Peach [Prunus persica (L.) Batsch] belongs to the Rosaceae family, genus Prunus (L.), and is the third most important temperate tree fruit in terms of worldwide production (FAOSTAT, 2021). Peach origin has been described in Asia and its domestication in China. From China, it was dispersed to Europe more than 2,000 years ago (Westwood, 1978; Hesse, 1985; Scorza and Sherman, 1996; Byrne et al., 2012). During the 16th century, peach was introduced to the Americas by the Spanish and Portuguese settlers, where after around 400 years the first formal breeding programme was established (Hesse, 1985; Scorza and Okie, 1990; Faust and Timon, 1995; Byrne et al., 2012).

Over the 1950s (1950–1960), researchers at the “Experimental Station of Aula Dei-CSIC” carried out an initial survey and collection aiming to determine and preserve the variability in the most cultivated stone fruit species that existed in Spain (Herrero et al., 1964). Thus, the first Spanish peach germplasm collection was established at the “Experimental Station of Aula Dei-CSIC” in Zaragoza (Spain), with the aim of organising, describing, and studying peach cultivars traditionally grown in Spain (Herrero et al., 1964). Initially, this collection was mostly composed of non-melting and clingstone cultivars due to their popularity among fruit growers and preferred consumption in Spain. In the successively propagated collections, during the following decades, peach cultivars with similar fruit typology were also introduced from other countries (Cambra, 1979a,b) aiming to study and select the most interesting accessions. In the case of the CITA collection, the exigency and interest of the European market for melting flesh cultivars, induced the establishment of a new peach collection focused on these types of fruit. From that moment on both collections worked coordinately, trying to include the maximum range of material used by Spanish growers. Germplasm banks are an efficient tool for genetic diversity preservation, providing phenotypic diversity to improve important traits in breeding. In fact, both collections studied here showed phenotypic variation for several important diseases and phenological and fruit quality traits (Font i Forcada et al., 2014; Cantín et al., 2021; Giménez et al., 2021).

The molecular analysis, of the different accessions maintained in the collections, allows the establishment of systematic relationships among accessions including their evolutionary relationships and the detection of misidentifications and possible errors made during other GeneBank operations (Bretting and Widrlechner, 1995). The Spanish peach germplasm has been previously explored using microsatellites or simple sequence repeats (SSRs), establishing genetic distances among accessions and marker-trait associations (Aranzana et al., 2003; Bouhadida et al., 2011; Alonso Segura et al., 2013; Font i Forcada et al., 2013). More recently, different resources using next-generation sequencing (NGS) technologies have been generated to study and manage peach germplasm collections (Verde et al., 2012, 2013, 2017), with the most recent being the new version of the peach 18K SNP v2 array (Gasic et al., 2019). Such tools are providing new knowledge for peach breeding regarding genetic control of fruit maturity (Pirona et al., 2013; Nuñez-Lillo et al., 2015; Elsadr et al., 2019), fruit quality and flower traits (Font i Forcada et al., 2019; Lobato et al., 2021), tolerance to diseases (Cirilli et al., 2017; Fu et al., 2021), agronomical traits (da Silva Linge et al., 2021; Fu et al., 2021), and genetic diversity (Thurow et al., 2020; Fu et al., 2021; Mas-Gómez et al., 2021). However, only a small fraction of the Spanish peach germplasm has been studied using NGS methodologies (Font i Forcada et al., 2019; Mas-Gómez et al., 2021).

More recently, the use of NGS methodologies has allowed the exploration of the somatic mutations occurring in bud sports in peach (Foster and Aranzana, 2018). In peach breeding, somatic mutations have been relatively frequent and used as a source of genetic variation (Okie, 1998; López-Girona et al., 2017; Foster and Aranzana, 2018). Such mutations can occur in specific histogenic layers (chimerical), generating individuals formed by genetically different cells (Burge et al., 2002; López-Girona et al., 2017; Foster and Aranzana, 2018). A chimerical natural mutation in the meristematic layer II revealed a new structural mutation causing the reversion from flat to round shape in peach (López-Girona et al., 2017). Moreover, specific regions with possible null alleles and different SNP genotypes in different groups of known sports have been identified using the previous 9K SNP v1 peach array (Micheletti et al., 2015).

Peach shows a high degree of self-compatibility which has caused considerable homozygosis levels in peach populations as the Spanish ones (Aranzana et al., 2003; Mas-Gómez et al., 2021), and probably helped to reduce genetic diversity (Li et al., 2013). In addition, a narrow genetic background has been found in peach breeding programmes due to the use of few founders (Li et al., 2013). Genetic diversity studies in peach have showed the highest level of genetic diversity in oriental peaches germplasm (Li et al., 2013), showing a fast decay of linkage disequilibrium (LD) in this population. However, peach germplasm of breeding programmes has shown similar patterns of LD and reduction of variability independently of the programme location (Xie et al., 2010; Micheletti et al., 2015). The higher number of alleles per locus of SSRs against the bi-allelic nature of SNPs, has frequently influenced genetic diversity and LD studies in peach, observing low values of genetic diversity and slow decays of LD when using SNPs (Micheletti et al., 2015; Thurow et al., 2020; Mas-Gómez et al., 2021). Population structure studies in peach have frequently shown differentiated groups by geographic origin (Font i Forcada et al., 2013; Li et al., 2013; Micheletti et al., 2015) and between melting and non-melting flesh varieties (Aranzana et al., 2010; Li et al., 2013; Chavez et al., 2014; Micheletti et al., 2015; Thurow et al., 2020). In Spanish peach germplasm, two diversification regions were identified by Bouhadida et al. (2011) suggesting the different regional environments and the peach industry requirements as main factors for the differentiation.

Here, we genetically characterised, on a genome-wide scale, the two National Peach Germplasm Collections placed at the Agrifood Research and Technology Centre of Aragon (CITA) and at the Experimental Station of Aula Dei (EEAD-CSIC), both located in Zaragoza (North Eastern Spain), and including most of the peach material traditionally used in Spain. We used the new peach 18K SNP v2 array to provide new findings of sports, synonymies/homonymies, SNP/trait associations of important fruit traits, and a deeper description of the population structure, genetic diversity, and LD decay.

The wide genetic characterisation performed using the new peach 18K SNP v2 array has shown multiple synonymies and homonymies in the Spanish peach collections not detected previously with low-density markers (Bouhadida et al., 2011; Alonso Segura et al., 2013; Font i Forcada et al., 2013). The majority of detected synonymies and homonymies are in agreement with available genotypes in GDR. Nevertheless, misidentifications and labelling errors were detected. This kind of errors occurs frequently in germplasm collections because of mix up of accessions, errors in the entry of plants, or unavailability of accurate genomic tools to distinguish accessions (Reed, 2004; Vanderzande et al., 2019).

The different parameters, such as IBS and IBD, and specific thresholds (e.g., 0.97, 0.977, or 0.98), have been established previously to classify two individuals as identical using NGS (Micheletti et al., 2015; Vanderzande et al., 2019; Montanari et al., 2020). Here, the used IBS threshold of 0.99, based on the group of known sports of ‘Springcrest’ (Micheletti et al., 2015), has clearly helped to detect a large number of synonymies and homonymies, therefore, avoiding genetic errors and redundancies in the downstream analysis. In addition, an error threshold of 0.5% was used in pedigree verifications showing a considerable number of pedigree errors. The presence of validated and historical pedigrees (Okie, 1998) agreed with the conclusions made using GDR genotype data, per example showing ‘Fortuna’ as the parent of ‘Andross’ or showing mistakes in ‘Merriam’ as a self-pollination of ‘Everts_5058’.

According to the pairwise IBS analysis performed in PLINK, ‘Starcrest’ showed considerable genotype differences with the rest of the components of the ‘Springcrest’ sport group (clone group 35). These differences were detected in a region of 13.5 Mb located in the upper part of chromosome 4, where ‘Starcrest’ showed 139 SNPs with a different genotype and 362 SNPs with null alleles (being located between the heterozygous and homozygous clusters in GenomeStudio™) (Supplementary Figure 12). In addition, ‘Starcrest’ showed a reduction in heterozygous SNPs in such a region, in which only 93 out of 1,108 high quality SNPs were heterozygous. Previously, eight SNPs with different genotypes and a large amount of possible null alleles were detected in the same region by Micheletti et al. (2015) between ‘Starcrest’ and ‘Springtime’, another sport of ‘Springcrest’ not present in these collections. These findings may indicate the existence of a chimeric mutation in the upper part of ‘Starcrest’ chromosome 4. The chromosome replacement of the detected region by the homologous would explain the high homozygosity found, as it has been previously suggested for other peach sports (López-Girona et al., 2017; Aranzana, 2021). Moreover, the SNP placement between heterozygous and homozygous clusters observed in the SNP plots, may indicate the existence of some cell layer(s) with heterozygous genotype(s) (Supplementary Figure 12). Previous studies have found SNPs associated with the harvest date in the upper part of chromosome 4 (Font i Forcada et al., 2019; da Silva Linge et al., 2021). This fact could be related with the observed harvest date of ‘Starcrest’, being earliest than the rest of their sports (Supplementary Tables 9, 10).

Despite of the commented precision of the new SNP array to distinguish between ‘Starcrest’ and ‘Springcrest’, two inconsistencies were generated by the SNPs, since accessions with differences in flesh colour (yellow and white) were considered as clones [clone group 22 (‘Sudanell’ group) and 31 (‘Rojo de Tudela’–‘Tambarría’)]. In this sense, the yellow flesh is caused by the accumulation of carotenoids in the chromoplasts due to the disruption of the allele ccd4 in LG1, which prevents the degradation of carotenoids (Adami et al., 2013; Falchi et al., 2013). Three mechanisms of disruption of this allele have been described (Falchi et al., 2013) and the physical position of this gene was found from 25,639,600 to 26,317,783 bps (Adami et al., 2013). The 32 SNPs of the region comprised between 25,545,000 and 26,317,783 bps were checked manually being identical among the mentioned clone groups. Moreover, all SNPs found, in the previous study reported by Micheletti et al. (2015), placed in the array showed the same genotypes among clones here. A possible hypothesis of these two discrepancies could be the existence of a different (fourth) mechanism involved in flesh colour, which is not represented in the SNP array used here.

A general low H_o mean was observed in this study which can be explained by the smaller size of populations in comparison with other studies (Aranzana et al., 2010; Chavez et al., 2014; Micheletti et al., 2015; Thurow et al., 2020), except in the case of Murcia population where H_o was higher than H_e, showing negative F_IS values (F_IS = −0.108). High rates of selfing in the northern Spanish populations could be the main reason of these results, showing highest F_IS values. In addition, this fact could also explain the slow decay of the LD observed here. All these results are clear descriptors of traditional Spanish accessions, which were selected from seed-propagated populations and selfing was the main mating system (Badenes et al., 1998; Aranzana et al., 2003).

According to previous studies (Herrero et al., 1964; Bouhadida et al., 2011; Cornille et al., 2015), the low and moderate differentiation values observed here, can be explained by a close origin of the used accessions and intense spread of cultivars along Spanish regions with similar growing conditions by fruit growers. At the same time, moderate values of differentiation were observed among Navarra population and some other relatively close northern populations, suggesting a different genetic background. The moderate differentiation of Murcia population with the rest of the populations, probably could be explained by the large geographical distance and their different soil and climatic conditions, which was also commented by Bouhadida et al. (2011).

As previously commented, the degree of population structure influences LD patterns within the genome (Thurow et al., 2020). In our study, clear differences were observed in the decay of LD between individual clusters, being slow in general. This was mainly due to the small sample size of each cluster. In this sense, a more rapid decay of LD was observed considering the whole population. High overall levels in the whole set have been observed in other species as wheat (Chao et al., 2010) with clear variation of LD decay among clusters. Based on this fact, a population level analysis was not considered in our study. In general, population structure analysis, by fastStructure and DAPC, divided the germplasm into well-defined clusters according to their genetic structure and fruit characteristics (flesh texture and typology).

Similar results have been previously observed (Aranzana et al., 2003, 2010; Chavez et al., 2014; Thurow et al., 2020). Interestingly, North American cultivars were mainly grouped in two clusters, which can be explained by their close relationships (Supplementary Figure 8; Okie, 1998; Li et al., 2013). A detailed understanding of the germplasm structure and clusters helps in parental selection in breeding programmes, increasing genetic diversity, and improving the potential gain from the selection (Pandey et al., 2021). In this sense, as preliminary results (after 1 year of phenotyping) some tolerant cultivars to M. fructicola were observed such as ‘Fraga B.D.’ and ‘Gallur’ (Cantín et al., 2021), which are related to FAST3, while ‘Montaced’, also tolerant (Cantín et al., 2021), had a higher affinity for FAST6. Regarding chilling injury susceptibility, the most tolerant cultivars, also according to preliminary results (Giménez et al., 2021) are distributed in FAST1 (‘Josepet’, ‘Plácido’, ‘Zaragozano Rojo’, ‘Risol’, and ‘Bonet IV’), FAST2 (‘Rubi Rich’), FAST3 (‘Zaragozano Amarillo’, ‘Oom Sarell’, ‘Baladín’, and ‘Alcañiz’) and FAST5 (‘Andross’, ‘Carson’, ‘Loadel’, and ‘Adriática’). These results provide new choices to breeders in parental selection and crossing designs.

The association analysis showed several strong SNP-trait associations with the studied traits. The leaf-gland phenotype trait acquired relevance because of glands contribution in biologic control (Mathews et al., 2009). In addition, a clear association of absence of leaf-glands with high susceptibility to peach powdery mildew (PPM) has been suggested (Dirlewanger et al., 1996). This morphological trait has been used by breeders as a tool to select descendants with glands in peach breeding programmes (Lambert et al., 2020). This trait was described as a Mendelian trait being the glands absence (ee), globose (Ee), and reniform (EE) (Connors, 1922) and placed on chromosome 7 (Dettori et al., 2001; Micheletti et al., 2015). More recently, using NGS technologies, Lambert et al. (2020) demonstrated that a MITE-like Moshan transposable element inserted in the candidate gene (Prupe.7G121100; 14,436,305–14,437,630 bp) controlling this trait, was responsible for the absence or globose-shape phenotype. SNP_IGA_776653 was the closest SNP to the candidate gene detected by Lambert et al. (2020), placed 0.3 Mb of the gene.

More interestingly, SnpEff analysis showed modifier effects in gene Prupe_7G125700, such gen encoded an ankyrine repeat domain (ANK) protein. This protein family has been associated to plant immunity response against different pathogens (Cao et al., 1997; Yan et al., 2002; Yang et al., 2012; Vo et al., 2015). In fact, a recent study identified a MELO3C002434 candidate gene for resistance to powdery mildew in melon, which encodes an ANK protein (Cao et al., 2021). Moreover, several KASP markers were designed around this gene obtaining successful results for marker-assisted breeding (MAS) in melon (Cao et al., 2021). Therefore, Prupe_7G125700 could play an important role in PPM resistance being a candidate gene for further studies in peach.

In the case of FH, seven new SNPs associated with this trait in this work were detected close to the retrotransposon insertion (from 15,897,836 to 15,899,002 bp) on chromosome 5 associated previously with the pubescence (Vendramin et al., 2014). These results agree with previous studies, showing that the region identified here could be an important conserved haplotype in glabrous individuals (Micheletti et al., 2015; Cao et al., 2016; Thurow et al., 2020; Tan et al., 2021). In fact, two significant SNPs (SNP_IGA_602331 and SNP_IGA_602512) located in this region and detected here, using mainly Spanish germplasm, were also detected with strong associations to FH by Micheletti et al. (2015). More recently, Tan et al. (2021) detected the highest associated SNP with FH in the same region (chr5:16,633,286; G/A), located ∼700 kb downstream of the major gene controlling this trait (MYB gene Prupe.5G196100). Although, these authors commented that no proper candidate genes are located in this region (Tan et al., 2021). The use of SnpEff in our analysis allowed us to detect the effect of the significant SNPs in the region in several candidate genes (Prupe_5G208400, Prupe_5G208500, Prupe_5G210500, and Prupe_5G208100). These genes belonging to important gene families involved in trichome formation and development in other species (Alvarez-Buylla et al., 2000; Li et al., 2011). Trichomes are hair-like appendages caused by the differentiation of epidermal cells (Vendramin et al., 2014). Modifier effects were detected in the genes Prupe_5G208400 and Prupe_5G208500, orthologous with MADS-BOX transcription factor genes. MADS-box proteins have been associated to trichomes development in Arabidopsis (Alvarez-Buylla et al., 2000; Willmann and Poethig, 2011), cotton (Lightfoot et al., 2008; Li et al., 2011), and petunia (Ferrario et al., 2004). In fact, a homology was detected between Prupe_5G208500 and AGL8 (AGAMOUS-LIKE 8), a MADS-box gene protein which showed high similarity with GhMADS11, a gene expressed specifically in cotton fibres (trichomes) and associated to fibre cell elongation (Li et al., 2011). Two modifier effects were detected in Prupe_5G210500, a gen orthologous with a bHLH transcription factor domain. Previous studies have identified bHLH TFs involved in trichome formation in Arabidopsis (Morohashi et al., 2007; Zhao et al., 2008) and tea plant (Liu et al., 2021). Specifically, the induction of two bHLH TFs, GL3, and EGL3, has been associated with triggering the trichome initiation pathways and the regulation of trichome development in a framework involving also MYB TFs (Morohashi et al., 2007; Zhao et al., 2008; Hao et al., 2021; Shangguan et al., 2021). In addition, modifier effects were detected with a gen orthologous with COBRA-LIKE 4 protein gene (Prupe_5G208100). COBRA-LIKE family genes were associated with important roles in specific types of cell expansion and cell wall biosynthesis, finding the COBRA-LIKE 9 highly expressed in trichomes (Jones et al., 2006; Brady et al., 2007). The detection of modifier effects in the mentioned candidate genes caused by significant SNPs detected here and two additional independent mutation events (in total three) in the major gene controlling this trait (Vendramin et al., 2014; Tan et al., 2021), may suggest a higher complexity in the expression of FH beyond from the proposed until so far.

Finally, we identified SNPs with significant association to flesh texture (FT) at the upper part of chromosome 4, in a region at the end part of chromosome 6 and in chromosome 7. FT has been related with endopolygalacturonase activity, and several previous studies have indicated the end region of chromosome 4 as the main region on the genetic control of peach texture (Peace et al., 2005; Meneses et al., 2016; Serra et al., 2017; Carrasco-Valenzuela et al., 2019; Giné-Bordonaba et al., 2020). Strong associated SNPs were identified in such region (Martínez-García et al., 2013; Micheletti et al., 2015; Thurow et al., 2020). In addition, SNPs strongly associated to this trait were also identified on chromosome 5 by several authors (Serra et al., 2017; Ciacciulli et al., 2018; da Silva Linge et al., 2021) and chromosome 8 (Ciacciulli et al., 2018). Previous significant SNPs detected by earlier studies using the 9K SNP array v1 (Micheletti et al., 2015; da Silva Linge et al., 2021) were included in our study, except one (SNP_IGA_821894 on chromosome 8). However, no significant association of these SNPs with FT was obtained here. A possible reason for this result may be the low number of melting accessions studied here (31 accessions), compared to the number of non-melting genotypes (120 accessions). This clearly affects the detection of the QTL for FT at the end part of chromosome 4 in our study, where the genetic control of FT has been previously identified, using a BC1 population (Dettori et al., 2001). Although the detected QTL in Chr4 could be a low probability event because of linkage mapping which uses few meiotic events.

On the other hand, a SNP detected here with high association with FT was SNP_IGA_477159 (p-value = 4.74E−04), located at 19,820,974 bp on scaffold 4. The SNP location is closed to the position of the two SNPs with strong association to flesh texture identified by Thurow et al. (2020), and located at 19,904,250 and 19,904,264 bp positions. These results also reflected the complex genetic architecture underlying this trait, which was also confirmed here due to the detection of significant SNPs in different regions or chromosomes (Peace et al., 2005; Martínez-García et al., 2013). Additional studies, including the development of more accurate ways of phenotyping this complex trait, must be developed to complete the genetic dissection of this central quality trait in peach.

The large number of synonymies and labelling errors detected, revealed the importance of genotyping germplasm collections with high-density markers to avoid redundancies and achieve an efficient management. Also, genomic rearrangements detected in ‘Starcrest’ (a sport of ‘Springcrest’) modifying regions related to harvest date indicate new lines of research. Genetic diversity results have shed new light in the predominant mating system in traditional Spanish peach varieties, the two main diversification regions (North and South of Spain) and the spread of peach cultivars. GWAS carried out detected new SNPs associated to TLG, FH, and FT using Spanish germplasm not previously analysed. The SnpEff analysis allowed the detection of PPM resistance and trichome development candidate genes to be further explored in breeding programmes. Finally, the genetic distance between accessions observed here, together with the phenotypic variation observed after evaluations of these materials, could ensure the adequate and representative diversity for future use in the improvement of important traits in peach breeding.

The original contributions presented in the study are included in the article/Supplementary Material, further inquiries can be directed to the corresponding author.

PJM-G, CC, and MM: conceptualization. PJM-G: methodology. JM-G, CC, MM, and PJM-G: investigation and writing – review and editing. PJM-G, CC, and JM-G: data curation and writing – original draft preparation. All authors contributed to the article and approved the submitted version.

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.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.