Development and deployment of a high-density linkage map identified quantitative trait loci for plant height in peanut (Arachis hypogaea L.)

Plant height is one of the most important architecture traits in crop plants. In peanut, the genetic basis of plant height remains ambiguous. In this context, we genotyped a recombinant inbred line (RIL) population with 140 individuals developed from a cross between two peanut varieties varying in plant height, Zhonghua 10 and ICG 12625. Genotyping data was generated for 1,175 SSR and 42 transposon polymorphic markers and a high-density genetic linkage map was constructed with 1,219 mapped loci covering total map length of 2,038.75 cM i.e., accounted for nearly 80% of the peanut genome. Quantitative trait locus (QTL) analysis using genotyping and phenotyping data for three environments identified 8 negative-effect QTLs and 10 positive-effect QTLs for plant height. Among these QTLs, 8 QTLs had a large contribution to plant height that explained ≥10% phenotypic variation. Two major-effect consensus QTLs namely cqPHA4a and cqPHA4b were identified with stable performance across three environments. Further, the allelic recombination of detected QTLs proved the existence of the phenomenon of transgressive segregation for plant height in the RIL population. Therefore, this study not only successfully reported a high-density genetic linkage map of peanut and identified genomic region controlling plant height but also opens opportunities for further gene discovery and molecular breeding for plant height in peanut.

peanut is too dwarf or too high, the combine harvesters could miss some pods or make some peanut straw and pods mixed together. Therefore, plant height has become one of the increasingly important agronomic trait in peanut breeding. Understanding the genetic basis of plant height is conducive to enhance resistance to lodging, increase peanut yield and improve efficiency of mechanized harvesting.
Quantitative trait locus (QTL) analysis is one of the major trait mapping approach to identifying genomic loci that control agronomic traits in crops. Simple sequence repeat (SSR) markers have been the choice of genetic markers and were used for developing genetic maps for cultivated peanuts which in turn facilitated several QTL mapping studies in peanut using F 2 and recombinant inbred lines (RIL) populations. Strikingly, many of these QTLs were detected for drought tolerance [17][18][19] , resistance to biotic stresses such as late leaf spot, rust and bacterial wilt resistance [20][21][22] , pod-and seed-related traits [23][24][25] and quality traits 23,[26][27][28] . However, only few studies focused on the plant height in peanut using F 2 mapping populations 23,25 . Three QTLs for plant height with 4.8~19.2% phenotypic variation explained (PVE) were identified in LG04.2, LG05.2 and LG06.2 using 94 F 2 lines in one year 25 . In our previous study, 3 QTLs for plant height with relatively large genetic distance ranging from 8 to 17 cM were identified based on an F 2 mapping population 23 . The resolution of the above mentioned QTLs for plant height was relatively low due to the limited numbers of markers and the lack of phenotypic evaluation in multiple environments.
In this study, the phenotypic data of plant height across three environments were collected for a RIL (F 6 generation) population from a cross between two peanut varieties, Zhonghua 10 (maternal parent) and ICG 12625 (paternal parent). The present study reports the development of high-density genetic linkage map for cultivated peanut; getting insights on the genetic basis of plant height in peanut and genome-wide identification of QTLs controlling plant height in peanut using the RIL population.

Results
Phenotypic variation of plant height. Phenotypic evaluation of two parental genotypes and RILs for three years (2013~2015) showed significant variation for plant height across the environments. Large phenotypic variation of plant height was observed between the RIL parents as well as in RIL population (  Fig. 1). The broad-sense heritability of plant height was estimated to be relatively high (91.4%) indicating strong control by genetic factors and less environmental effects. Two-way analysis of variance (ANOVA) also revealed that genetic, environmental effects and genotype by environment interaction significantly influenced plant height (Table 2).
Marker polymorphism and genetic map construction. Screening of a set of 7,641 SSR and 427 transposon markers on two parental genotypes of the RIL population resulted in identification of 1,217 (1,175 SSRs and 42 transposon) polymorphic markers. Genotyping of these polymorphic markers amplified 1,272 loci, of which 1,163 markers had one locus for each marker, 53 markers had two loci for each marker and one marker had three loci. Subsequently, a high-density genetic map containing 1,219 loci was constructed covering a total map length of 2,038.75 cM of the peanut genome with an average map density of 1.67 cM per loci (Fig. 2, Table 3). All  A total of 8 QTLs explaining more than 10% of PV were identified in three environments. However, all identified QTLs jointly explained 53.3~65.5% of plant height variation based on multiple linear regressions. Epistasis analysis revealed two nominal interactions of pair-wise QTLs (P < 0.02 and P < 0.05), qPHA4.3 and qPHB4.3b, qPHB6.2b and qPHB6.2c, which contributed little to plant height variation (R 2 = 0.059 and R 2 = 0.061), respectively (Supplementary Dataset 2). Thus, we observed that the epistasis is significantly less important relative to additive effect in present study. Of the18 QTLs identified in three environments, 7 QTLs (38.9%) were detected on LG A04 with 9.45~20.52% PVE, and 5 QTLs (27.8%) were detected on LG B06 with 5.00~7.77% PVE, suggesting that there were QTL clusters on LG A04 and B06. It is worth mentioning that four major QTLs detected on LG A04 in 2013 environment collectively explained as much as 84.98% PVE and had relatively high LOD values with a range of 6.39~7.98. These results indicated that LG A04 is rich in genes controlling plant height. In addition, 7 QTLs detected on LG A04 and one QTL detected on LG B03 in three environments had negative additive genetic effects (Fig. S1), which revealed maternal parent Zhonghua 10 as the source of alleles improving the plant height. The remaining 10 QTLs had positive additive genetic effects, suggesting that the alleles of these QTLs for increased high plant height came from the paternal parent ICG 12625.
To further dissect the QTLs controlling plant height, we integrated summary of QTL information in multiple environments via meta-analysis. Of 18 QTLs of plant height, four QTLs were identified as the reproducible QTL in multiple environments, whose confidence intervals (CI) were overlapped with at least one QTL (Fig. 1). These sets of reproducible QTLs were subsequently integrated into consensus QTL using meta-analysis. Specifically, two reproducible QTLs, qPHA4.1c (CI: 61.5~62.7 cM) and qPHA4.3 (CI: 61.9~62.6 cM), were integrated into a consensus QTL named cqPHA4a (CI: 61.6~62.2 cM). Another two QTLs, qPHA4.1d (70.1~71.5 cM) and qPHA4.2b (67.0~71.4 cM), were integrated into another consensus QTL named cqPHA4b (CI: 69.7~71.0 cM) (Tables 4 and 5). Comparing consensus QTL with their original QTLs improved the resolution by 2-3.4 fold, indicating the QTL meta-analysis was feasible to finely explore the QTL originally detected in multiple environments in the present study.
LGs  Recombination of QTL and transgressive segregation in the RIL population. We observed the extensively phenotypic variations in the RIL population that exhibited a large-scale transgressive segregation for plant height (Fig. 1), almost accounting for half of all lines in the RIL population. There were 6 RILs with shorter plant height than Zhonghua 10 and 43 RILs taller than ICG 12625 in 2013, 16 RILs shorter than Zhonghua 10 and   62 RILs taller than ICG 12625 in 2014 and 18 RILs shorter than Zhonghua 10 and 54 RILs taller than ICG 12625 in 2015. The results suggested that the increasing-effect alleles for plant height may be resided in both the parents, which was congruent to the finding of the additive directions of QTLs detected in the present study (Fig. S1).
To explore the genetic basis of the transgressive segregation of plant height, we evaluated the allelic recombination of QTLs in the RIL population. On the basis of the average plant height across three environments, the whole RIL population were divided into three PH-groups, named 'Less than Zhonghua 10' (n = 11), 'Between Zhonghua 10 and ICG 12625' (n = 75) and 'More than ICG 12625' (n= 54), respectively (Fig. 4A). The relationship between plant height and the compositions of QTL alleles for each PH-group was tested using ANOVA analysis. A nominal difference between three PH-groups was observed for the number of QTL parental alleles harbored in each line (P = 0.015), while the significantly differences were found for the number of positive alleles (P = 2.8E-13) and sum of additive values (P = 8.9E-16) for all QTLs in each lines, respectively (Fig. 4B). Additionally, the two statistics (i.e., the number of positive alleles and sum of additive values) demonstrated an identical trend of relationship with the plant height variations in the whole population, which could explain 46% PV for plant height in the population (Fig. 4C).

Discussion
The construction of a genetic linkage map with optimum density is a prerequisite for conducting QTL analysis in the biparental population. More importantly, genome coverage level and marker density of the genetic map significantly impacts not only the sensitivity of QTL detection but also affects number of identified QTL 30 . The unavailability of optimum genomic resources especially SSR markers together with low level of DNA polymorphism in cultivated gene pool have hindered development of dense genetic maps for a long time in peanut 31 . Among marker types, SSRs gained wide acceptance in the scientific community because of their abundance and their ease of use for DNA fingerprinting, checking adulteration and impurity, genetic diversity, trait mapping and molecular breeding. In case of cultivated peanut, the first SSR-based genetic map was constructed using only 135 SSR markers, covering a total of 1,270.5 cM map distance which covers half of the peanut integrated consensus map 29 . In the past decade, thousands of SSR markers have been developed in peanut from complementary DNAs (cDNAs), SSR-enriched genomic DNA libraries, and BAC-ends 25,[32][33][34][35][36][37] , which greatly accelerated the construction of genetic map and QTL analysis in peanut 21,[23][24][25][27][28][29][38][39][40] . Since the polymorphism of SSRs in a specific population largely depended on the genetic diversity between the parents of an experimental population, several such efforts were required to make available large number of SSRs in the public domain for genetic and QTL mapping studies. In fact 2016 has been a great year because of the availability of genome sequence for both the subgenomes of the tetraploid peanut 15,41 , thereby initiating an era where large number of structural variations have been identified and can be used as genetic markers for generating high throughput genotyping data on genetic populations. Additionally, the previously studies on QTL analysis in peanut were based on either the limited SSRs (no more than 400) 19,21,27,28,39 or the F 2 mapping population [23][24][25] posing difficulty in understanding the genetic basis of the complex traits whose phenotypes needed to be investigated repeatedly in multiple environments. In the present study, 1,175 SSR and 42 transposon polymorphic markers were used to genotype a peanut RIL population. Subsequently, we constructed a high-density linkage map with a total of 1,219 loci with the map length of 2,038.75 cM, covering nearly 80% of peanut genetic map represented by the integrated peanut map 29 . More than 85% of the 1,219 loci segregated in the population at the expected ratio of 1:1, which was significantly more than that of other studies 23,25,29,39,40 . The high proportion of mapped loci following Mendelian segregation has provided good quality and precise genetic map for conducting QTL analysis. Prior to this study, Shirasawa et al. 29 constructed a linkage map with 1,469 loci of which majority were SSRs using mere 91 individuals of a RIL population. It is important to note that above mentioned population was developed from a cross between an elite and an artificial amphidiploid (A. ipaensis × A. duranensis) 4× , and the population size was not big enough (n = 91) to conduct high resolution QTL detection i.e., too for complex traits. Therefore, it is worth mentioning here that the genetic map provided in our study is one of the highest-density maps of cultivated peanut in a permanent experimental population, which would not only give us the opportunity to deeply explore the genetic basis of agriculturally important traits but will also be beneficial to improve the draft sequence of the peanut tetraploid reference genome 15 .
Plant height is highly heritable trait due to accurate trait measurements leading to availability of reliable phenotyping data for conducting genetic analysis and selection 5 . Tremendous efforts have been made to identify QTLs or candidate genes for plant height in rice 4,6-8 , maize 9-11 , wheat 12 25 corresponded to linkage groups A04, B05 and B06 in this study, respectively, by comparing the markers of the two linkage maps. Although 7 QTLs and 5 QTLs were identified on the linkage group A04 and B06, respectively for plant height in this study, there were no repeated markers between these 12 QTLs and above mentioned QTLs in the study of Shirasawa et al. 25 . Among these identified QTLs, one QTL, qPHB4.3b, was detected on the LG B04 near the marker AHGS2429, which was located on the confidence interval of qHMSB4 in our previous study for plant height 23 . However, other QTLs were different from those detected in previous studies 23,25 , which might be novel QTLs for plant height. In our study, qPHA4.1c and qPHA4.3, and qPHA4.1d and qPHA4.2b, respectively overlapped on the linkage group with each other and subsequently were integrated into consensus QTLs cqPHA4a and cqPHA4b, respectively. Our study detected 8 of 18 QTLs explaining more than 10% PVE, providing insights into genetic controls of plant height in peanut. Nearly one third of missing heritability could be due to limited RIL population size.
In the present study, 8 QTLs with negative additive effects and 10 QTLs with positive additive effects for plant height were identified in the RIL population, implying not all increasing-effect alleles of detected QTLs originated from the high-phenotype parent (ICG 12625) and vice versa. We found that all the lines were recombinants at the 18 detected QTLs, indicating the allelic recombination would be the reason where the RIL population contained nearly a half of lines that expressed as the transgressive segregation of plant height to the parents. This conclusion could be supported by the fact that the tall RILs (from 'More than ICG 12625' group) did not harbor majority of the alleles from ICG 12625 but the positive alleles for QTL per se (P = 2.8E-13). The findings revealed that allelic recombination and favorable-allele pyramiding would be an important way to create the more phenotypic diversity and ultimately improve plant height for achieving a balance of vegetable and productive developments in peanut. In RIL population, either the number of additive alleles or sum of additive effect of the detected QTLs for each RIL were significantly correlated to the observed phenotype of plant height (P < 0.05), which suggested that the simple accumulation of QTL alleles allowed to explain the genetic basis of plant height in the peanut RIL population (R 2 = 0.46). It is worth noting that, we detected two consensus QTLs, namely cqPHA4a and cqPHA4b that stably showed significantly additive effects across three environments. Interestingly, the alleles from the high-phenotype parent (ICG 12625) had an ability to decrease plant height of 5.8~7.67 cm relative to the allele from the low-phenotype parent (Zhonghua 10). These large-effect QTLs provided an effective tool for us to improve plant height of the peanut cultivars by introgression few QTLs/genes via molecular marker-assistant selection.

Methods
Plant materials and phenotyping. In this study, a peanut RIL (F 6 generation) population with 140 lines was developed using the single seed decent procedure at Oil Crops Research Institute (OCRI) of Chinese Academy of Agricultural Sciences (CAAS). Wuhan, China. The maternal parent, Zhonghua 10 (A. hypogaea var. vulgaris), is a cultivar with middle plant height developed by the OCRI-CAAS, Wuhan, China in 2004. The paternal parent, ICG 12625 (PI497597, A. hypogaea var. aequatoriana), is a germplasm with high plant height received from the International Crop Research Institute for the Semi-Arid Tropics (ICRISAT), Hyderabad, India. The complete RIL population together with parental lines were planted in experimental field in OCRI-CAAS, Wuhan, China in the consecutive years from 2013 to 2015 using a randomized complete block design with two replications. Each plot contained one row, with 10~12 plants in each row, 10 cm between plants within each row and 30 cm between the rows. Field management followed the standard agricultural practices. At least six plants were selected randomly from each line to investigate the plant height, which measured as the distance from the base of the above-ground plant to the tip of the main stem. Shapiro-Wilk test was used to evaluate the normality of plant height distribution in each year.
With treating the year as a single environment, two-way analysis of variance (ANOVA) was performed to evaluate the effect of genotype and environment on phenotypic variance of plant height in R function "lm" 42 . The line mean based broad-sense heritability for plant height was calculated as: