The Genetic Structure of the Field Pea Landrace

: “Roveja di Civita di Cascia” is a landrace of Pisum sativum grown in marginal land habitats of the Apennines, Central Italy, and is one of the eleven herbaceous crop landraces listed in the Regional Register of local varieties. The objective of the present paper was to assess its genetic structure using 62 morphological traits and ﬁve microsatellites. As many as 55 traits showed signiﬁcant di ﬀ erences with the control entries ( P. sativum subsp. sativum var. arvense and var. sativum ). We tested P. sativum Simple Sequence Repeats (SSRs) for their transferability to “Roveja”, and found that only 12 out of 35 performed well. Of these, we demonstrated that ﬁve were su ﬃ cient to assess the genetic structure of this landrace, characterized by several private alleles, di ﬀ erentiating it from Paladio and Bluemoon, which were used as controls. Phenotypic and genotypic data evidenced a genetic structure based on a blend of several pure-bred lines. The sustainability of on-farm landrace conservation is discussed.


Introduction
Pea (Pisum sativum L., family Fabaceae, genus Pisum, 2n = 14) is an annual herbaceous plant predominantly autogamous, native in the Middle East, between the Caucasus and Mesopotamia, with a secondary center of diversification in the Mediterranean Basin and in Ethiopia [1][2][3]. Several archeological evidences date its presence in the Middle East and Central Asia as early as 10,000 B.C., where, together with other legumes and cereals, it represented an important component of the diet of those civilizations [4][5][6]. Pea (field pea, Pisum arvense, and purple pea, P. elatius) and other Neolithic crops reached Europe through the migration or colonization of farmers and shepherds from the Near East [7]. Archeological sites and historical data report evidences of its cultivation since the Stone and Bronze Ages, where it seems to have represented, together with lentil, barley and spelt, the basis of animals and human nutrition [8].
The botanical classification of the genus Pisum has not yet been completely clarified. Originally, it was considered a genus composed by five species, later a monotypic genus, and more recently, a genus with two species [3,[9][10][11][12]. According to the most recently accredited classification, the genus Pisum includes the wild species P. fulvum found in Jordan, Syria, Lebanon and Israel, and the cultivated species P. abyssinicum from Yemen and Ethiopia and P. sativum distributed throughout the world and consisting of both wild (P. sativum subsp. elatius) and cultivated subspecies (P. sativum L. subsp. sativum) [2].
Primer sequence and allele range for validated loci were fed by Multiplex Manager [39] to determine the best sets of loci to include in a multiplex protocol. Multiplex Manager was used with the option of grouping all validated loci within the minimum number of PCRs, avoiding allele range overlap and primer interactions.
PCRs of the five selected loci were carried out with the Type-it Microsatellite PCR Kit (Qiagen, Hilden, Germany) containing 1X Type-it master mix with 0.2 µM of each fluorescent forward primer labeled with 6-FAM or ROX dyes (Sigma, St. Louis, Missouri, United States) and reverse unlabeled primer and 20 ng of template DNA and H 2 O to a final volume of 20 µL. All amplifications were performed in a GeneAmpPCRSystem 9700 (Applied Biosystems, Foster City, California, United States) consisting of a denaturing step at 95 • C for 5 min followed by 30 cycles at 95 • C for 30 s, Ta for 90 s and 72 • C for 30 s, and a final elongation step of 30 min at 60 • C.
PCR products were separated and analyzed on a 3130 XL DNA Analyzer (Applied Biosystems, Foster City, California, United States). The size of the amplified products was determined with an internal standard DNA (GeneScan 500 Liz, Thermo Fischer Scientific, Waltham, Massachusetts, United States) and the scorable peaks were assigned by GeneMapper software (Applied Biosystems, Foster City, California, United States).

Statistical Analyses
Qualitative traits were arranged in contingency tables and the difference among proportions was assessed by Pearson's chi-squared test. Quantitative traits were analyzed by univariate one-way Analysis of Variance (ANOVA). Principal Component Analysis (PCA) was performed using data of all traits and the number of principal components was determined using the minimum eigenvalue criterion recommended by Kaiser [40]. Univariate analyses were carried out by SAS software (Cary, NC, United States) [41], multivariate PCA by PAST software (University of Oslo, Norway) [42].
Prior to any statistical analysis of SSR data, common PCR artifacts leading to genotyping error were investigated. Presence of null alleles, large allele dropout and extreme stuttering was inferred by 1000 bootstraps and a 95% confidence interval (CI) using Micro-Checker 2.2.3 (Norwich Research Park Science, Norwich, United Kingdom) [43].
Bayesian model-based clustering implemented in STRUCTURE 2.3.3 (Pritchard Lab, Stanford University, United States ) [50,51] was performed to assess the genetic structure at the population level, as well as to detect genetic stocks contributing to this germplasm collection. Ancestry model with admixture and correlated allele frequency model was set to get the estimates of the posterior probability of data. STRUCTURE analysis was performed testing 20 independent runs with K from 1 to 21. The length of the burn-in period was set at 100,000, and the number of Markov Chain Monte Carlo (MCMC) repeats after burn-in were set at 500,000. The best K-value was determined through the 1K method [52] by using STRUCTURE HARVESTER 0.6.193 [53]. A membership coefficient qI ≥ 0.8 was used to assign the individuals to clusters. Individuals with membership coefficients qI < 0.8 were grouped together and considered "genetically admixed" [54][55][56][57].

Phenotypic Characterization
The Chi-square test and the analysis of variance showed significant differences among entries in 55 out of 62 traits. Traits with no differences were: (i) Anthocyanin coloration in at least one part of the plant (leaves, stem or seeds), (ii) leaves and pod color (green for all entries), (iii) absence of pods parchment, (iv) presence of pod suture strings and (v) presence of stipule flecking. All details which referred to significant differences among entries are reported in Table A2. Compared to P. sativum var.
sativum Paladio (PS), all landrace lines of "Roveja" were characterized by longer stems, with more nodes, more branches and the presence of anthocyanin coloration.
Moreover, they showed more leaflets per leaf, but shorter and thinner, of different shape (the broadest part of the leaflet was located in the middle rather than at the base) and a more pronounced dentation. Stipules were also significantly smaller, in length and width, and hence, in area. Petioles were significantly shorter, although the variability among the "Roveja" lines was consistent, with some of them (CC_41, 26, 08, 11) not differing from Paladio. Except for lines 58 and 06, the landrace "Roveja" flowered significantly later and showed more flowers per node; the color of the wings was pink to reddish purple and the standards were cream rather than white. Other remarkable differences with Paladio were significantly shorter pods (almost half size) and with thinner walls; their shape at the distal part was essentially blunt while plants of Paladio showed a pointed end. Pod curvature was also slightly more pronounced in "Roveja". The mean number of ovule per pod was significantly lower (7.5 vs. 8.4, respectively), but the pod number per plant was higher (14.0 vs. 5.6, respectively), some of which were empty. In terms of seed starch, all accessions of "Roveja" showed simple grain starch compared with the compound grain starch of Paladio, and this was reflected on cotyledons wrinkling, absent or slight in the former, very marked in the latter (2.0 vs. 8.5, respectively). Other striking differences were on cotyledon color (orange in "Roveja", green and yellow in Paladio) and on testa marbling (very apparent in the former and almost absent in the latter). Interestingly, among the 43 accessions of "Roveja" there were significant differences for the hilum color, ranging from accessions whose color was the same as that of testa (CC_43, 37, 13) up to a very dark color (CC_38, 53,42,26). The different reproductive strategies between "Roveja" and PS, already shown as the number of ovules per pod and pods per plant, were confirmed also in terms of seed yield (4.3 vs. 6.6 g/plant), number of seeds per plants (51 vs. 29), 100-seed weight (10.0 vs. 24.3 g) and seed size (seed surface of 20.9 vs. 45.8 mm 2 ). Bluemoon, used as our control entry, although being a botanical variety arvense, for some traits (antochyanin stem color, stem length, color of the standard, pod wall, cotyledon color, testa marbling and 100-seed weight) was more similar to Paladio, for the others closer to the landrace "Roveja".
The first two principal components obtained from phenotypic traits ( Figure 1) were able to explain almost the entire variance: 95.0 and 4.2%, respectively. The traits whose orthogonal dimensions were able to maximally separate the observations were the dimension of leaflets and of stipules (area, perimeter, length and width), positively correlated with the first component, and the number of pods and of seeds per plant positively correlated with the second component. dentation. Stipules were also significantly smaller, in length and width, and hence, in area. Petioles were significantly shorter, although the variability among the "Roveja" lines was consistent, with some of them (CC_41, 26, 08, 11) not differing from Paladio. Except for lines 58 and 06, the landrace "Roveja" flowered significantly later and showed more flowers per node; the color of the wings was pink to reddish purple and the standards were cream rather than white. Other remarkable differences with Paladio were significantly shorter pods (almost half size) and with thinner walls; their shape at the distal part was essentially blunt while plants of Paladio showed a pointed end. Pod curvature was also slightly more pronounced in "Roveja". The mean number of ovule per pod was significantly lower (7.5 vs. 8.4, respectively), but the pod number per plant was higher (14.0 vs. 5.6, respectively), some of which were empty. In terms of seed starch, all accessions of "Roveja" showed simple grain starch compared with the compound grain starch of Paladio, and this was reflected on cotyledons wrinkling, absent or slight in the former, very marked in the latter (2.0 vs. 8.5, respectively). Other striking differences were on cotyledon color (orange in "Roveja", green and yellow in Paladio) and on testa marbling (very apparent in the former and almost absent in the latter). Interestingly, among the 43 accessions of "Roveja" there were significant differences for the hilum color, ranging from accessions whose color was the same as that of testa (CC_43, 37, 13) up to a very dark color (CC_38, 53,42,26). The different reproductive strategies between "Roveja" and PS, already shown as the number of ovules per pod and pods per plant, were confirmed also in terms of seed yield (4.3 vs. 6.6 g/plant), number of seeds per plants (51 vs. 29), 100-seed weight (10.0 vs. 24.3 g) and seed size (seed surface of 20.9 vs. 45.8 mm 2 ). Bluemoon, used as our control entry, although being a botanical variety arvense, for some traits (antochyanin stem color, stem length, color of the standard, pod wall, cotyledon color, testa marbling and 100-seed weight) was more similar to Paladio, for the others closer to the landrace "Roveja". The first two principal components obtained from phenotypic traits ( Figure 1) were able to explain almost the entire variance: 95.0 and 4.2%, respectively. The traits whose orthogonal dimensions were able to maximally separate the observations were the dimension of leaflets and of stipules (area, perimeter, length and width), positively correlated with the first component, and the number of pods and of seeds per plant positively correlated with the second component. The gray region in Figure 1 circumscribes 95% of all individuals; eight individuals outside it were CC lines (three of which of CC_58) and five of them Paladio (PS_03, 08, 04, 02 and 06). The latter group of plants was scattered along the right side of PC1, characterized by a larger dimension of leaflets and petioles, thus confirming the differences found by univariate statistical analyses. From  dentation. Stipules were also significantly smaller, in length and width, and hence, in area. Petioles were significantly shorter, although the variability among the "Roveja" lines was consistent, with some of them (CC_41, 26, 08, 11) not differing from Paladio. Except for lines 58 and 06, the landrace "Roveja" flowered significantly later and showed more flowers per node; the color of the wings was pink to reddish purple and the standards were cream rather than white. Other remarkable differences with Paladio were significantly shorter pods (almost half size) and with thinner walls; their shape at the distal part was essentially blunt while plants of Paladio showed a pointed end. Pod curvature was also slightly more pronounced in "Roveja". The mean number of ovule per pod was significantly lower (7.5 vs. 8.4, respectively), but the pod number per plant was higher (14.0 vs. 5.6, respectively), some of which were empty. In terms of seed starch, all accessions of "Roveja" showed simple grain starch compared with the compound grain starch of Paladio, and this was reflected on cotyledons wrinkling, absent or slight in the former, very marked in the latter (2.0 vs. 8.5, respectively). Other striking differences were on cotyledon color (orange in "Roveja", green and yellow in Paladio) and on testa marbling (very apparent in the former and almost absent in the latter). Interestingly, among the 43 accessions of "Roveja" there were significant differences for the hilum color, ranging from accessions whose color was the same as that of testa (CC_43, 37, 13) up to a very dark color (CC_38, 53,42,26). The different reproductive strategies between "Roveja" and PS, already shown as the number of ovules per pod and pods per plant, were confirmed also in terms of seed yield (4.3 vs. 6.6 g/plant), number of seeds per plants (51 vs. 29), 100-seed weight (10.0 vs. 24.3 g) and seed size (seed surface of 20.9 vs. 45.8 mm 2 ). Bluemoon, used as our control entry, although being a botanical variety arvense, for some traits (antochyanin stem color, stem length, color of the standard, pod wall, cotyledon color, testa marbling and 100-seed weight) was more similar to Paladio, for the others closer to the landrace "Roveja". The first two principal components obtained from phenotypic traits ( Figure 1) were able to explain almost the entire variance: 95.0 and 4.2%, respectively. The traits whose orthogonal dimensions were able to maximally separate the observations were the dimension of leaflets and of stipules (area, perimeter, length and width), positively correlated with the first component, and the number of pods and of seeds per plant positively correlated with the second component. The gray region in Figure 1 circumscribes 95% of all individuals; eight individuals outside it were CC lines (three of which of CC_58) and five of them Paladio (PS_03, 08, 04, 02 and 06). The latter group of plants was scattered along the right side of PC1, characterized by a larger dimension of leaflets and petioles, thus confirming the differences found by univariate statistical analyses. From Civita di Cascia lines (CC_lin); Sustainability 2019, 11, x FOR PEER REVIEW dentation. Stipules were also significa were significantly shorter, although t some of them (CC_41, 26, 08, 11) not d "Roveja" flowered significantly later a pink to reddish purple and the st differences with Paladio were signifi their shape at the distal part was essen curvature was also slightly more pron significantly lower (7.5 vs. 8.4, respec respectively), some of which were em simple grain starch compared with th cotyledons wrinkling, absent or sli respectively). Other striking differen yellow in Paladio) and on testa mar latter). Interestingly, among the 43 ac hilum color, ranging from accessions a very dark color (CC_38, 53,42,26). T already shown as the number of ovule seed yield (4.3 vs. 6.6 g/plant), numbe g) and seed size (seed surface of 20.9 being a botanical variety arvense, for standard, pod wall, cotyledon color Paladio, for the others closer to the lan The first two principal compone explain almost the entire variance: dimensions were able to maximally s stipules (area, perimeter, length and w number of pods and of seeds per plan dentation. Stipules were also significantly smaller, in length and width, and hence, in area. Petioles were significantly shorter, although the variability among the "Roveja" lines was consistent, with some of them (CC_41, 26, 08, 11) not differing from Paladio. Except for lines 58 and 06, the landrace "Roveja" flowered significantly later and showed more flowers per node; the color of the wings was pink to reddish purple and the standards were cream rather than white. Other remarkable differences with Paladio were significantly shorter pods (almost half size) and with thinner walls; their shape at the distal part was essentially blunt while plants of Paladio showed a pointed end. Pod curvature was also slightly more pronounced in "Roveja". The mean number of ovule per pod was significantly lower (7.5 vs. 8.4, respectively), but the pod number per plant was higher (14.0 vs. 5.6, respectively), some of which were empty. In terms of seed starch, all accessions of "Roveja" showed simple grain starch compared with the compound grain starch of Paladio, and this was reflected on cotyledons wrinkling, absent or slight in the former, very marked in the latter (2.0 vs. 8.5, respectively). Other striking differences were on cotyledon color (orange in "Roveja", green and yellow in Paladio) and on testa marbling (very apparent in the former and almost absent in the latter). Interestingly, among the 43 accessions of "Roveja" there were significant differences for the hilum color, ranging from accessions whose color was the same as that of testa (CC_43, 37, 13) up to a very dark color (CC_38, 53,42,26). The different reproductive strategies between "Roveja" and PS, already shown as the number of ovules per pod and pods per plant, were confirmed also in terms of seed yield (4.3 vs. 6.6 g/plant), number of seeds per plants (51 vs. 29), 100-seed weight (10.0 vs. 24.3 g) and seed size (seed surface of 20.9 vs. 45.8 mm 2 ). Bluemoon, used as our control entry, although being a botanical variety arvense, for some traits (antochyanin stem color, stem length, color of the standard, pod wall, cotyledon color, testa marbling and 100-seed weight) was more similar to Paladio, for the others closer to the landrace "Roveja". The first two principal components obtained from phenotypic traits ( Figure 1) were able to explain almost the entire variance: 95.0 and 4.2%, respectively. The traits whose orthogonal dimensions were able to maximally separate the observations were the dimension of leaflets and of stipules (area, perimeter, length and width), positively correlated with the first component, and the number of pods and of seeds per plant positively correlated with the second component. The gray region in Figure 1 circumscribes 95% of all individuals; eight individuals outside it were CC lines (three of which of CC_58) and five of them Paladio (PS_03, 08, 04, 02 and 06). The latter group of plants was scattered along the right side of PC1, characterized by a larger dimension of leaflets and petioles, thus confirming the differences found by univariate statistical analyses. From dentation. Stipules were also significantly smaller, in length and width, and hence, in area. Petioles were significantly shorter, although the variability among the "Roveja" lines was consistent, with some of them (CC_41, 26, 08, 11) not differing from Paladio. Except for lines 58 and 06, the landrace "Roveja" flowered significantly later and showed more flowers per node; the color of the wings was pink to reddish purple and the standards were cream rather than white. Other remarkable differences with Paladio were significantly shorter pods (almost half size) and with thinner walls; their shape at the distal part was essentially blunt while plants of Paladio showed a pointed end. Pod curvature was also slightly more pronounced in "Roveja". The mean number of ovule per pod was significantly lower (7.5 vs. 8.4, respectively), but the pod number per plant was higher (14.0 vs. 5.6, respectively), some of which were empty. In terms of seed starch, all accessions of "Roveja" showed simple grain starch compared with the compound grain starch of Paladio, and this was reflected on cotyledons wrinkling, absent or slight in the former, very marked in the latter (2.0 vs. 8.5, respectively). Other striking differences were on cotyledon color (orange in "Roveja", green and yellow in Paladio) and on testa marbling (very apparent in the former and almost absent in the latter). Interestingly, among the 43 accessions of "Roveja" there were significant differences for the hilum color, ranging from accessions whose color was the same as that of testa (CC_43, 37, 13) up to a very dark color (CC_38, 53,42,26). The different reproductive strategies between "Roveja" and PS, already shown as the number of ovules per pod and pods per plant, were confirmed also in terms of seed yield (4.3 vs. 6.6 g/plant), number of seeds per plants (51 vs. 29), 100-seed weight (10.0 vs. 24.3 g) and seed size (seed surface of 20.9 vs. 45.8 mm 2 ). Bluemoon, used as our control entry, although being a botanical variety arvense, for some traits (antochyanin stem color, stem length, color of the standard, pod wall, cotyledon color, testa marbling and 100-seed weight) was more similar to Paladio, for the others closer to the landrace "Roveja". The first two principal components obtained from phenotypic traits ( Figure 1) were able to explain almost the entire variance: 95.0 and 4.2%, respectively. The traits whose orthogonal dimensions were able to maximally separate the observations were the dimension of leaflets and of stipules (area, perimeter, length and width), positively correlated with the first component, and the number of pods and of seeds per plant positively correlated with the second component. The gray region in Figure 1 circumscribes 95% of all individuals; eight individuals outside it were CC lines (three of which of CC_58) and five of them Paladio (PS_03, 08, 04, 02 and 06). The latter group of plants was scattered along the right side of PC1, characterized by a larger dimension of leaflets and petioles, thus confirming the differences found by univariate statistical analyses. From Cermis 4767 (CE); Sustainability 2019, 11, x FOR PEER REVIEW dentation. Stipules were also significantly s were significantly shorter, although the va some of them (CC_41, 26, 08, 11) not differin "Roveja" flowered significantly later and sh pink to reddish purple and the standard differences with Paladio were significantly their shape at the distal part was essentially curvature was also slightly more pronounce significantly lower (7.5 vs. 8.4, respectively) respectively), some of which were empty. In simple grain starch compared with the com cotyledons wrinkling, absent or slight in respectively). Other striking differences we yellow in Paladio) and on testa marbling latter). Interestingly, among the 43 accessio hilum color, ranging from accessions whose a very dark color (CC_38, 53,42,26). The dif already shown as the number of ovules per seed yield (4.3 vs. 6.6 g/plant), number of se g) and seed size (seed surface of 20.9 vs. 45 being a botanical variety arvense, for some standard, pod wall, cotyledon color, testa Paladio, for the others closer to the landrace The first two principal components o explain almost the entire variance: 95.0 dimensions were able to maximally separat stipules (area, perimeter, length and width) number of pods and of seeds per plant posit The gray region in Figure 1 circumscr were CC lines (three of which of CC_58) a latter group of plants was scattered along th of leaflets and petioles, thus confirming the d entation. Stipules were also significantly smaller, in length and width, and hence, in area. Petioles ere significantly shorter, although the variability among the "Roveja" lines was consistent, with ome of them (CC_41, 26, 08, 11) not differing from Paladio. Except for lines 58 and 06, the landrace Roveja" flowered significantly later and showed more flowers per node; the color of the wings was ink to reddish purple and the standards were cream rather than white. Other remarkable ifferences with Paladio were significantly shorter pods (almost half size) and with thinner walls; heir shape at the distal part was essentially blunt while plants of Paladio showed a pointed end. Pod urvature was also slightly more pronounced in "Roveja". The mean number of ovule per pod was ignificantly lower (7.5 vs. 8.4, respectively), but the pod number per plant was higher (14.0 vs. 5.6, espectively), some of which were empty. In terms of seed starch, all accessions of "Roveja" showed imple grain starch compared with the compound grain starch of Paladio, and this was reflected on otyledons wrinkling, absent or slight in the former, very marked in the latter (2.0 vs. 8.5, espectively). Other striking differences were on cotyledon color (orange in "Roveja", green and ellow in Paladio) and on testa marbling (very apparent in the former and almost absent in the atter). Interestingly, among the 43 accessions of "Roveja" there were significant differences for the ilum color, ranging from accessions whose color was the same as that of testa (CC_43, 37, 13) up to very dark color (CC_38, 53,42,26). The different reproductive strategies between "Roveja" and PS, lready shown as the number of ovules per pod and pods per plant, were confirmed also in terms of eed yield (4.3 vs. 6.6 g/plant), number of seeds per plants (51 vs. 29), 100-seed weight (10.0 vs. 24.3 ) and seed size (seed surface of 20.9 vs. 45.8 mm 2 ). Bluemoon, used as our control entry, although eing a botanical variety arvense, for some traits (antochyanin stem color, stem length, color of the tandard, pod wall, cotyledon color, testa marbling and 100-seed weight) was more similar to aladio, for the others closer to the landrace "Roveja". The first two principal components obtained from phenotypic traits ( Figure 1) were able to xplain almost the entire variance: 95.0 and 4.2%, respectively. The traits whose orthogonal imensions were able to maximally separate the observations were the dimension of leaflets and of tipules (area, perimeter, length and width), positively correlated with the first component, and the umber of pods and of seeds per plant positively correlated with the second component. The gray region in Figure 1 circumscribes 95% of all individuals; eight individuals outside it ere CC lines (three of which of CC_58) and five of them Paladio (PS_03, 08, 04, 02 and 06). The atter group of plants was scattered along the right side of PC1, characterized by a larger dimension f leaflets and petioles, thus confirming the differences found by univariate statistical analyses. From The gray region in Figure 1 circumscribes 95% of all individuals; eight individuals outside it were CC lines (three of which of CC_58) and five of them Paladio (PS_03, 08, 04, 02 and 06). The latter group of plants was scattered along the right side of PC1, characterized by a larger dimension of leaflets and petioles, thus confirming the differences found by univariate statistical analyses. From Figure 1 is also graphically evident that the total variability of "Roveja" lines was greater than that of CC_ori, the original seed lot, as well as of the CE and CA used as controls. This is important in conserving and managing landraces of inbred species because in our case 43 individuals were sufficient to preserve and retain enough diversity. This is confirmed also by examining for each trait the range of variation reported in Table A2.

Genotypic Characterization
A set of 35 pea genomic and EST SSRs were evaluated for their transferability from P. sativum to "Roveja". Genomic SSR markers are attractive, because they show a high level of polymorphism but, being unlinked to transcribed regions, do not have a defined gene function. On the contrary, EST-SSR are known for their high level of cross species transferability because of being located within transcribed regions of DNA characterized by a low mutation rate; for this reason they are more conserved but less polymorphic [58][59][60]. Although belonging to the same species, the pea SSR showed interesting differences, hereafter described. Out of 35 tested SSRs only 22 (62.8%) displayed clear and repeatable amplicons of expected or of approximate size, a surprising result considering the genetic proximity between the two taxa. Nonetheless, in Roveja only 15 of them were found polymorphic and therefore used for subsequent analysis. The 15 Roveja SSR loci were sequenced and aligned with the corresponding P. sativum sequences stored in the National Center for Biotechnology Information (NCBI) database, and the presence of SSR loci was confirmed in twelve of them (80%), revealing a high degree of conservation of SSR flanking regions in four of them: PeaCLHPPS, AA430902, PSMPB14 and PSGAPA1 ( Figure S1).
Furthermore, the sequences revealed some differences in almost all analyzed loci, due to: (i) variations in length of the repetitive pattern and/or (ii) different repetitive motives (Table 1). For instance, comparing the Roveja and P. sativum motif at locus PSMPSAD237, the former contains a repeat of (AGAT) 33 (AG) 7 , while the latter contains a repeat of ATCT; locus AA321 shows a (TC) 12 motif in the former vs. a (TC) 17 (AC) 15 motif in the latter; and locus PSAD270 contain a (CT) 22 (ATCT) 4 motif vs. a TC motif [25]. Finally, in some loci the same type of repeat resulted conserved across the two taxa. Loci PeaCLHPPS and AA430902 exhibited the same length and repetitive motives in Roveja and in pea, whereas loci PSMPB14 and PSGAPA1 showed differences in the length variations of the microsatellite repeats. On the other hand, it is well known that SSR polymorphism is the result of differences in the number of repeats of the motif caused by polymerase strand-slippage in DNA replication, unequal crossing-over or by recombination errors [61,62]. Therefore, different individuals, and even more, different species, exhibit variations as differences in repeat numbers.
Despite the autogamous nature of the reproductive system and of only five SSR markers used, we found a great variability in terms of the number of alleles among the analyzed loci, particularly for PSMPSAD237 and PSAD270, (with 19 and 17 alleles, respectively, Table 1). This result could be explained by the intrinsic characteristics of SSR markers, in general characterized by high mutation rates [63,64], positively correlated with the length of the repetitive motif rather than to the motif itself. Therefore, SSRs with a large number of repeats, as in our case, are more inclined to mutate due to the increased probability of slippage [62,[65][66][67].
The five nuclear SSRs produced scorable amplicons with a total of 53 alleles showing an average polymorphism of 70%. The average number of alleles per locus was as low as 3.33 (ranging from 1 to 12) and the number of effective alleles per locus was even lower (Ne = 2.22) ( Table 2). were monomorphic at all five loci considered. Considering only the polymorphic loci, the observed heterozygosity (Ho) was nil in 7 cases out of 30, and 17 entries were not in Hardy-Weinberg equilibrium due to the lower proportion of heterozygote compared to the expected. Therefore, despite the fact that the color of the flower is able to attract pollinators, "Roveja" showed high rates of selfing (mean F values 0.69 and F IS = 0.68).
Nevertheless, beside the low gene flow due to the reproductive system, the high level of gene fixation is also attributable to genetic drift, with an F ST value of 0.50, indicating very strong differentiation among entries (Table 3). By examining the result of PCoA (Figure 2), it is also evident that the differentiation is not among entries of "Roveja" (CC_lin, CC_ori, CA and CE), but between these and PS and BM. The two coordinates were able to explain as much as 98% of total variation, with Coordinate 1 able to differentiate the "Rovejas" from BM and Coordinate 2 from PS.  By examining the result of PCoA (Figure 2), it is also evident that the differentiation is not among entries of "Roveja" (CC_lin, CC_ori, CA and CE), but between these and PS and BM. The two coordinates were able to explain as much as 98% of total variation, with Coordinate 1 able to differentiate the "Rovejas" from BM and Coordinate 2 from PS.
The five SSRs were also used to determine the genetic structure among entries. The average log-likelihood values for Ks from 1 to 20 and the distribution of ΔK values [52] indicated two peaks, corresponding to K = 2 and K = 14. The hierarchical genetic structure was investigated at K = 14 and a threshold value qI ≥ 0.80 was used to assign individuals to the clusters (Figure 3).
Out of 253 individuals, 67 (26%) were not classified in any of the 14 clusters and classified in the genetically admixed group (ADM). The group ADM included all individuals from several lines of "Roveja", from CC and CA. With the exception of Cluster 12, all the others included at least one complete set of individuals belonging to the same line/entry ( Figure 3).  dentation. Stipules were also significantly smaller, in length and width, and hence, in area. Petioles were significantly shorter, although the variability among the "Roveja" lines was consistent, with some of them (CC_41, 26, 08, 11) not differing from Paladio. Except for lines 58 and 06, the landrace "Roveja" flowered significantly later and showed more flowers per node; the color of the wings was pink to reddish purple and the standards were cream rather than white. Other remarkable differences with Paladio were significantly shorter pods (almost half size) and with thinner walls; their shape at the distal part was essentially blunt while plants of Paladio showed a pointed end. Pod curvature was also slightly more pronounced in "Roveja". The mean number of ovule per pod was significantly lower (7.5 vs. 8.4, respectively), but the pod number per plant was higher (14.0 vs. 5.6, respectively), some of which were empty. In terms of seed starch, all accessions of "Roveja" showed simple grain starch compared with the compound grain starch of Paladio, and this was reflected on cotyledons wrinkling, absent or slight in the former, very marked in the latter (2.0 vs. 8.5, respectively). Other striking differences were on cotyledon color (orange in "Roveja", green and yellow in Paladio) and on testa marbling (very apparent in the former and almost absent in the latter). Interestingly, among the 43 accessions of "Roveja" there were significant differences for the hilum color, ranging from accessions whose color was the same as that of testa (CC_43, 37, 13) up to a very dark color (CC_38, 53,42,26). The different reproductive strategies between "Roveja" and PS, already shown as the number of ovules per pod and pods per plant, were confirmed also in terms of seed yield (4.3 vs. 6.6 g/plant), number of seeds per plants (51 vs. 29), 100-seed weight (10.0 vs. 24.3 g) and seed size (seed surface of 20.9 vs. 45.8 mm 2 ). Bluemoon, used as our control entry, although being a botanical variety arvense, for some traits (antochyanin stem color, stem length, color of the standard, pod wall, cotyledon color, testa marbling and 100-seed weight) was more similar to Paladio, for the others closer to the landrace "Roveja". The first two principal components obtained from phenotypic traits ( Figure 1) were able to explain almost the entire variance: 95.0 and 4.2%, respectively. The traits whose orthogonal dimensions were able to maximally separate the observations were the dimension of leaflets and of stipules (area, perimeter, length and width), positively correlated with the first component, and the number of pods and of seeds per plant positively correlated with the second component. The gray region in Figure 1 circumscribes 95% of all individuals; eight individuals outside it were CC lines (three of which of CC_58) and five of them Paladio (PS_03, 08, 04, 02 and 06). The latter group of plants was scattered along the right side of PC1, characterized by a larger dimension of leaflets and petioles, thus confirming the differences found by univariate statistical analyses. From Civita di Cascia lines (CC_lin); Sustainability 2019, 11, x FOR PEER REVIEW dentation. Stipules were also significantly s were significantly shorter, although the va some of them (CC_41, 26, 08, 11) not differi "Roveja" flowered significantly later and sh pink to reddish purple and the standar differences with Paladio were significantly their shape at the distal part was essentially curvature was also slightly more pronounc significantly lower (7.5 vs. 8.4, respectively) respectively), some of which were empty. In simple grain starch compared with the com cotyledons wrinkling, absent or slight in respectively). Other striking differences w yellow in Paladio) and on testa marbling latter). Interestingly, among the 43 accessio hilum color, ranging from accessions whose a very dark color (CC_38, 53,42,26). The dif already shown as the number of ovules per seed yield (4.3 vs. 6.6 g/plant), number of se g) and seed size (seed surface of 20.9 vs. 45 being a botanical variety arvense, for some standard, pod wall, cotyledon color, testa Paladio, for the others closer to the landrace The first two principal components o explain almost the entire variance: 95.0 dimensions were able to maximally separat stipules (area, perimeter, length and width) number of pods and of seeds per plant posi The gray region in Figure 1 circumscr were CC lines (three of which of CC_58) a latter group of plants was scattered along th of leaflets and petioles, thus confirming the ntation. Stipules were also significantly smaller, in length and width, and hence, in area. Petioles re significantly shorter, although the variability among the "Roveja" lines was consistent, with me of them (CC_41, 26, 08, 11) not differing from Paladio. Except for lines 58 and 06, the landrace oveja" flowered significantly later and showed more flowers per node; the color of the wings was k to reddish purple and the standards were cream rather than white. Other remarkable ferences with Paladio were significantly shorter pods (almost half size) and with thinner walls; ir shape at the distal part was essentially blunt while plants of Paladio showed a pointed end. Pod rvature was also slightly more pronounced in "Roveja". The mean number of ovule per pod was nificantly lower (7.5 vs. 8.4, respectively), but the pod number per plant was higher (14.0 vs. 5.6, pectively), some of which were empty. In terms of seed starch, all accessions of "Roveja" showed ple grain starch compared with the compound grain starch of Paladio, and this was reflected on tyledons wrinkling, absent or slight in the former, very marked in the latter (2.0 vs. 8.5, pectively). Other striking differences were on cotyledon color (orange in "Roveja", green and llow in Paladio) and on testa marbling (very apparent in the former and almost absent in the ter). Interestingly, among the 43 accessions of "Roveja" there were significant differences for the um color, ranging from accessions whose color was the same as that of testa (CC_43, 37, 13) up to ery dark color (CC_38, 53,42,26). The different reproductive strategies between "Roveja" and PS, eady shown as the number of ovules per pod and pods per plant, were confirmed also in terms of d yield (4.3 vs. 6.6 g/plant), number of seeds per plants (51 vs. 29), 100-seed weight (10.0 vs. 24.3 and seed size (seed surface of 20.9 vs. 45.8 mm 2 ). Bluemoon, used as our control entry, although ing a botanical variety arvense, for some traits (antochyanin stem color, stem length, color of the ndard, pod wall, cotyledon color, testa marbling and 100-seed weight) was more similar to ladio, for the others closer to the landrace "Roveja".
The first two principal components obtained from phenotypic traits ( Figure 1) were able to plain almost the entire variance: 95.0 and 4.2%, respectively. The traits whose orthogonal ensions were able to maximally separate the observations were the dimension of leaflets and of pules (area, perimeter, length and width), positively correlated with the first component, and the mber of pods and of seeds per plant positively correlated with the second component. The gray region in Figure 1 circumscribes 95% of all individuals; eight individuals outside it re CC lines (three of which of CC_58) and five of them Paladio (PS_03, 08, 04, 02 and 06). The ter group of plants was scattered along the right side of PC1, characterized by a larger dimension leaflets and petioles, thus confirming the differences found by univariate statistical analyses. From ation. Stipules were also significantly smaller, in length and width, and hence, in area. Petioles e significantly shorter, although the variability among the "Roveja" lines was consistent, with e of them (CC_41, 26, 08, 11) not differing from Paladio. Except for lines 58 and 06, the landrace veja" flowered significantly later and showed more flowers per node; the color of the wings was to reddish purple and the standards were cream rather than white. Other remarkable rences with Paladio were significantly shorter pods (almost half size) and with thinner walls; shape at the distal part was essentially blunt while plants of Paladio showed a pointed end. Pod ature was also slightly more pronounced in "Roveja". The mean number of ovule per pod was ificantly lower (7.5 vs. 8.4, respectively), but the pod number per plant was higher (14.0 vs. 5.6, ectively), some of which were empty. In terms of seed starch, all accessions of "Roveja" showed le grain starch compared with the compound grain starch of Paladio, and this was reflected on ledons wrinkling, absent or slight in the former, very marked in the latter (2.0 vs. 8.5, ectively). Other striking differences were on cotyledon color (orange in "Roveja", green and w in Paladio) and on testa marbling (very apparent in the former and almost absent in the r). Interestingly, among the 43 accessions of "Roveja" there were significant differences for the m color, ranging from accessions whose color was the same as that of testa (CC_43, 37, 13) up to ry dark color (CC_38, 53,42,26). The different reproductive strategies between "Roveja" and PS, dy shown as the number of ovules per pod and pods per plant, were confirmed also in terms of yield (4.3 vs. 6.6 g/plant), number of seeds per plants (51 vs. 29), 100-seed weight (10.0 vs. 24.3 d seed size (seed surface of 20.9 vs. 45.8 mm 2 ). Bluemoon, used as our control entry, although g a botanical variety arvense, for some traits (antochyanin stem color, stem length, color of the dard, pod wall, cotyledon color, testa marbling and 100-seed weight) was more similar to dio, for the others closer to the landrace "Roveja". The first two principal components obtained from phenotypic traits ( Figure 1) were able to ain almost the entire variance: 95.0 and 4.2%, respectively. The traits whose orthogonal ensions were able to maximally separate the observations were the dimension of leaflets and of les (area, perimeter, length and width), positively correlated with the first component, and the ber of pods and of seeds per plant positively correlated with the second component. The gray region in Figure 1 circumscribes 95% of all individuals; eight individuals outside it e CC lines (three of which of CC_58) and five of them Paladio (PS_03, 08, 04, 02 and 06). The r group of plants was scattered along the right side of PC1, characterized by a larger dimension aflets and petioles, thus confirming the differences found by univariate statistical analyses. From

Cermis 4767 (CE);
Sustainability 2019, 11, x FOR PEER REVIEW dentation. Stipules were also significantly smalle were significantly shorter, although the variabil some of them (CC_41, 26, 08, 11) not differing fro "Roveja" flowered significantly later and showed pink to reddish purple and the standards w differences with Paladio were significantly short their shape at the distal part was essentially blunt curvature was also slightly more pronounced in significantly lower (7.5 vs. 8.4, respectively), but respectively), some of which were empty. In term simple grain starch compared with the compoun cotyledons wrinkling, absent or slight in the respectively). Other striking differences were on yellow in Paladio) and on testa marbling (very latter). Interestingly, among the 43 accessions of hilum color, ranging from accessions whose colo a very dark color (CC_38, 53,42,26). The differen already shown as the number of ovules per pod a seed yield (4.3 vs. 6.6 g/plant), number of seeds p g) and seed size (seed surface of 20.9 vs. 45.8 mm being a botanical variety arvense, for some traits standard, pod wall, cotyledon color, testa mar Paladio, for the others closer to the landrace "Rov The first two principal components obtaine explain almost the entire variance: 95.0 and dimensions were able to maximally separate the stipules (area, perimeter, length and width), posi number of pods and of seeds per plant positively The gray region in Figure 1 circumscribes 9 were CC lines (three of which of CC_58) and fiv latter group of plants was scattered along the rig of leaflets and petioles, thus confirming the differ ion. Stipules were also significantly smaller, in length and width, and hence, in area. Petioles ignificantly shorter, although the variability among the "Roveja" lines was consistent, with f them (CC_41, 26, 08, 11) not differing from Paladio. Except for lines 58 and 06, the landrace ja" flowered significantly later and showed more flowers per node; the color of the wings was o reddish purple and the standards were cream rather than white. Other remarkable nces with Paladio were significantly shorter pods (almost half size) and with thinner walls; hape at the distal part was essentially blunt while plants of Paladio showed a pointed end. Pod ure was also slightly more pronounced in "Roveja". The mean number of ovule per pod was cantly lower (7.5 vs. 8.4, respectively), but the pod number per plant was higher (14.0 vs. 5.6, tively), some of which were empty. In terms of seed starch, all accessions of "Roveja" showed grain starch compared with the compound grain starch of Paladio, and this was reflected on ons wrinkling, absent or slight in the former, very marked in the latter (2.0 vs. 8.5, tively). Other striking differences were on cotyledon color (orange in "Roveja", green and in Paladio) and on testa marbling (very apparent in the former and almost absent in the Interestingly, among the 43 accessions of "Roveja" there were significant differences for the color, ranging from accessions whose color was the same as that of testa (CC_43, 37, 13) up to dark color (CC_38, 53,42,26). The different reproductive strategies between "Roveja" and PS, y shown as the number of ovules per pod and pods per plant, were confirmed also in terms of ield (4.3 vs. 6.6 g/plant), number of seeds per plants (51 vs. 29), 100-seed weight (10.0 vs. 24.3 seed size (seed surface of 20.9 vs. 45.8 mm 2 ). Bluemoon, used as our control entry, although a botanical variety arvense, for some traits (antochyanin stem color, stem length, color of the rd, pod wall, cotyledon color, testa marbling and 100-seed weight) was more similar to o, for the others closer to the landrace "Roveja". he first two principal components obtained from phenotypic traits ( Figure 1) were able to n almost the entire variance: 95.0 and 4.2%, respectively. The traits whose orthogonal sions were able to maximally separate the observations were the dimension of leaflets and of s (area, perimeter, length and width), positively correlated with the first component, and the r of pods and of seeds per plant positively correlated with the second component. he gray region in Figure 1 circumscribes 95% of all individuals; eight individuals outside it C lines (three of which of CC_58) and five of them Paladio (PS_03, 08, 04, 02 and 06). The roup of plants was scattered along the right side of PC1, characterized by a larger dimension lets and petioles, thus confirming the differences found by univariate statistical analyses. From The five SSRs were also used to determine the genetic structure among entries. The average log-likelihood values for Ks from 1 to 20 and the distribution of ∆K values [52] indicated two peaks, corresponding to K = 2 and K = 14. The hierarchical genetic structure was investigated at K = 14 and a threshold value qI ≥ 0.80 was used to assign individuals to the clusters (Figure 3).  Individuals of CC_ori were allocated into several clusters (1,6,7,9,10,11,12 and ADM), similarly to those of CA (cluster 7 and ADM) and CE (cluster 11, 13 and 14). In summary, combining the results of allele frequencies, F-statistics and STRUCTURE, it is evident that the genetic structure of the landrace "Roveja di Civita di Cascia" is based on numerous lines, mostly homozygote at several loci, significantly different from BM and PS, but intermixed with the original population CC_ori and with CA and CE.
The AMOVA showed that the within individual variation was 13%, among entries was 19%, while most of it (68%) was among individuals within entries. In fact, the differences among individuals within entries are likely to be due to homozygous, fixed alleles at different loci.
Out of 53, as many as 26 alleles were private, i.e., unique to the various entries (Table 4), confirming the low gene flow among populations, and at the same time, explaining the goodness of differentiations among entries that STRUCTURE was able to evidence. Individuals of CC_ori were allocated into several clusters (1,6,7,9,10,11,12 and ADM), similarly to those of CA (cluster 7 and ADM) and CE (cluster 11, 13 and 14). In summary, combining the results of allele frequencies, F-statistics and STRUCTURE, it is evident that the genetic structure of the landrace "Roveja di Civita di Cascia" is based on numerous lines, mostly homozygote at several loci, significantly different from BM and PS, but intermixed with the original population CC_ori and with CA and CE.
The AMOVA showed that the within individual variation was 13%, among entries was 19%, while most of it (68%) was among individuals within entries. In fact, the differences among individuals within entries are likely to be due to homozygous, fixed alleles at different loci.
Out of 53, as many as 26 alleles were private, i.e., unique to the various entries (Table 4), confirming the low gene flow among populations, and at the same time, explaining the goodness of differentiations among entries that STRUCTURE was able to evidence.
In summary, the landrace "Roveja di Civita di Cascia" is composed by several pure bred lines and the environmental conditions, the farming system and the mating system are likely to be key factors in determining its genetic structure.

Discussion
Although our landrace Roveja and pea belong to the same species, the results of the present study showed differences concerning phenotypic, genetic and genotypic aspects. In the first case the differences with the control variety Paladio emerged from both univariate and multivariate analyses. In the second case there were differences in the motif of the SSR repetitions, and in the third case, differences in the number of repetitions. These results confirm the on-going debate on the botany of the genus Pisum.
The case of "Roveja di Civita di Cascia" is emblematic, because a number of crops grown in Central Italy have the same history and a similar social context. In the 1960s, many farmers from hilly and mountain areas of South and Central Italy moved to large cities and to industrial areas in the North of the country. This caused a consistent abandonment of marginal lands and a parallel loss of agricultural genetic diversity. Nevertheless, the landraces of several crops remained part of the farming systems due to rural traditions linked to food and feed [68][69][70]. Fortunately, the risk of the genetic erosion of local germplasm was opportunely recognized earlier than other countries [14], and in the 1990s, four Regions of Central Italy (Tuscany, Umbria, Latium and Marche) promulgated laws aimed at collecting and conserving ex-situ and in-situ genetic resources. A detailed inventory of available accessions was published by Negri in 2003 [71], and recent studies confirmed the richness of the agricultural biodiversity of this area due to peculiar soil, climatic and social conditions [15,[72][73][74][75][76]. These laws are concerned with the use of genetic resources in relation to the rural development. In particular, the relationship between genetic resources, territory, typical products and local traditions are a focal point in establishing political measures aiming at preserving crop landraces.
Sustainability in agriculture can be declined in several ways: (i) type and amounts of external inputs (fertilizers, pesticides, water), (ii) suitable exploitation of modern technologies (tools improving water use efficiency, equipment for a precision agriculture), (iii) economic return for a single crop, for a given farm, or for a whole region; (iv) environmentally sound practices, etc. Dealing with living organisms, and particularly under marginal agricultural conditions, sustainability must be entrusted to the genetic variability of plants, animals and microorganisms, paying particular attention to its extent, nature and evolution.
Within this framework the size of the variability of "Roveja di Civita di Cascia" was shown to be adequate, and it is likely to remain so if it continues to be grown according to the Production Regulations [77] approved by the farmer's association. Its genetic variability was evident, and all the same, enclosed within boundaries sufficient to reveal its uniqueness: in four out of five SSR markers we found as many as 10 private alleles. This result is of importance because it grants traceability in case of frauds, and this can be done at low costs. Moreover, the alleles frequencies can be monitored in time, assessing their changes caused by natural evolutive factors (gene flow, mutations, genetic drift, selection) and anthropic interventions. The former are expected to occur normally, but also as a result of the rapid changes of climatic factors. The latter are likely to occur in the event of breeding programs aiming at improving the productivity.
In such circumstances it is recommended to carry out a mass selection, discarding few plants with evident symptoms of diseases and/or with few pods rather than selecting few superior genotypes. In all cases, care must be taken in monitoring the changes in the overall variability, including distinctive traits.
The case examined in the present paper is one example of on-farm conservation of local landraces. The role played by farmers in conserving such valuable genetic resources is difficult to quantify economically. In such cases, public institutions can provide incentives as subsidies to the farmers to safeguard this kind of crops. This form of assistance can guarantee the safeguard in the short period, but it is not sustainable forever, as it is dependent upon public funds. The case of "Roveja di Civita di Cascia" is an example of an efficient germplasm conservation, as nowadays the crop provides enough income to the rural community.
The average production of Roveja is about 0.7-1.0 t ha −1 compared to 1.5-2 t ha −1 of commercial varieties of pea used as feed (e.g., Bluemoon), and the average price of the former is around 15-20 € kg −1 compared to about 20 cents of the latter. Even considering the higher production costs of Roveja (packaging, labeling, marketing, etc.), the economic return for the rural community is remarkable and gives sustainability to the whole system. The promotion efforts put in place were appreciated by the market, so that in 2004 Roveja was awarded the Slow Food Presidium.
Rather than relying on subsidies, the model of "Roveja di Civita di Cascia" should be followed by many other local crops: The higher market prices compared to common products justify not only the safeguarding of the environment and of genetic resources but also local traditions, strenghtening the links of the rural community to its land. Stem: length (cm) StLenght Table A1. Cont.