Chromosome level high-density integrated genetic maps improve the Pyrus bretschneideri ‘DangshanSuli’ v1.0 genome

Pear (Pyrus spp.) is an important temperate fruit crop that is popular worldwide because of its sweet and juicy flesh, excellent eating quality, and nutritional, medicinal and economic value [1, 2]. In the ancient Chinese pharmacopoeia Compendium of Materia Medica (1596), pears were thought to resolve phlegm and relieve cough and asthma. In 2016, the area used for pear cultivation in China exceeded 1.113 million ha and yielded a harvest of 18.70 million tons of fruit (National Bureau of Statistics of the PRC, <http://data.stats.gov.cn/search.htm?s=%E6%A2%A8>, September 26, 2018, data last accessed), which accounted for more than 70% of global pear production (FAOSTAT, <http://www.fao.org/faostat/en/#data/QC>, September 26, 2018, data last accessed).

At present, conventional hybrid breeding is still the main method for generating new pear cultivars. However, the efficiency of this method is low in pear, because of its large tree size, long juvenile phase, and high genome heterozygosity [3–5]. The tens of thousands of hybrid trees required for breeding occupy a large area of land for many years and consume significant manpower and financial resources. Molecular marker-assisted selection (MAS) based on Quantitative Trait Locus (QTL) mapping of important agronomic traits allows breeders to screen hybrids at the seedling stage and eliminate those lacking potential [4, 6]. Thus, this practice can reduce the land, labor and financial investment required for crop breeding, especially for perennial fruit trees [4]. Developing molecular markers linked to important agronomic traits is an important part of molecular breeding.

Significant progress has been made in developing and applying genetic markers in pears over the last two decades. Iketani et al. constructed two genetic maps from the segregation data of 82 F1 individuals and mapped the resistance allele of pear scab and the susceptibility allele of black spot in different linkage groups of ‘Kinchaku’ in 2001 [7]. Since then, researchers have used RAPD, SRAP, AFLP, SSR and other molecular markers to construct pear genetic maps and identified QTLs for agronomic traits [8–12]. In recent years, a hybrid population of ‘Bayuehong (BYH)’ and ‘Dangshansuli (DS)’, which was established by the Zhengzhou Fruit Research Institute of the Chinese Academy of Agricultural Sciences, has been used by several research teams to construct genetic maps and map QTLs for dozens of fruit traits, including fruit weight, fruit shape index, soluble solid content, and fruit maturity [3, 5, 6, 13–16].

Genetic maps constructed using RAPD, AFLP and SSR markers are not ideal for map-based cloning or identification of genes responsible for important traits because of the low density of the mapped markers. Following the application of next generation sequencing (NGS) technology, high density pear genetic maps constructed with SNP markers were used for QTL mapping and map-based gene cloning [6, 17, 18]. For example, Yao et al. [18] identified the MYB114 gene responsible for the red skin trait using a high-density BYH × DS (BD) map [3]. Furthermore, high-density genetic maps can guide the anchoring of scaffolds to pseudo-chromosomes. Wu et al. anchored 796 scaffolds of 386.7 Mb (or 75.5%) of DS v1.0 genome sequences to 17 pseudo-chromosomes according to a genetic map consisting of 2005 SNP markers [1]. Chagné et al. anchored 868 scaffolds of 171.3 Mb (29.7% of the genome) to the 17 pseudo-chromosomes of Pyrus communis ‘Bartlett’ according to an integrated map consisting of 2279 markers (1391 and 888 apple and pear SNPs, respectively, genetically mapped using seven apple full-sib families and five pear segregating populations) [2, 19, 20]. Recently, Li et al. [6] anchored another 291.5 Mb of the ‘Bartlett’ genome scaffolds in 17 pseudo-chromosomes using three high-density SNP genetic maps (OH × LBJ, BYH × DS-JXB and PH-CG) with European pear genetic backgrounds, which dramatically increased the anchored portion of the ‘Bartlett’ genome from 29.7% in the original assembly to 50.5%. The assembled genomes of DS and Bartlett provide useful tools for fine mapping of genes/QTLs and map-based gene cloning, and they have laid a solid foundation for future pear genetics and genomics research. However, pseudo-chromosome genome assemblies do not yet exist for ‘Bartlett’ or DS.

Scaffold anchoring of the two published pear genomes was performed using genetic maps constructed with F1 populations that were relatively small in size, which may have resulted in errors and conflicts, some of which were discovered in subsequent analysis [6, 21]. High-density consensus genetic maps constructed by integrating high-quality genetic maps have been used to improve the quality of pear genome assemblies [6]. In this study, we used two relatively large F1 populations of Asian pear to construct genetic maps. By integrating these two new maps with the BD map [3], we were able to improve the scaffolding of the DS reference genome v1.1 to the chromosome level <http://genedenovoweb.ticp.net:81/pear/>.

With the development of high-throughput sequencing technology and the popularity of simplified genomic sequencing technologies such as RAD [24], GBS [25] and SLAF [26], genetic maps with high (or even ultra-high) marker density have laid a foundation for downstream applications such as QTL mapping and chromosome assembly [27–30]. Wu et al. constructed a genetic map of 2005 SNP markers in pear using a BYH × DS cross and anchored scaffolds of the DS genome [1]. Subsequently, they improved the genetic map and identified QTLs by increasing the number of offspring from 56 to 102 and increasing the number of markers to 3241 (the BD map) [3]. Montanari et al. successfully mapped 829 polymorphic pear markers and 569 polymorphic apple markers from their newly developed apple and pear Infinium® II 9 K SNP array to the parental genetic maps for five pear segregating populations, and they found that the number of markers per map varied from 318 to 645 [20]. These parental genetic maps were then used, along with maps for seven apple full-sib families [19], to anchor 171.3 Mb of the assembled ‘Bartlett’ genome into 17 LGs [2]. Recently, Wang et al. built a high-density linkage map with 4865 markers (4664 SNP markers and 201 SSR markers) of the segregating population of ‘Red Clapp’s Favorite’ × ‘Mansoo’ (RM map), spanning 2703.61 cM, with an average distance of 0.56 cM between adjacent markers [31]. To date, the BD and RM maps are the densest pear genetic maps. In this study, we selected only five of the polymorphic parental SNP markers equally distributed on each scaffold to reduce marker redundancy for construction of the MH map and the YM map. This practice did not maximize the number of markers or marker density of these genetic maps, but it led to the generation of maps on which the markers were relatively evenly distributed.

During comparative mapping of multiple genetic maps, some markers may be revealed to be located at different positions of a LG or may even be assigned to different LGs. This phenomenon is observed even between male and female maps from a single outcross population [32], and it may be caused by segregation distortion of markers. Segregation distortion of markers affects the recombination distance between markers and the order of the markers on LGs [33, 34]. Wang et al. found that two distorted markers (NAUpy59t and ZFRIt051) were able to change the order of other markers on a LG [32]. Differences between the genetic backgrounds of parents in different mapping populations can also lead to differences in marker order [31, 32]. In this study, the parent material for the MH map consisted of ‘Mantianhong’ (P. pyrifolia) and ‘Hongxiangsu’ (P. bretschneideri), whereas the parent material for the YM map consisted of ‘Yuluxiang’ (P. bretschneideri) and ‘Mantianhong’, and the genetic backgrounds of the two populations were relatively similar. However, the parents of the BD map were ‘Bayuehong’ (P. communis) and DS (P. bretschneideri), so their backgrounds were quite different from those of the MH and YM populations [35–38]. Therefore, the markers of the MH and YM maps have relatively good collinearity between each other compared to that of the markers of the BD map. In addition, the presence of incorrect insertions of contigs in some scaffolds [21] and large differences in population size may also be responsible for differences in the order of the markers among the different maps analyzed in this study.

The quality of the genome assembly has an important and direct effect on the anchoring of scaffolds into pseudo-chromosomes [6, 39, 40]. Pootakham et al. constructed an integrated linkage map and anchored 28,965 contigs, covering only 12% of the published rubber tree genome, of which 78% of sequences were identified as repetitive DNA, and the average scaffold length was 1.84 kb [41, 42]. Westbrook et al. mapped 3305 loblolly pine scaffolds onto 12 linkage groups using 3762 markers and found that most scaffolds were too short to span two or more markers [43]. In pear, there were 2103 scaffolds (with an N50 scaffold length of 540.8 kb) and 142,083 scaffolds (with an N50 scaffold length of 88.114 kb) in the DS and ‘Bartlett’ genome assemblies, respectively, however, in comparison with the ‘Bartlett’ genome assembly, more sequences of the DS assembly had been anchored into pseudo-chromosomes with fewer markers [1, 2, 6]. Genetic maps are useful tools for guiding scaffold anchoring into pseudo-chromosome assembly, and the quality and density of genetic maps are important factors [6, 39]. In this study, the polymorphic markers of the parents of two populations were first filtered according to their positions before the genetic map was constructed, so the number of mapped markers in the two genetic maps was not the largest among published genetic maps, but the number of anchored scaffolds of these two maps were both close to that of the previously published BD map, and the total length of the anchored scaffolds in each of these two maps exceeded the total length of the scaffold anchored by the BD map (excluding the SSR markers in the BD map). Moreover, in comparison with the BD map, the direction of more than 95% of anchored scaffolds (anchored with two or more markers) can be confirmed by both newly constructed maps, while the direction of only 68.78% of anchored scaffolds can be determined according to the BD map. This finding indicates that the MH and YM maps had fewer redundant markers and higher quality in comparison with the BD map, and thus they are more useful as guides for pseudo-chromosome assembly.

The new DS v1.1 genome is a greatly improved version in comparison with DS v1.0, with particular improvement to pseudo-chromosomes 1 and 7. In this study, seven scaffolds previously anchored to pseudo-chromosome 1 of DS v1.0 were transferred to pseudo-chromosome 7 of DS v1.1. Of the 38 scaffolds anchored to pseudo-chromosome 1 of DS v1.1, 19 were newly anchored in this study. Only 12 of the other 19 scaffolds were also anchored to pseudo-chromosome 1 according to the BD map, while the other seven were scattered among the other pseudo-chromosomes of DS v1.0. Li et al. reported similar findings in a recent study, in which they integrated existing genetic maps and performed scaffold anchoring. They anchored nine new scaffolds on DS pseudo-chromosome 1 and re-anchored a large number of markers located on LG1 of the BD map (BYH × DS-JXB) to pseudo-chromosome 7 of the DS genome [6]. Five of the nine scaffolds (scaffold 341.0, scaffold 467.0, scaffold 638.0, scaffold 797.0 and scaffold 872.0) newly anchored by Li et al. also appeared in the list of 19 newly anchored scaffolds in this study, which demonstrated the reliability of the DS v1.1 assembly. The partially homologous blocks on pseudo-chromosomes 1 and 7 may be the reason that a large number of markers that should have been mapped onto LG7 in the BD map were instead mapped onto LG1 [1, 6, 23]. However, the failure to anchor more scaffolds onto pseudo-chromosome 1 may also be related to the small size of the population that was used to construct the BD map, as well as lethal genes present in chromosome 1, which can cause interspecific hybrid necrosis [44].

In our previous study, a modified QTL-seq analysis was performed on the red/green fruit skin trait locus of Asian pear at the scaffold level, and the scaffolds linked to the red/green locus were then mapped to LG5 according to the high density BD map [3, 21]. In this study, we directly performed chromosome-level QTL-seq analysis using the new DS v1.1 genome assembly and located the red/green locus in an interval consistent with that reported in previous studies [18, 21], which greatly reduced the workload required to map scaffolds to linkage groups and supported the accuracy of DS v1.1. The peak of the candidate interval in the |∆(SNP-index)| plot (Additional file 2: Figure S1) was split into two parts by a 3.26 Mb gap because most part of each of the three scaffolds (NW_008988076.1, NW_008988141.1 and NW_008988091.1) within the gap were unlinked to the R/G locus [21]. As the three individual genetic maps are highly consistent in the entire region of LG5 (Fig. 2), we were able to rule out the possibility of assembly error regarding the genetic maps. In short, although there were small errors in some scaffolds, the DS v1.1 genome assembly provides a foundation for studies aimed at locating important agronomic traits at the chromosome level.

In this study, two sets of maps, MH and YM, were constructed and further merged with a previously published BD map to build a new integrative MH-YM-BD map. Eight hundred ninety-eight scaffolds, of which 88.31% were directionally anchored with two or more markers, were then successfully anchored into 17 pseudo-chromosomes of the DS v1.1 genome according to the MH-YM-BD map, accounting for 78.8% (400.57 Mb) of the assembled genome size. Errors in each pseudo-chromosome were corrected. Pseudo-chromosomes 1, 5 and 7 were extended by 50, 20 and 27.5%, respectively. Seven scaffolds from pseudo-chromosome 1 were transferred into pseudo-chromosome 7 in the newly constructed DS v1.1 genome. The DS v1.1 genome, which has high collinearity with the apple genome and was used to accurately locate the red/green locus of Asian pear, provides a critical tool that is essential for studies of pear genetics, genomics and molecular breeding.