Dissecting the genetic diversity of cultivated tomato (Solanum lycopersicum) germplasm resources: a comparison of ddRADseq genotyping and microsatellite analysis via capillary electrophoresis and high-resolution melting

Abstract

Understanding the genetic diversity of crops is of fundamental importance for the efficient use and improvement of germplasm resources. Different molecular genotyping systems have been implemented for population structure and phylogenetic relationships analyses, among which, microsatellites (SSRs) and single nucleotide polymorphisms (SNPs) markers have been the most widely used. This study reports the efficacy of SNPs detected via double-digest restriction-site-associated DNA sequencing (ddRADseq) and SSRs analyzed via capillary electrophoresis (CE) and high-resolution melting (HRM) in tomato. In total, 21,020 high-quality SNPs, 20 CE-SSRs, and 17 HRM-SSR markers were assayed in a panel of 72 accessions that included a diversified set of landraces, long-shelf-life cultivars and heirlooms with different origins and fruit typology. The results showed how the population structure analysis was consistent using the three genotyping methods, although SNPs were more efficient in distinguishing cultivar types and in measuring the degree of accessions’ similarity. Compared to CE-SSR, the analysis of microsatellites via HRM yielded a slightly higher number of alleles (98 vs. 96). HRM-SSR demonstrated a distinction between European and non-European germplasm, better resolving the collection’s diversity and being more consistent with SNP data. Phylogenetic trees drawn with independent marker data, detected specific groups of accessions showing robust clusters, highlighting how heirlooms were less heterogeneous than landraces. In addition, the fixation index (F_ST) revealed a high genetic differentiation between heirlooms and long-shelf-life cultivars, with SNP and SSR-HRM data emphasizing the distinction between cherry and plum types and CE-SSR data between cherry and oxheart types. In all instances, a greater molecular variance was found within the different considered biological statuses, provenances, and typologies rather than among them. This work presents the first attempt to compare the three tomato genotyping techniques in tomato. Findings highlighted how the markers used are complementary for genetic diversity analysis, with SNPs providing better insight and HRM-SSR as a viable alternative to capillary electrophoresis to dissect the genetic structure.

Supplementary Information

The online version contains supplementary material available at 10.1007/s13205-024-04141-0.

Introduction

Genetic diversity represents the pool of alleles and genotypes present in organisms and affects their morphology, physiological characteristics and their adaptation to the environment (Tripodi 2023). In plants, discerning the genetic variation present in germplasm collections is essential for both determining strategies for genetic diversity conservation and for effectively utilizing it for breeding and selection. In addition, variety discrimination is trusted a main requirement for recognizing origin trademarks, such as protected designation of origin (PDO) or for implementation of breeders’ rights (Dimitrakopoulou and Vantarakis 2023). Therefore, a trustworthy discriminatory approach is essential when it is difficult to distinguish diversity on the basis of phenotype. Early in the 1980s, biological methods such as allozymes and biochemical markers were implemented for crop genetic studies (Tripodi 2023). The advent of molecular markers has greatly reshaped the field of evolutionary biology and population genetics. Genetic or molecular markers are nucleotide sequences that can be used to track variations or polymorphisms (deletions, insertions, duplications, and translocations) that exist among individuals within specific regions of DNA (Hasan et al. 2021), thus allowing a deep analysis of genetic variation at a much higher resolution than alloenzymes or biochemical markers. The advantages of molecular markers rely on their wide distribution in the genome, the lack of influence of the environment, and the possibility of being assayed in any tissue and developmental stage, offering the ability to compare individuals without bias (Cuyas et al. 2023). Furthermore, molecular markers are powerful for discern the genetic variability in highly related individuals, leading to a greater understanding of the mechanisms underlying the diversity and evolution of populations. An optimal genetic marker should be highly polymorphic, codominant, precise, repeatable, high-throughput, and cost-effective (Amiteye 2021).

Different types of molecular markers pioneered the study of plant genetic diversity and gene mapping including restriction fragment length polymorphism (RFLP), intersimple sequence repeat (ISSR), random amplified polymorphic DNA (RAPD), amplified fragment length polymorphism (AFLP), simple sequence repeats (SSRs), start codon targeted (SCOT) (Amiteye 2021; Rai 2023). SSRs are DNA sequences consisting of a short tandem repeat motif of 1–6 base pairs with lengths varying between 5 and 40 repeats (Geethanjali et al. 2024). Given their multiallelic nature, the high variability and reproducibility, as well as the possibility of automation in capillary electrophoresis, these markers have been widely used for genetic diversity studies in several plant species (Geethanjali et al. 2024; Tripodi 2023). In recent years, thousands of microsatellites have been discovered thanks to progress of next-generation sequencing technologies (NGS) that enabled the release of comprehensive transcriptomes and genomes (Le Nguyen et al. 2019). At the same time, NGS projects have enhanced the understanding of variations in genome sequences throughout discovery of millions of single nucleotide polymorphisms (SNPs) (Tripodi 2022). The biallelic nature of SNPs makes them less informative when compared to SSRs which are able to detect more than six alleles per locus (Heffernan et al. 2024). In contrast, owing to their very high frequency, their uniform distribution in the genome and low mutation rate compared to microsatellites, SNPs are better suited for automation and standardization in high-throughput technologies thus, allowing the generation of a larger amount of data (Singh et al. 2023; Tripodi 2022). Genotyping by sequencing (GBS) and restriction-site-associated DNA sequencing (RADseq) are the most popular and low-cost strategies for genome-wide SNP discovery and large-scale genotyping in crops (Wang et al. 2024). Both are known as reduced-representation approaches relying on the use of restriction enzymes to fragment the genome, followed by high-throughput sequencing of the fragments. Double-digest restriction-site-associated DNA sequencing (ddRADseq) is a modification of RAD sequencing that uses two restriction enzymes, instead of one, to reduce genome complexity (Magbanua et al. 2023). The combination of two enzymes increases the number of restriction sites in the genome and results in a greater diversity of fragment sizes, which can improve both marker resolution and genome coverage. GBS, RADseq and ddRADseq have been applied to a wide range of application, being used for genome-wide association and QTL mapping studies (Hossain et al. 2019; Tripodi et al. 2021), construction of genetic maps (Calayugan et al. 2024; Scheben et al. 2020), genotyping of large crop collections (Esposito et al. 2020; Yamashita et al. 2019). Another promising approach for investigating the genetic variability among individuals is high-resolution melting (HRM) analysis (Gatica-arias et al. 2023; Tsanakas et al. 2018). HRM detects genetic variants measuring the difference in the melting of double-strand DNA. It is based on the principle that the melting temperature (Tm) of a DNA fragment is affected by its base composition and sequence (Vossen et al. 2009). Using a saturated DNA intercalating dye and a real-time detection system, it is possible to capture the fluorescence change of melting curve underlying the allelic differences among individuals. This method is efficient, accurate and cost-effective to genotype both SSRs and SNPs (Heo and Chung 2020). HRM has been used both for gene mapping (Cho et al. 2022), food traceability (Lagiotis et al. 2020; Pereira et al. 2018) and evaluation of the diversity of different crop species (Anthoons et al. 2022). The application of HRM for the analysis of microsatellites has also been reported (Heo et al. 2019).

Tomato (Solanum lycoperiscum L.) is one of the most consumed vegetable crops and a main staple for human nutrition as it is rich in antioxidant bioactive compounds and primary source of lycopene (Atherton and Rudich 2012). After the discovery of the New World in the sixteenth century, tomato has been spread from its native area in Central-South America’s first in Europe and then in the rest of the world (Razifard et al. 2020), where the selection processes led to the development of a variegated panorama of landraces, heirlooms, and improved varieties adapted to different uses and/or market destinations (Esposito et al. 2020). This diversity is reflected in the color, shape, and size of the berries. Currently, it is estimated that more than 10,000 varieties are part of the tomato gene pool (Tripodi 2022). Understanding the molecular diversity is essential for their preservation, exploitation, and utilization. To that end, disparate studies have been performed via microsatellites (Castellana et al. 2020; Pozharskiy et al. 2023), GBS or ddRADseq methodologies (Blanca et al. 2022; Esposito et al. 2020) in tens or hundreds of accessions. However, fewer attempts have focused on the use of HRM. Various studies have compared the consistency of SNP and SSR marker data for assaying the genetic relationships in crop species. Examples have been reported in Arabidopsis (Fisher et al. 2017), alfalfa (Annichiarico et al. 2015), and sunflower (Filippi et al. 2015). In addition, microsatellites and HRM have been applied in combination for food forensic studies (Bosmali et al. 2012; Chedid et al. 2020) and genotype identification (Gomes et al. 2018).

Our primary goal was to assess the assaying ability of capillary electrophoresis with fluorescence detection, high-resolution melting analysis, and ddRADseq for SNP discovery in tomato in terms of its ability to measure genetic diversity, population structure, and phylogenetic relationships. We tested 72 phenotypically different accessions mostly local varieties, and heirlooms. To the best of our knowledge, no research has evaluated the advantages and disadvantages of these three genotyping technologies for the study of germplasm diversity in tomato. This work represents the first attempt to address this gap. The three marker systems’ prospects, outcomes, and consistency are as follows: described and discussed.

Materials and methods

Plant material

The plant material consisted of 72 cultivated tomato (S. lycopersicum) accessions held at the genebank of the Research Centre for Vegetable and Ornamental Crops (CREA, Pontecagnano, SA; Monsampolo del Tronto, AP; Italy) (Table 1). The collection included landraces, long-shelf-life cultivars and heirlooms belonging to different varietal types (beefsteak, cherry, globe, oxheart, and plum) with diverse origins (Fig. 1). Seeds were sown in nursery seed trays with peat and seedlings grown in a glasshouse at a controlled temperature (min 16°C–max 28 °C).

Table 1.

Details of the 72 tomato (S. lycopersicum) genotypes analyzed in the present study

Name	Code*	Provenance	Name	Code*	Provenance
Black Tula	HL_B1	Russia	Pomo Max	LR_C3	Italy
Mexican Ribbed	HL_B2	Mexico	del Vesuvio	LR_C4	Italy
Apricot Ribbed	HL_B3	na	7536	LR_C5	Italy
Tangerine Ribbed	HL_B4	United Kingdom	Giallo Pro	LR_C6	Italy
Art pink tiger	HL_C1	na	Pom Nero Pro	LR_C7	Italy
Sun Black	HL_C2	na	Malareto	LR_C8	Italy
Snow white	HL_C3	United States	Giallo a grappoli	LR_C9	Italy
Gajo de melon	HL_C4	United States	99,190	LR_C10	Italy
Black Cherry	HL_C5	United States	Varrone	LR_G1	Italy
Rutgers I	HL_G1	United States	Laura	LR_G2	Italy
Ailsa Craig LA3174	HL_G2	United States	Gaetani	LR_O1	Italy
Black truffle	HL_G3	United States	Belmonte Max	LR_O2	Italy
Blanche du Quebe	HL_G4	Canada	Rita2	LR_O3	Italy
Violet Jasper	HL_G5	China	Noc Corbarino	LR_P1	Italy
Siberienne Rose	HL_G6	Russia	Nocerino Nocera	LR_P2	Italy
Ailsa Craig LA3193	HL_G7	United States	Pizzutello Pin	LR_P3	Italy
Michael Pollan	HL_P1	United States	San Marzano 622	LR_P4	Italy
Cream sausage	HL_P2	United States	Fiaschetto a mandorla	LR_P5	Italy
Pomo Luca	LR_B1	Italy	San Marzano 113	LR_P6	Italy
Rotonda Ligure	LR_B2	Italy	LSM Pro	LR_P7	Italy
Cost Fiorentino Novoli	LR_B3	Italy	San Marzano LPRO	LR_P8	Italy
Marmande FTR	LR_B4	France	Campano Bal	LR_P9	Italy
Genovese	LR_B5	Italy	Piennolo Pro 2	LSL_C1	Italy
Pera loc 2	LR_B6	Italy	Vesuvio SM small	LSL_C10	Italy
Vomano Agrif	LR_B7	Italy	DelVesuvio 25	LSL_C11	Italy
Valdaso Basili	LR_B8	Italy	Regina Ostuni	LSL_C12	Italy
Pera Abr1	LR_B9	Italy	Piennolo Pugliese	LSL_C2	Italy
Pisanello	LR_B10	Italy	Regina	LSL_C3	Italy
Pomodoro Stella Pisa	LR_B11	Italy	Molese	LSL_C4	Italy
Gigante RR	LR_B12	Italy	Molteno	LSL_C5	Italy
Ponderosa	LR_B13	Italy	RedPear	LSL_C6	Italy
Marmande Cap	LR_B14	France	Piennolo21	LSL_C7	Italy
Marmande Csmall	LR_B15	France	Pop25	LSL_C8	Italy
Cost Fiorentino	LR_B16	Italy	VesuvioPOP8	LSL_C9	Italy
Nero Pro	LR_C1	Italy	Nocerino	LSL_P1	Italy
Cherry Pro	LR_C2	Italy	SMGAR	LSL_P2	Italy

Code: accession code including details of biological status and varietal types (see legend); Provenance = Country of origin

^*HL Heirloom, LR Landrace, LSL long-shelf-life cultivar, B beefsteak, C cherry, G globe, O oxheart, P plum

Fig. 1.

Overview of the phenotypic diversity of the collection studied. Accessions with different biological status were considered including heirlooms (HL), landraces (LR) and long-shelf-life cultivars (LSL). Fruit morphology include beefsteak (B), cherry (C), globe (G), oxheart (O) and plum (P) types. More details of the accessions displayed in the figure are reported in Table 1

DNA isolation and ddRAD sequencing

Genomic DNA was prepared according to the procedure described in Taranto and collaborators (2016). For each genotyping assay, the same batch of DNA was used. The 72 accessions’ SNP data using ddRAD sequencing were obtained from the previously disclosed dataset (Esposito et al. 2020). The whole matrix consisting of 246,936 SNPs was filtered applying a MAF of 0.05 and retaining loci that were represented in at least 90% of the population. In total, 21,020 high-quality SNPs were selected and used for downstream analysis.

Capillary electrophoresis microsatellite detection

A total of 20 microsatellites were selected in this study for assaying genetic diversity (Table 2) (Areshchenkova, and Ganal 2002; Bae et al. 2010; He et al. 2003; Smulders et al. 1997). SSR genotyping was performed via amplification of the template DNA with forward primers 5′-end labeled with the fluorescent dye FAM and reverse primers unlabeled. Amplification of the 72 accessions was performed in 96-well PCR (polymerase chain reaction) plates, with each reaction containing 15 ng of genomic DNA, 2 µL of 5X Colorless GoTaq^® Buffer, 0.2 µM of each primer, 0.2 mM of each dNTP, and 1.5 units of GoTaq^® G2 DNA Polymerase (Promega, Madison, WI, USA) in a total volume of 10 µL. PCR amplifications consisted of denaturation at 94 °C for 2 min, followed by 34 cycles of three steps at 94 °C for 30 s, specific T° annealing for 30 s, and 72 °C for 45 s, final extension of 72 °C for 7 min and soaking at 12 °C. Reactions were carried out in a C-1000 Touch Thermal Cycler (Bio-Rad Hercules, CA, USA). The PCR products were separated by electrophoresis in 2% agarose gels. After PCR, 1:40 dilution was used for capillary electrophoresis. Then, 1 µL of the diluted amplicons was mixed with 0.3 µL of the GeneScan-600 LIZ (GS 600) size standard (Thermo Fisher Scientific, Waltham, MA, USA) and 13.7 µL of Hi-Di formamide (Thermo Fisher Scientific, Waltham, MA, USA). Fragment analyses were carried out in a SeqStudio™ Genetic Analyzer (Thermo Fisher Scientific, Waltham, MA, USA). Capillary fragment separation was analyzed with GeneMapper™ Software v.6. Signal peak height and allele size, were calculated on the basis of the GS 600 molecular weight standard.

Table 2.

Forward and reverse primer sequences, annealing temperatures and alleles detected for the microsatellites used to genotype 72 tomato accessions through a capillary electrophoresis assay and high-resolution melting analysis

SSR name	Forward primer	Reverse primer	Alleles detected*	Tm (°C)
LEat008a	AAGCGCGAGCTCTCTCTGATCTC	CCACGATCTCCGCCATATGC	92, 99, 100	55 °C
LEct001b	TCCAATTTCAGTAAGGACCCCTC	CCGAAAACCTTTGCTACAGAGTAGA	98, 99, 101, 102, 107, 108	53 °C
LEta003b	GCTCTGTCCTTACAAATGATACCTCC	CAATGCTGGGACAGAAGATTTAATG	103, 104, 105, 106, 108, 108	53 °C
LEta015b	ATATGCATGGACAAATCTTGAGGG	CTCGCGCATCAAATTAATGTATCAG	101, 103, 105, 107, 109, 112	55 °C
LEaat002b	GCGAAGAAGATGAGTCTAGAGCATAG	CTCTCTCCCATGAGTTCTCCTCTTC	95, 96, 102, 102, 104, 106, 107	53 °C
LEaat007b	CAACAGCATAGTGGAGGAGG	TACATTTCTCTCTCTCCCATGAG	89, 96, 97, 98, 101, 102	53 °C
LEaac001b	AGGAAGAGCGTGAGTCTGAAC	TCCTGCGCCACTTTAGAG	106, 107	55 °C
LEctt001b	CCTCTCTTCACCTCTTTACAATTTCC	CACTGGTCATTAAGTCTACAGCC	90, 90, 91, 97, 97, 100	55 °C
LEtat002b	ACGCTTGGCTGCCTCGGA	AACTTTATTATTGCCACGTAGTCATGA	192, 195, 198,	55 °C
LE20592b	CTGTTTACTTCAAGAAGGCTG	ACTTTAACTTTATTATTGCCACG	162, 164, 165, 167, 169	53 °C
LE21085b	CATTTTATCATTTATTTGTGTCTTG	ACAAAAAAAGGTGACGATACA	96, 99, 101, 102, 116, 117, 119, 120	50 °C
LECBPE3c	CCTACAAAAACTGCCTCT	TTATATCAATACAACAACATT	115, 117	56 °C
LEDIH4REc	TTTTGTAATCATCTTGGAAAC	ATTGTGTTATGATGATATTTG	86	55 °C
LELAT59c	AACAACATTTCACAAAGTGCT	CGTCTCAATGAGACAACAAGT	67, 68, 69, 70	55 °C
LELEUZIPc	GGTGATAATTTGGGAGGTTAC	CGTAACAGGATGTGCTATAGG	99, 102, 103,	55 °C
LEPRP4c	TTCATTTCTTGCAACTACGAT	CATACTAGCAACATCAAAGGG	198, 199, 229, 230	55 °C
TC1107a	TCCATCTCTCTCTAGACCTTTCT	TTCTTAAATCCTCTCACTCA	87, 89, 94, 95	56 °C
TCIC-419d	TGAGCAACATAAATGCATGTATGGC	AGAACAACTGTAGTGGTTCCATCACC	96, 101, 103, 105, 106, 444, 445, 446	55 °C
TMS56a	GATCTCAAAGGATGAACAATAC	TCATTAGGAGATTCTTTGTATCA	121, 121, 122, 123, 124, 125, 126	55 °C
TMS59a	TGAACGGGCCTTCTGTTATC	ATCATCATTATAGTTCTTAAGTGAT	100, 100, 101, 102, 108	55 °C

a = Areshchenkova and Ganal (2002), b = He et al. (2003), c = Smulder et al. (1997), d = Bae et al. (2010)

^*Via capillary electrophoresis

High-resolution melting analysis

SSR-HRM analysis was performed via a CFX96 Real-Time PCR System (Bio-Rad, Inc., Hercules, CA, USA). For each marker, a reaction mixture containing 2 µL (15 ng/µL concentration) of genomic DNA, 5 µl of 2 × Precision Melt Supermix (Bio-Rad), 0.4 µL of 200 nM of both forward and reverse nonlabeled primers and distilled water were added to reach the final volume of 10 µL. HRM-PCR was conducted in a 96-well Bio-Rad instrument with a thermal cycling temperature of 95 °C for 2 min, followed by 39 cycles of 95 °C for 15 s, primer annealing for 20 s, and 72 °C for 40 s. The amplification was immediately followed by the high-resolution melting steps: 95 °C for 30 s and 64 °C for 1 min, then the HRM increased from 65 °C to 95 °C with increments of 0.2 °C/cycle every 10 s. Precision Melt Analysis ™ Software (Bio-Rad, Hercules, CA, USA) was used to analyze the melting profile and genotype discrimination. Melting curves were then generated and clustered into various genotype groups. Clusters were assigned on the basis of a percent confidence threshold > 95%.

Data analysis

A genetic diversity summary of the SNP matrix was performed via the Geno Summary tool implemented in Tassel v5.2.15 (Bradbury 2007). Considering the biallelic nature of the SNPs, the expected heterozygosity according to the Hardy–Weinberg equilibrium (H) was calculated according to the formula:

$$H=1-p^2-q^2$$

where p and q each represent the frequency of the different alleles for each SNP.

The efficiency of microsatellites via capillary electrophoresis and high-resolution melting analysis was measured by calculating different statistical parameters including the polymorphic information content (PIC), effective multiplex ratio (EMR), marker index (MI), discriminating power (DP) and resolving power (RP). The analyses were implemented in the iMEC program (Amiryousefi et al. 2018).

The PIC represents the marker’s capacity to identify polymorphisms in the population (Botstein et al. 1980), thus providing level of informativeness. The index was estimated according to the formula:

$$\text{PIC}=2f\times \left(1-f\right)$$

where f is the frequency of the polymorphism present at a given locus, and (1 − f) is the frequency of an absent polymorphism for each locus.

The EMR (Powell et al 1996) is defined as the product of the number of polymorphic loci (npl) for each marker and their percentage over the total number of loci (ntl):

$$\text{EMR}=\text{n}pl\times \frac{\text{n}pl}{\text{n}tl}$$

MI (Powell et al. 1996) estimates the overall efficiency of the molecular marker assay and is calculated according to the following formula:

$$\text{MI}=\text{PIC}\times \text{EMR}$$

DP (Tessier et al. 1999) is the likelihood that two randomly selected individuals exhibit different banding patterns and, therefore, are discernible from the others:

$$\text{DP}=1\times \text{C}p$$

where Cp is the confusion probability calculated as Cp = Σ ci = Σ pi Npi-1/N − 1, which consider the i-th pattern of the given j-th primer, present at frequency pi in a set of varieties, with N individuals. Cp is equal to the sum of all ci for all of the patterns generated by the marker.

The RP (Prevost and Wilkinson 1999) is the ability of a marker to distinguish between large number of genotypes and was calculated as

$$\text{RP}=\mathrm{\Sigma }\times \text{Ib}$$

where Ib represents the polymorphism informativeness, which takes values of 1 − [2 (0.5 − p)] with p being the proportion of each genotype containing the polymorphism.

For SNP markers, genetic structure was determined via the model-based ancestry estimation obtained with ADMIXTURE software (Alexander et al. 2015) with K values ranging from 1 to 10. Tenfold cross-validation (CV) procedure with five iterations was performed, and CV scores were used to determine the best K value. For the CE-SSR and HRM-SSR, the population structure was inferred via Bayesian-based cluster algorithm implemented in the software STRUCTURE version 2.3.4 (Pritchard et al. 2000). The admixture model analysis, assuming correlation among allele frequencies, was considered. The Markov chain Monte Carlo (MCMC) method was used for allele frequency estimation and identification of the best number of populations (K). Runs were performed with 50,000 burn-in cycles and 50,000 MCMC iterations with number of genetic clusters (K) from 1 to 10 and 5 independent runs for each K. The most likely numbers of K were determined via the StructureSelector freeware (Li and Liu 2018). Accessions were considered to belong to a specific K cluster if its membership coefficient (qi) was ≥ 0.50, whereas the genotypes with qi values lower than 0.5 at each assigned K were considered admixed. A phylogenetic tree was drawn via the neighbor joining method and the maximum likelihood model with 1000 bootstraps. Analyses were conducted in MEGA X software (Kumar et al. 2018). Principal component analysis (PCA) for SNPs and SSRs was performed in Tassel v5.2.15, and a biplot was drawn via ggplot2 (Wickham 2016). GenAlEx V6.5 (Peakall et al. 2012) was used to calculate the molecular variance (AMOVA) and genetic divergence (F_ST) as factors of variation among and within the populations for the three genotyping methods used. AMOVA was performed with 999 permutations to test for significance. To study the correlations between pairs of molecular datasets, the Mantel test using Pearson’s r value and 10,000 permutations was performed (Mantel 1967).

Results

Single-nucleotide polymorphism analysis

The 21,020 high-quality SNPs identified via ddRADseq spanned the entire genome at an average density of 38.64 kb across the 12 chromosomes (Chr), ranging from 12.42 kb (Chr 5) to 17.09 kb (Chr 08) (Fig. 2a). About half of the SNPs were identified on chromosomes 4 and 5, whereas 5.87% of them were located on chromosome 0 (1235) (Fig. 2b). In total, 56 gap regions with sizes > 1 Mb were found on all chromosomes, with the highest number of gaps on chromosome 8. The gap regions did not exceed 3 Mb with the largest region being 2.96 Mb in size on chromosome 8. Overall, 91% of the SNPs were positioned at a distance lower than 1 Mb. Across the whole set, the PIC values ranged from 0.033 to 0.375 (Fig. 2c), with a mean of 0.240. The minimum average PIC value was encountered on Chr 5 (0.118), whereas the maximum value was found on Chr 12 (0.299). On average, the heterozygosity was 0.255, reaching values above 0.300 only on chromosomes 1, 8 and 12 (Fig. 2d). The observed transition/transversion ratio was 1.29. Among the transition events, A > G and C > T were the most abundant (18.697% and 18.225%, respectively), whereas C > A and G > T were the most represented transversion events (6.52% and 6.23%, respectively).

Fig. 2.

SNPs detected through double-digest restriction-site-associated DNA sequencing: a Distribution of 21,020 SNPs across the 12 tomato chromosomes. The number of SNPs is represented within a 1 Mb window size. The horizontal axis shows the chromosome (Chr) length (Mb); each bar represents a chromosome, with Chr 1 at the top and Chr 12 at the bottom. The different colors depict SNP density following the gradient in the legend on the right; b bar chart illustrating the overall number of SNP markers on each chromosome; c bar chart showing the average polymorphic information content (PIC) for each chromosome; d heterozygosity level for SNP markers in the 12 chromosomes

Microsatellite analysis

For genetic diversity analysis, 20 microsatellites were used in the present study (Table 2). Among these, 17 SSRs produced clear genotyping profiles generated from both HRM and fragment analysis by capillary electrophoresis (CE-SSR). The markers LECBPE3 and TC1107 did not provide clear HRM profiles; therefore, only the CE-SSR results were considered. Furthermore, LEDIH4RE was monomorphic for the 72 accessions studied and thus was excluded from the final data analysis.

Through capillary electrophoresis analysis, a total of 96 alleles were detected with an average number of 4.80 alleles per locus. The markers TCIC-419 and LE21085 presented the maximum number of alleles (8), while the minimum number of alleles was observed for LEaac001 and LECBPE3 (2) (Table 3). The mean expected (He) and observed (Ho) heterozygosity were 0.383 and 0.040, respectively, indicating an overall medium level of diversity. Among the 19 SSRs considered, the mean PIC was 0.307, with values ranging from 0.244 for LE21085 to 0.50 for LECBPE3. The marker LEPRP4 showed highest EMR and MI, with values of 2.775 and 0.929, respectively. The lowest EMR was found for LELEUZIP and TMS59 (1.014), whereas for MI, LEaat007 presented the lowest value of 0.267. The highest RP was found for LEta015 (2.611), while the lowest values were shown by LELEUZIP (0.085).

Table 3.

Number of loci (N°), expected heterozygosity index (He), average heterozygosity (H.av), polymorphic index content value (PIC), effective multiple ratio (EMR), marker index (MI), discrimination power (DP), and resolving power (RP) obtained from the SSRs used in the present work via capillary electrophoresis (CE) and high-resolution melting HRM) analysis

Marker	CE-SSR								HRM-SSR
	N°	He	Ho	PIC	EMR	MI	DP	RP	N°	He	H.av	PIC	EMR	MI	DP	RP
LE20592	5	0.378	0.039	0.306	1.264	0.387	0.937	1.917	5	0.320	0.033	0.269	1.000	0.269	0.961	0.508
LE21085	8	0.285	0.030	0.244	1.375	0.336	0.971	1.250	7	0.245	0.025	0.215	1.000	0.215	0.980	0.611
LEaac001	2	0.411	0.043	0.326	1.423	0.464	0.496	0.901	4	0.375	0.038	0.305	1.000	0.305	0.938	0.222
LEaat002	7	0.318	0.033	0.267	1.389	0.372	0.961	2.444	9	0.198	0.020	0.178	1.000	0.178	0.988	1.889
LEaat007	6	0.293	0.031	0.250	1.069	0.267	0.969	1.694	7	0.245	0.025	0.215	1.000	0.215	0.980	1.943
LEat008	3	0.464	0.048	0.356	1.097	0.391	0.867	1.417	3	0.460	0.047	0.367	0.985	0.362	0.893	0.147
LECBPE3#	2	0.500	0.052	0.375	1.029	0.386	0.737	0.171	–	–	–	–	–	–	–	–
LEct001	6	0.338	0.035	0.281	1.292	0.363	0.954	2.306	7	0.245	0.025	0.215	1.000	0.215	0.980	2.000
LEctt001	6	0.440	0.046	0.343	1.958	0.672	0.894	0.306	4	0.375	0.038	0.305	1.000	0.305	0.938	0.222
LEDIH4RE*#	1	0	0	0	1	0.000	0	0	–	–	–	–	–	–	–	–
LELAT59	4	0.465	0.048	0.357	2.528	0.902	0.601	1.111	4	0.375	0.038	0.305	1.000	0.305	0.938	1.167
LELEUZIP	3	0.448	0.047	0.347	1.014	0.352	0.887	0.085	3	0.460	0.047	0.366	0.986	0.361	0.893	0.139
LEPRP4	4	0.425	0.044	0.335	2.775	0.929	0.520	2.451	6	0.278	0.028	0.239	1.000	0.239	0.973	1.111
LEta003	6	0.313	0.033	0.264	1.167	0.308	0.963	2.333	5	0.320	0.033	0.269	1.000	0.269	0.960	1.611
LEta015	6	0.340	0.035	0.283	1.306	0.369	0.953	2.611	13	0.142	0.014	0.132	1.000	0.132	0.994	2.000
LEtat002	3	0.450	0.047	0.349	1.028	0.359	0.884	1.444	4	0.375	0.038	0.305	1.000	0.305	0.938	0.507
TC1107#	4	0.379	0.039	0.307	1.015	0.311	0.936	0.324	–	–	–	–	–	–	–	–
TCIC-419	8	0.398	0.041	0.319	2.194	0.700	0.925	2.056	10	0.180	0.018	0.164	1.000	0.164	0.990	2.000
TMS56	7	0.301	0.031	0.256	1.292	0.330	0.966	2.583	2	0.500	0.051	0.375	1.000	0.375	0.752	0.556
TMS59	5	0.323	0.034	0.271	1.014	0.275	0.959	1.887	5	0.320	0.033	0.269	1.000	0.269	0.960	0.909
Total	96.0	7.27	0.76	5.84	27.23	8.47	16.38	29.29	98.0	5.412	0.552	4.491	16.97	4.480	16.06	17.54
Average	4.80	0.383	0.040	0.307	1.433	0.446	0.862	1.542	5.76	0.318	0.032	0.264	0.998	0.264	0.944	1.032

*Monomorphic

^#Not suitable for HRM analysis

SSRs via high-resolution melting analysis detected a similar number of patterns (98) on the basis of melting curves, with an average number of distinctive profiles per locus of 5.76. The level of heterozygosity was lower but comparable to values reported by CE-SSR. The two systems detected similar numbers of alleles (Fig. 3) with substantial differences observed for LEta015 and TM56, which presented greater (13) and lower (2) numbers of alleles, respectively. For the statistical marker parameters, no values greater than those observed with capillary analysis were identified, except for DP. Indeed, the average PIC was 0.264, with values ranging from 0.132 for LEta015 to 0.375 for TMS56. These two markers presented the same highest and lowest MI due to EMR not reaching values above 1.000 with an average of 0.998. The average DP was 0.944 with the highest value (0.994) observed for LEta015. The average RP was 1.032 with the highest value of 2.000 observed for LEct001, LEta015 and TCIC-419.

Fig. 3.

Comparison of microsatellite polymorphisms detected by high-resolution melting and capillary electrophoresis. On the left, the HRM-shifted melting normalized melting curve and the difference curves are shown. Each curve indicates the melting status for each accession. Different colored curves represent the different cluster of assignment of the accessions studied. On the right, fragment analysis profiles of fluorescence-labeled microsatellites separated by capillary electrophoresis. The number above each peak indicate the alleles sizes. a marker LE20593; b marker LEaac001

Population structure

The population structure analyses based on the SNPs, CE-SSR and HRM-SSR data are shown in Fig. 4. Coefficients of membership are reported in Supplementary Table 1. Based on ADMIXTURE analysis, the collection was partitioned into 3 subclusters that represent the most likely number of subpopulations (K) as displayed by the CV error graph (Fig. 4a). The first cluster (K1) included 29 accessions, mostly landraces with few heirlooms and long-shelf-life (LSL) cultivars. This cluster included individuals from all varietal groups. The second cluster (K2) consisted of 9 accessions with cherry typology and included both landraces and long-shelf-life cultivars from Italy as well as heirlooms from the United States. The third cluster (K = 3) included 33 accessions pertaining to distinct typologies and biological states. Only one accession was considered admixed, as it had values for the highest cluster membership coefficient (qi) lower than 0.5. The PCA biplot graph in the first two components explained 37.36% of the variation, with accessions of K1 distributed in both positive and negative axes of the first and second components, whereas those within K2 and K3 were scattered on the negative axis of the first and second components, respectively.

Fig. 4.

Genetic structure of the 72 tomato accessions. For each marker system is shown the bar plot with individuals represented by a thin vertical line, which is partitioned into K-colored segments whose length is proportional to the estimated membership coefficient (q) (right panel). The loading plot in the first two components, showing the diversity of the 72 studied accessions are displayed (central panel). Accessions in the PCA biplot are annotated considering the K subdivision following the population structure analysis. The cross-validation error for different admixture or structure models (K = 1–10) are shown (left panel). a ADMIXTURE analysis with 21,020 ddRADseq-SNPs markers showing 3 (K = 3) groups according to the most informative K value (lowest value); b STRUCTURE analysis with 96 CE-SSR polymorphic loci showing 2 (K = 2) groups according to the most informative K value (highest value); c STRUCTURE analysis with 98 HRM-SSR polymorphic loci showing 3 (K = 3) groups according to the most informative K value (highest value)

The population structure data from the CE-SSR analysis (Fig. 4b) revealed two likely subpopulations including 34 (K1) and 38 (K2) accessions without any distinction in terms of varietal type, origin or biological status. The PCA revealed less variation than the SNP markers did, explaining 16.04% of the variation in the first two PCs. Accessions were mostly differentiated on the first principal components with individuals of the K2 population positioned on the negative PC1 axis, while those of the K1 on the positive one.

On the basis of HRM-SSR data, three likely subclusters were identified. The first (K1) was represented by 20 accessions and included several Italian landraces and LSL cultivars as well as heirlooms recovered from non-European countries (Canada, China, Mexico, Russia, the United Kingdom, the United States). The second cluster (K2) included 27 accessions and was exclusively represented by landraces and LSL varieties from Italy and France. The last small cluster (K3) grouped 13 accessions from Italy and from the United States, being highly represented by cherry types. A high number of admixed accessions (13) were found via HRM-SSR data. As observed with the CE-SSR data, K1 and K2 differed on the first principal component, being positioned on their respective negative and positive axes. Instead, the accessions belonging to K3 were found to be more distributed between the first two subpopulations.

PCA charts depicting the distribution of accessions according to typology, biological status, and provenance for the three genotyping methods are shown in Fig. 5.

Fig. 5.

Loading plot of the first two components, showing the genetic diversity of the 72 studied tomato accessions according to typology (left panels), biological status (central panels), and provenance (right panels). a ddRADseq-SNP markers; b CE-SSR markers; c HRM-SSR markers. For each genotyping system, the variation explained in the first and second components is reported in Fig. 4

On the basis of varietal types and provenance, no clear separation of the accessions was observed, although the SNP data better discerned cherry types from the rest. In addition, SNPs highlighted specific groups of accessions with Italian and North American origins. Instead, HRM-SSR discriminated European and non-European germplasm with a greater efficacy. This genotyping method also effectively distinguished between landraces and heirlooms, with the LSL varieties tending to cluster closely with the latter.

Phylogenetic relationships

Neighbor-joining phylogenetic trees based on the maximum likelihood model for different genotyping data are reported in Fig. 6. In all instances, the population was divided into several subgroups, with accessions having different levels of similarity. Although the same phylogenetic relationships were not found for all the accessions, we observed specific clusters using the 3 different genotyping methods. As an example, the two cherry heirlooms HL_C3 (‘Snow white’) and HL_C4 (‘Gajo de melon’), which are appreciated on the U.S. market, and with yellow-orange fruits color at maturity, were grouped at a high similarity level with SNP and CE-SSR data. Although the same cluster was not observed when HRM-SSR was used, HL_C3 clustered close to another Italian cherry LSL cultvar (LSL_C5) in agreement with the SNP data. Robust clustering with SNPs and CE-SSR markers were also observed for the heirloom varieties HL_G4 (‘Blanche du Quebe’), HL_G5 (‘Violet Japer’) and HL_P1 (‘Michael Pollan’) as well as for pairs of landraces with globe (LR_G1 ‘Varrone’; LR_G2, ‘Laura’) or plum and cherry (LR_P4, ‘San Marzano 622’; LR_C8, ‘Malareto’) shapes. SNP and HRM-SSRs highlighted a close relationship between pairs of beefsteak tomato types such as LR_B14 (‘Marmande Cap’) and LR_B15 (‘Marmande Csmall’) as well as HL_B1 (‘Black Tula’) and HL_B2 (‘Mexican Ribbed’). Only two plum accessions from the Campania region (Italy): LR_P8 (‘San Marzano LPRO’) and LR_P9 (‘Campano Bal’) presented higher levels of similarity with both the CE and HRM microsatellite data. The three methods highlighted the close kinship of heirlooms having orange-ribbed external fruit color, HL_B3 (‘Apricot Ribbed’) and HL_B4 (‘Tangerine Ribbed’) and Italian landraces with cherry shape LR_C7 (‘Pom Nero Pro’) and LR_C8 (‘Malareto’).

Fig. 6.

Neighbor-joining phylogenetic trees via the maximum likelihood model with 1,000 bootstraps. Individual dendrograms drawn with ddRADseq-SNP (a), CE-SSR (b) and HRM-SSR (c) markers are shown. Accessions that demonstrated consistent grouping with the different molecular data are highlighted by colored squares connected by dotted lines. Accessions with different biological status are represented with different colored symbols: landraces with red circles, long-shelf-life cultivars with blue squares, heirlooms with yellow rhombuses. For accession acronyms see Table 1

Overall, among the three genotyping methodologies, we observed a greater number of shared clusters using SNP data.

Population genetic differentiation

The genetic differences within and among the groups constituting the tomato collection were inferred via calculation of the AMOVA and F_ST values. Analyses were performed for each genotyping method subdividing the population according to their biological status (landrace, long-shelf-life cultivar, and heirloom), cultivar type (beefsteak, cherry, globe, oxheart, and plum), and provenance (European and non-European). The results are reported in Supplementary Tables 2 and 3. In all instances, the percentage of molecular variance was greater within each population than among populations, being greater when considering the biological status. SNP data revealed the same variation ratio (98–2% within-among populations) in all the considered populations. CE-SSR data indeed revealed a high intrapopulation variance in biological status (99%) with a slight decrease within cultivar types (98%) and provenances (97%). A higher degree of variation was found with HRM-SSR data with intrapopulation variability in the range of 11–14% and greater values among cultivar types. The fixation index (F_ST) values revealed, in all instances, greater genetic differentiation between heirlooms and long-shelf-life cultivars. Concerning varietal types, the SNP and SSR-HRM data highlighted a greater difference between cherry and plum tomatoes, while CE-SSR data between cherry and oxheart cultivars. Similar pairwise F_ST values were also recorded between European and non-European germplasm via SNP and HRM-SSR markers, while CE-SSR suggested a higher genetic difference rate.

The degree of relationship between the genotyping matrices revealed significant correlations in all pairwise comparisons: HRM-SSR/CE-SSR (R = 0.245, p < 0.0001), HRM-SSR/SNP (R = 0.233, p < 0.0001), and CE-SSR/SNP (R = 0.270, p < 0.0001) and comparison values for the 3 matrices HRM-SSR/CE-SSR/SNP (R = 0.226, p < 0.0001).

Discussion

The primary challenge facing modern agriculture is the need for greater food security in light of the world’s population growth and overall climate change. In this context, traditional varieties are becoming more widely acknowledged given their importance as a source of useful alleles to improve the resilience and sustainability of crops. Thus, it is essential to understand their genetic diversity to enhance the management of germplasm resources toward effective exploitation in breeding programs. This study aims to dissect the population structure and phylogenetic relationships of a diverse set of tomato varieties, through different categories of molecular markers, with the aim of evaluating the level of informativeness of these genotyping systems for crop genetic diversity research. The collection was part of an association mapping panel (Tripodi et al. 2021) and included diverse locally adapted varieties that were established through genetic resources valorization initiatives. The rationale arises from the need to discern the accuracy of the approaches applied to outline their potential and drawbacks. Microsatellites were the primary marker of choice until the beginning of the ‘omics’ era. The significant progress of cutting-edge sequencing technologies enhanced the discovery of thousands of SNPs providing high genome-density scan. Several factors, such as the flexibility of analysis, the marker informativeness and the cost per sample, need to be considered prior to the adoption of any genotyping methodology (Kockum et al. 2023). The analysis of microsatellites via capillary sequencing is labor intensive for the preparation of labeled amplicons and inspection of the peaks detected. Furthermore, amplification of microsatellites may induce artifacts affecting allele size and related genotyping information. High-resolution melting expedites the process since labeled primers and post-PCR handling procedures are not needed, thus reducing the associated costs of each assay. However, the sensitivity of detection may be affected by variation in DNA concentration among samples and/or systematic mistakes in pipetting, leading to melt curve clustering bias (Slomka et al. 2017). Despite the complexity of library preparation, ddRAD sequencing may be simple to carry out owing to the networks of genotyping platforms offering services. Furthermore, SNP markers are not affected by the chemistry and/or platforms used, allowing major interoperability between research groups. However, the downstream analysis of the raw data requires greater computational skills than microsatellites do.

In this study, the development of SNP markers was outsourced to the IGA technology service (Udine, Italy), while SSR genotyping was performed in the laboratory of CREA via an internal pipeline. For the number of samples considered, the expenses associated with genotyping using ddRADseq and CE-SSR were comparable, whereas HRM analysis allowed for a roughly one-third cost reduction. However, considering the number of genotyped loci, ddRADseq analysis allowed cost reductions of approximately 200% and 60% compared with those of CE-SSR and HRM-SSR, respectively.

In tomato, microsatellites have been widely used to characterize both traditional and improved cultivars. Gonias et al. (2019) used nine microsatellites in a set of 107 Greek landraces, modern varieties, and commercial hybrids. As in our study, the landraces constituted different subgroups and did not clearly separate from the rest, due to the complex selective history of these materials which determines the formation of subgroups that inherit specific traits (Vilarinho et al. 2015; Tripodi et al. 2023). However, the average number of alleles per locus was greater than that in our study, possibly linked to the different markers used as well as more heterogeneous materials. One of the constraints of our choice of markers was the size of the amplicon required for HRM analysis, which should not exceed the range of 100–200 bp to obtain optimal profiles (Grazina et al. 2021). This limited the possibility of identifying highly polymorphic markers based on information in the literature. Pozharskiy et al. (2023) investigated the diversity of transcontinental Asian tomatoes through 13 SSRs, including common markers identified in the present study. In agreement with our findings, LE20592 and LE21085 presented similar polymorphic information content, although the number of detected alleles detected was lower. The average number of SSR alleles found agreed with previous studies testing the same microsatellites (He et al. 2003; Castellana et al. 2020; Caramante et al. 2021; Tripodi et al. 2022), despite their polymorphic information content differed. This primarily depends on the germplasm analyzed and its level of diversity. Overall, the marker dataset showed average information content proving to be adequate for downstream genetic diversity inferences.

Similar results were obtained from the population structure analysis using separate marker datasets, indicating that the collection was divided into two major groups. SNP and HRM-SSR data indicated a smaller third group. Remarkably, this final group included identical cherry accessions, emphasizing their robust genetic structure. The likely number of detected K clusters showed poor structuration of the collection, in agreement with the AMOVA results, highlighting high intrapopulation variation and suggesting high gene flow between individuals. The absence of modern varieties influences the genetic structure of the population as observed by Castellana et al. 2020. Indeed, the collection was mostly represented by ancient cultivars retrieved from Mediterranean areas. For heirlooms, despite having been developed by specific breeding programs, they mostly represent the selected variability of the United States according to market and consumer preferences. Comparison of phylogenetic trees highlighted more robust clusters for heirlooms likely because of the lower heterogeneity compared with that of landraces. A better region wise grouping was found with the CE-SSR than with the HRM-SSR and SNP markers. These results agree with those of Choudhury and collaborators (2023) who reported that microsatellites able to detect greater geographic isolation of rice ecotypes than SNP markers. However, the SNP data allowed better discrimination of cultivar types compared with the other two methods, which are based on phylogenetic relationships and F_ST values. A main driver for varietal classification of tomatoes is the size and shape of the fruits (Bhattarai et al. 2018), both of which are under the control of many genetic loci (Tanksley 2004; Monforte et al. 2013). The advantages of genome-wide markers include in the greater probability of detecting polymorphisms in genomic regions involved in the variation of quantitative traits, that in this case was not possible with microsatellites, given the lower number of loci identified. According to earlier research, the number of SNPs has a significant impact on their ability to resolve population genetic structure. In sunflower, Filippi and collaborators (2015) reported differences in individual subpopulations structure assignments, indicating how distance methods are less effective to identify reproducible groups via SSR and SNP markers. In rice, Singh and collaborators (2013) highlighted a similar broad grouping pattern with both SSR and SNP markers resolving individuals in different subclusters in the phylogenetic dendrograms. Similarly, in eggplant (Gramazio et al. 2017), differences between evolutionary trees separating brinjal, scarlet and gboma accessions through SNP and SSR data were observed. These findings, in agreement with our results, suggest similar hierarchical groupings but different relationships among individuals on the basis of the polymorphisms detected, thus highlighting how SNPs may improve the estimation of genetic relatedness.

We also explored the possibility of HRM as a viable alternative to capillary electrophoresis for discrimination of microsatellites. This approach has been widely used in several crops, including vegetables (An et al. 2017; Jeong et al. 2010), cereals (Shabanimofrad et al. 2015; Yu et al. 2011), legumes (Bosmali et al. 2012; Razaida et al. 2021), fruit trees (Di Stefano et al. 2012; Gomes et al. 2018), and officinals (Li and Xiong 2018; Nunziata et al. 2018) but only in minor extent have been applied in tomato (Tripodi et al. 2022). For several markers, HRM-SSR analysis provided higher number of alleles, than CE-SSR, due to the higher probability of detecting polymorphisms in the bounding regions of microsatellites (Di Stefano et al. 2012). Moreover, the collection was resolved with greater variation via HRM-SSR providing more consistency with SNP data in terms of the number of subpopulations detected and divergences among the categories of individuals, thus confirming that this method is suitable for large-scale genotyping of tomato.

The three datasets were highly significantly correlated, as shown by the Mantel test, suggesting that all the datasets are complimentary for genetic diversity investigation. Implementation of additional studies in other species and including different gene pools would provide a validation of the results described here for the three genotyping methods.

Conclusion

This study tested the ability of different genotyping systems to explore the genetic diversity of germplasm resources. As model species, we considered tomato, investigating a heterogeneous collection of traditional varieties. To better understand the potentiality and constraints of the applied genotyping methods, we analyzed the population structure and inferred phylogenetic relationships. Overall, we showed that either the SNP or the HRM/CE -SSR panels used here are appropriate for estimating genetic diversity. Indeed, the three systems led to the same general conclusion for the investigation of both genetic structure and diversity within and among populations. Estimates of genetic differentiation (F_ST) among populations derived from SNPs were higher than both CE- and HRM-SSRs although significantly correlated. We also emphasized the distinctive insights that each genotyping technique offered. SNP and HRM-SSR markers highlighted a high relatedness at the population structure level with a greater number of populations detected. SSRs demonstrated superior sample grouping, better discerning the diversity of European and non-European germplasms. Furthermore, SNPs improved the understanding of phylogenetic relationships, allowing a better clustering of accessions according to common characteristics. This outcome is undoubtedly predicted considering the significant number of loci impacted by SNPs. Our results also demonstrated that HRM analysis combined with microsatellites provides a valid and more economical alternative system that can be effectively used to uncover the diversity of crops. The implementation of these molecular genotyping systems can provide valuable information toward full understanding of genetic variation and evolutionary relationships in plant species.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

P. T. conceived the project, performed laboratory and data analysis, and drafted the manuscript. G.C. provided the tomato germplasm collection. R.D.A. and P. T. performed capillary microsatellites and high-resolution melting analyses, A.C. reviewed the molecular data. All the authors revised and approved the final manuscript.

Funding

This study was supported by the RGV-ORFLORA project funded by the Ministry of Agriculture, Food Sovereignty and Forests and by the European Union Horizon 2020 Research and Innovation program for funding this research under Grant agreement No. 774244 (Breeding for Resilient, Efficient and Sustainable Organic Vegetable Production; BRESOV).

Data availability

The data that support the findings of this study are available from the corresponding author.

Declarations

Conflict of interest

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.