grosso”, whose LRM value was equal to 0.47, with “Riviezzo Piccolo”, appeared closely related to the cultivar “Gianfreda” (subgroup G-1A). Additionally, both “Riviezzo Grosso” and “Riviezzo Piccolo” were quite distant from the cultivar “Riviezzo”, that was instead included in subgroup G-1B together with “Rachele tenera”, as also observed by [
48].
One group of genotypes (“Ciavea”, “Zin zin”, “Portone Gioia”, “Don Carlo”, “San Michele strada” and “Montranese”), with moderately high LRM values ranging from 0.41 to 0.44, together formed a branch of the tree in cluster G-2A. Interestingly, only four genotypes constituted cluster G-2B, namely the cultivars “Mangini”, “Sciacovelli Altamura”, “Amendolara” and “Casa Perrini”.
Some cases of homonymy were discovered, since genotypes with the same name but genetically polymorphic were clearly distinguished (
Figure 1). Among them, the cultivars “Caputo” and “Caputo_2”, “Irene Lanzolla” and “Irene Lanzolla_2”, “Pizzutella” and “Pizzutella_2”, “San Giuseppe” and “San Giuseppe_2”, “Stivalone” and “Stivalone_2”, “Viscarda” and “Viscarda_2”, and finally “Zanzanello_1” and “Zanzanello_2” are noteworthy, as the components of each pair belonged to clearly different subgroups or to different main clades in some cases, thus resulting phylogenetically distant. Moreover, the case of the two homonymous cultivars “Mollese Grossa” and “Mollese Grossa_2” was also resolved because surprisingly, they were quite distant from one another and clustered in different clades, G-2A and G-1A, respectively. By contrast, “Mollese Grossa” was close to “Mollese di Canneto” and belonged to the same branch of the cultivar “Mollese Spadalunga”, while “Mollese Grossa_2” was grouped together with “Tuono-Troito-Stilla”. Despite the name similarity, other two genotypes named “Mollese Troia” and “Mollese Manfredonia” skipped these groupings and fell into another different cluster, i.e., the subgroup G-1B.
Figure 1.
Dendrogram of the Apulian almond population based on genetic distance. Genotypes names are colored according to the main clusters they belong to (blue for G-1, green for G-2, and orange for G-3). Reference cultivars are colored black.
Figure 1.
Dendrogram of the Apulian almond population based on genetic distance. Genotypes names are colored according to the main clusters they belong to (blue for G-1, green for G-2, and orange for G-3). Reference cultivars are colored black.
2.3. Population Structure
Structure analysis allowed us to determine the genetic constitution of the Apulian almond population. Genotypes whose membership coefficient (qi) was higher than 0.6 were assigned to a defined cluster, otherwise they were considered to be of admixed ancestry. In accordance with the Evanno criterion [
50], a strong signal for K = 3 was obtained (
Figure 2), thus indicating three main, clearly genetically distinct groups that mirrored the genetic distance-based clustering, plus one admixed group (
Figure 3). The first cluster (blue) grouped together 62 genotypes that mostly corresponded to the G-1A clade; the second cluster (orange) consisted of 48 genotypes that included genotypes of the G-1B clade, while the third cluster (grey) included 45 genotypes belonging to the G-2 clade. Again, all the commercial varieties used as references fell into one single cluster (n.1). Reflecting the phylogenetic tree, the cultivars “Tuono” and “Troito”, as well as “Mollesca di Ruvo”, “Stilla” and “Piangente”, grouped together in cluster 2, thus providing further confirmation of their genetic similarity.
Many cases of admixture (32 in total) were evident within and among gene pools, likely as a result of hybrid origin and allele sharing among these genotypes. As an example, the cultivars “Filippo Ceo” and “Falsa Barese” were assigned to the admixed group in this study, highlighting their different allelic composition, as also confirmed by [
35,
48]. Nevertheless, contrasting results may emerge depending on the sample dataset of the collection analyzed and the kind of markers used, as in the research works by [
17,
39], that assumed a common origin for these two cultivars, or else that one could have originated from the other.
The alleles richness of the almond Apulian germplasm, confirmed by the high number of alleles scored, with a relatively congruent number of SSR loci analyzed, reflected the high level of diversification within the Apulian germplasm. This diversification was also verified and confirmed in both the neighbor-joining and structure plot that gathered the genotypes analyzed within clearly distinguishable groups.
Figure 2.
Estimation of the optimum number of clusters for the Apulian almond population according to Evanno’s method. For each K value, the DeltaK is reported, thus indicating that K = 3 is the uppermost probable number of genetically homogenous groups in the collection analyzed.
Figure 2.
Estimation of the optimum number of clusters for the Apulian almond population according to Evanno’s method. For each K value, the DeltaK is reported, thus indicating that K = 3 is the uppermost probable number of genetically homogenous groups in the collection analyzed.
Figure 3.
(
A) Bar plot of all individual almond genotypes generated using STRUCTURE software version 2.3.4 [
51]. For the burning phase, 30,000 iterations were set, followed by 1000 repetitions for K values, ranging from 1 to 10, with 10 runs for K. The three reconstructed subpopulations are distinguished by different colors: blue for cluster 1 (
B), orange for cluster 2 (
C) and grey for cluster 3 (
D). Each genotype is represented by a vertical bar that can be portioned into different colored segments indicating its genetic background. Multiple colors show the admixed genetic constitution of some individuals. Cluster assignment was based on a membership threshold set at >0.6.
Figure 3.
(
A) Bar plot of all individual almond genotypes generated using STRUCTURE software version 2.3.4 [
51]. For the burning phase, 30,000 iterations were set, followed by 1000 repetitions for K values, ranging from 1 to 10, with 10 runs for K. The three reconstructed subpopulations are distinguished by different colors: blue for cluster 1 (
B), orange for cluster 2 (
C) and grey for cluster 3 (
D). Each genotype is represented by a vertical bar that can be portioned into different colored segments indicating its genetic background. Multiple colors show the admixed genetic constitution of some individuals. Cluster assignment was based on a membership threshold set at >0.6.
2.4. Phenotypic Characterization
A subset of 109 cultivars was subjected to a detailed phenotypic characterization. Descriptive statistics of the analyzed traits are shown in
Table 3, including means, minimum, maximum and coefficient of variation (CV), and in
Figure 4, illustrating the distribution of Apulian almonds in terms of the most important agronomic traits. The selected genotypes showed good levels of phenotypic variation as regards all traits, as confirmed by the relatively high coefficient of variation (CV) values (
Table 3), as had also been highlighted by [
16] in a study of a large population of Apulian almonds. Among the traits measured, the highest level of variation was found for nut ventral suture (CV = 89.70%) while kernel tegument color intensity showed the lowest differences among genotypes (CV = 20.78%). Among technological traits, mean values of nut and kernel size were medium (
Table 3 and
Figure 5), but as shown in
Figure 4, some cultivars exhibited values rated in ranges from very small to very large. Nutshell resistance to cracking resulted hard in most of the population examined; the mean number of double kernels was low-medium; all these results are in agreement with [
16]. Flowering time was intermediate in most of the cultivars, but a group of 31 genotypes showed an early blooming time, while another group of 29 genotypes resulted late (
Figure 4). Ripening time in most of the cultivars (62) was intermediate, with only few cultivars exhibiting very early (4) and very late (4) ripening times (
Figure 4).
Principle component analysis assigned most of the traits to four components which explained 46.1% of the total variation (
Table 4). The first two components, which accounted for 27.4% of the total variation, highlighted technological characteristics such as nut size, nut shape, percentage of double kernels in PC1 and nut ventral suture, nutshell softness and kernel size in PC2. The phenological traits showed the highest factor loadings in PC3 (flowering time) and PC4 (ripening time).
Based on the first two components, the PCA plot grouped the almond cultivars according to their technological features resemblance (
Figure 6). Proceeding from positive to negative values of PC1, genotypes were characterized by a lower percentage of double kernels and higher nut size. Proceeding from positive to negative values of PC2, cultivars were defined by a lesser nutshell softness and smaller kernel size. The distribution of almond genotypes on the PC1 and PC2 plots confirmed the wide variability of the samples. A similar wide variability for some of the phenotypic traits analyzed was confirmed by other investigations conducted on a different Apulian almond population [
9,
17].
Table 3.
List of morphological and phenological traits detected in the almond collection and their descriptive statistics.
Table 3.
List of morphological and phenological traits detected in the almond collection and their descriptive statistics.
Trait | Mean | Expression Level | Minimum | Maximum | Coefficient of Variation (CV %) |
---|
Tree habit | 5.11 | medium | 1.00 | 9.00 | 38.35 |
Tree vigor | 6.05 | medium-strong | 1.00 | 9.00 | 25.40 |
Color of petals | 1.27 | white | 1.00 | 3.00 | 43.85 |
Leaf blade color | 5.50 | green | 3.00 | 7.00 | 22.78 |
Nut size | 4.82 | medium | 1.00 | 9.00 | 33.69 |
Nut shape (side view) | 2.59 | round-oblong | 1.00 | 4.00 | 42.80 |
Nutshell color intensity | 4.45 | light-medium | 1.00 | 9.00 | 38.62 |
Nutshell incision (pores) | 2.27 | medium porous | 1.00 | 4.00 | 33.24 |
Nut ventral suture | 1.73 | firmly closed | 1.00 | 5.00 | 89.70 |
Nutshell softness | 2.93 | hard | 1.00 | 7.00 | 45.49 |
Kernel shape | 1.98 | elliptic | 1.00 | 3.00 | 22.76 |
Kernel size | 4.58 | medium | 1.00 | 7.00 | 35.21 |
Kernel tegument color intensity | 5.27 | medium | 3.00 | 7.00 | 20.78 |
Kernel taste | 3.18 | sweet | 3.00 | 7.00 | 26.41 |
Percentage of double kernels | 4.50 | low-medium | 3.00 | 7.00 | 33.72 |
Flowering time | 4.96 | intermediate | 3.00 | 7.00 | 30.03 |
Ripening time | 5.42 | intermediate | 1.00 | 9.00 | 28.87 |
Figure 4.
Distribution of the 109 almond genotypes for morphological and phenological traits scored according to GIBA code.
Figure 4.
Distribution of the 109 almond genotypes for morphological and phenological traits scored according to GIBA code.
Figure 5.
Nut and kernel morphology of some Apulian almond genotypes.
Figure 5.
Nut and kernel morphology of some Apulian almond genotypes.
Table 4.
Eigenvalues and proportion of total variability for the first eight principal components from PCA analysis of the almond genotypes studied.
Table 4.
Eigenvalues and proportion of total variability for the first eight principal components from PCA analysis of the almond genotypes studied.
Traits | PC1 | PC2 | PC3 | PC4 |
---|
Tree habit | 0.09 | −0.13 | −0.50 | 0.07 |
Tree vigor | −0.09 | −0.07 | 0.22 | 0.61 |
Color of petals | −0.35 | −0.47 | 0.30 | −0.25 |
Leaf blade color | 0.31 | −0.07 | −0.25 | −0.26 |
Nut size | −0.63 | 0.49 | −0.25 | 0.16 |
Nut shape (side view) | −0.73 | 0.08 | −0.20 | 0.06 |
Nutshell color intensity | −0.43 | −0.27 | 0.44 | −0.05 |
Nutshell incision (pores) | −0.33 | 0.26 | 0.16 | −0.09 |
Nut ventral suture | −0.21 | 0.50 | 0.29 | −0.23 |
Nutshell softness | 0.30 | 0.54 | −0.10 | −0.41 |
Kernel shape | −0.50 | −0.05 | −0.40 | −0.25 |
Kernel size | −0.23 | 0.61 | −0.24 | 0.43 |
Kernel tegument color intensity | −0.35 | 0.25 | 0.49 | 0.15 |
Kernel taste | −0.16 | −0.27 | 0.33 | 0.05 |
Percentage of double kernels | 0.56 | 0.31 | 0.08 | 0.38 |
Flowering time | −0.41 | −0.29 | −0.47 | −0.07 |
Ripening time | −0.02 | −0.49 | −0.26 | 0.52 |
% of Variance | 14.97 | 12.41 | 10.14 | 8.56 |
Figure 6.
Factor scores of the first two principal components (PCs) for almond genotypes.
Figure 6.
Factor scores of the first two principal components (PCs) for almond genotypes.
The neighbor-joining clustering method utilizing all the morphological and phenological data again subdivided the Apulian almond population into three main clades with comparable dimensions (
Figure 7). Indeed, 39 cultivars belonged to cluster M-1, 37 were included in cluster M-2, and 33 were included in cluster M-3, the least abundant. Moreover, within each cluster it was possible to identify at least two subgroups. The subclusters M-1A and M-3A mostly included almonds with large size nuts and kernels, but the greater nut size did not always correspond to the greater kernel size. Genotypes showing the highest values of shell hardness fell into subgroups M-2A and M-3A. The main cluster M-2 grouped all genotypes with the highest percentage of double kernels. Almonds with the most bitter kernel taste were included in subcluster M-1B.
The observed morphological variability mostly mirrored the results of the molecular characterization obtained with SSR markers, especially the M-1 clade, that was almost exclusively composed of genotypes included in the G-1 clade. Moreover, most of the genotypes with a high genetic similarity according to their LRM values were once again found to belong to the same branches of the morphological tree. For example, the high genetic similarity between the cultivars “Stilla” and “Tuono” (LRM = 0.5), “Bianchetta” and “Lunghina” (LRM = 0.49), “Del lago” and “Rachele tenera” (LRM = 0.48) also emerged from the phenotypic analysis, these genotypes being included in the same morphological subclusters.
However, some exceptions and discrepancies between the morphological and the genetic analysis emerged, as also reported by [
47], which highlighted how some traits, such as kernel oil composition, for example, could be strongly affected by the season, location and climate of the tree growing areas, despite being primarily genotype-dependent. Thus, as expected, also in our study, some genotypes sharing most of the genetic background for the considered microsatellite loci resulted phenotypically rather distant, probably due to their different growing environments. This is, for example, the case of “Mollesca di Ruvo” versus “Tuono” and “Stilla”, that shared high genetic similarity, but in the morphological analysis they fell into different clades. On the contrary, two cases of synonymies, that were clearly genetically resolved by SSR analysis, were found to be close to each other in terms of phenological and morphological characterization, thus explaining the attribution of the same name to these genotypes (this is the case of “Pizzutella” and “Pizzutella_2” and of “Stivalone” and “Stivalone_2”) (
Figure 7).
Figure 7.
Morphological dendrogram generated by the neighbor-joining clustering method using all 17 phenotypic descriptors on a subset of 109 almond genotypes.
Figure 7.
Morphological dendrogram generated by the neighbor-joining clustering method using all 17 phenotypic descriptors on a subset of 109 almond genotypes.