Quantitative trait locus analysis of cucumber fruit texture using double-digest restriction-site-associated DNA sequencing

Fruit �rmness and crispness are important traits of cucumber because they directly affect its commercial value. We performed quantitative trait locus (QTL) analysis of these fruit texture traits with the double-digest restriction-site-associated DNA sequencing (ddRAD-Seq) technique and detected 31 QTLs for fruit �rmness: 11 for placenta �rmness, 12 for skin �rmness, and 8 for �esh �rmness; and 25 QTLs for crispness-related scores: 10 for �esh crispness index, 8 for �esh apparent fractal dimension by Richardson plot, and 7 for �esh apparent fractal dimension by box-counting (Kolmogorov’s dimension). Several QTLs associated with �esh �rmness and crispness are located near regions for fruit length, diameter, and length-to-diameter ratio, and for resistance to powdery mildew and downy mildew, indicating that gene linkage is likely to limit breeding e�ciency. Our results will contribute to the development of informative DNA markers closely linked to genes for desirable fruit texture traits that are required for effective selection of new cultivars.


Introduction
Fruit texture traits, such as rmness and crispness, are one of the most important characteristics in cucumber (Cucumis sativus L.), because these are directly related to consumer preference and demand.Grown and bred around the world, the wide range of cucumber cultivars differ greatly in rmness and crispness.In breeding programs, the genetic structure of the breeding materials and the inheritance of target traits are important for e cient breeding.The genetic and physiological mechanisms of fruit texture traits have been investigated by many researchers (Jones 1954 Shimomura et al. 2016).Although breeding efforts for many fruit vegetables have focused on improvement in texture, few advances have been made in identi cation of genetic mechanisms, because aspects of texture other than rmness are di cult to measure quantitatively.Cucumber fruit rmness has been quanti ed by scienti c instruments for many years (Breene et al. 1972;Jeon et al. 1973; Kanno and Kamimura 1978;Thompson et al. 1982).However, rmness is just one component of texture, and other components such as crispness and juiciness are also important.
Improvements in computer and instrument performance have allowed researchers to develop new methods for the quantitative evaluation of fruit texture traits by mechanical (Horie et al. 2004;Yoshioka et al. 2009Yoshioka et al. , 2010)), acoustic (Sakurai et al. 2005), and biorheological (Dan et al. 2003; Kohyama et al. 2009Kohyama et al. , 2013) ) measurements.
Considerable advances in the genetic analysis of cucumber fruit traits have been made possible by the development of molecular techniques, such as genetic linkage mapping using molecular markers and quantitative trait locus (QTL) analysis.Many genetic analyses associated with fruit length, diameter, and length-to-diameter ratio (L/D) have been performed.The rst report of QTL analysis for cucumber morphological traits identi ed 18 QTLs: 5 for length, 3 for diameter, 3 for seed-cavity size, 2 for color, 4 for L/D, and 1 for seed-cavity-to-diameter ratio (Kennard and Havey 1995).Later researchers reported many QTLs for fruit morphological traits (Serquen et al. 1997;Fazio et al. 2003 Wang et al. 2020a, b).Methods of single nucleotide polymorphism (SNP) genotyping by next-generation sequencing (NGS), such as restriction-site-associated DNA sequencing (RAD-Seq) and genotyping-by-sequencing techniques, have been widely adopted on account of their low cost and exibility (Davey et al. 2011).In cucumber, NGS investigations include genetic analysis of fruit-related traits.Wei et al. (2014) reported the rst QTL analysis using an NGS-derived genetic map in cucumber; using the speci c-length ampli ed fragment sequencing technique they identi ed 7 QTLs for fruit length and 2 QTLs for fruit weight.However, no QTL analysis has yet been performed for fruit texture traits in cucumber.
Here, we report the rst QTL analysis of fruit rmness and crispness, using double-digest RAD-Seq (ddRAD-Seq) analysis.We investigated three populations of F 2 progeny of parents with high and low crispness: two of Japanese-type cucumber × weedy cucumber line CS-PMR1, and one of Japanese-type cucumber × Beit Alpha-type cucumber.CS-PMR1, derived from PI197088, has strong resistance to both powdery mildew and downy mildew, and can be used to breed new cultivars with resistance to these diseases (Fukino et al. 2013;Morishita et al. 2003;Sakata et al. 2006).However, the development of DNA markers closely linked to genes for both disease resistance and fruit texture traits is needed for the e cient breeding of new cultivars.Here, we document the inheritance of fruit texture traits and discuss the prospects for breeding new cucumber cultivars with disease resistance and desirable fruit texture traits.

Plant materials
Phenotypic data on fruit texture traits were collected in 2012, 2014, and 2016 at the Institute of Vegetable and Floriculture Science, National Agriculture and Food Research Organization, Mie, Japan.We used three populations of F 2 progeny of parents with high and low crispness (Fig. 1).Populations of 116 and 128 F 2 progeny (EC1 and EC2, respectively) were derived from a cross between high-crisp line 'Encore 10' ("E"), an inbred line derived from a commercial F 1 hybrid cultivar, and low-crisp line CS-PMR1 ("C"; weedytype cucumber).EC1 and EC2 populations were sown in plastic pots on 11

Evaluation of fruit texture
We evaluated the texture of the sampled fruits by puncture test as described previously (Yoshioka et al. 2009(Yoshioka et al. , 2010;;Shimomura et al. 2016).A 15-mm-thick transverse slice was cut from the middle of each fruit.The samples were punctured with a Texture Analyzer (Stable Micro Systems Ltd., Surrey, UK) tted with a 3.0-mm-diameter cylindrical plunger.The rate of advance was 150 mm min − 1 , and data were acquired at a resolution of 500 points per second.Transverse sections were punctured in the direction of the long axis of the fruit to a distance of 8 mm: the placenta was punctured once, and the eshy part was punctured at three separate locations.The middle two-thirds of each force-deformation curve (puncture depth 2.5 to 7.6 mm) was used to calculate placenta rmness (PFI), esh rmness (FFI), and three esh crispness scores: esh crispness index (FCI) and the apparent fractal dimension of each curve by the Richardson plot (FFDR) and box-counting (Kolmogorov's dimension; FFDK) methods (Russ 1994;Peleg 1997; Double-digest restriction-site-associated DNA sequencing (ddRAD-Seq) Library construction and sequencing analysis of the three populations and the parents were performed as described by Shirasawa et al. (2016) with minor modi cations.The ddRAD-Seq libraries were constructed with two combinations of restriction enzymes, PstI and MspI (Thermo Fisher Scienti c, Waltham, MA, USA).Digested DNA was ligated to adapters using T4 DNA ligase (Takara Bio Inc., Shiga, Japan) and puri ed using Agencourt AMPure XP cleanup reagent (Beckman Coulter, Brea, CA, USA) to eliminate short (< 300 bp) DNA fragments.Puri ed DNA was ampli ed by PCR with indexed primers.Ampli ed DNA fragments were puri ed using the Qiagen Mini Elute Kit (Qiagen, Hilden, Germany), and fragments 300-1000 bp in length were fractionated by 2.0% agarose gel electrophoresis (Kanto Chemical Co., Inc., Tokyo, Japan) with a QIAquick Gel Extraction Kit (Qiagen).These libraries were sequenced on a HiSeq X sequencer (Illumina, Inc., San Diego, USA) in 150-bp paired-end reads.

NGS data acquisition and SNP calling
The sequencing data without low-quality reads were trimmed and ltered in FaQCs v.

SSR markers
We used simple sequence repeat (SSR) markers from previous studies (Fukino et al. 2008;Ren et al. 2009;Cavagnaro et al. 2010) to test for polymorphism between parents of EC1.PCR ampli cation was carried out using a post-labeling method for multiplexed genotyping analysis with a bar-coded split tag (BStag) as described by Shimizu and Yano (2011) with minor modi cations.PCR was performed in a 10-µL reaction mixture that consisted of 1-5 ng of genomic DNA, 0.5 pmol of both the tagged forward primer and the uorescently labeled BStag primer, 2 pmol of reverse primer, and 2× Qiagen Multiplex PCR Master Mix (Qiagen).DNA was ampli ed in a Mastercycler Pro 384 (Eppendorf Corporation, Hamburg, Germany) under the following protocol: 95 °C for 5 min; 33 cycles of 95 °C for 20 s, 55 °C for 1 min 30 s, 72 °C for 30 s; an additional three cycles of 95 °C for 20 s, 49 °C for 1 min 30 s, 72 °C for 30 s; and 68 °C for 10 min.The sizes of the ampli ed fragments were estimated on an automated DNA analyzer (model 3730xl, Applied Biosystems, Foster City, CA, USA) with a GeneScan-500LIZ size standard (Applied Biosystems).Fragment length was determined in GeneMapper software (Applied Biosystems).

Linkage map construction and QTL analysis
Three genetic linkage maps were constructed in AntMap v. 1.2 software (Iwata and Ninomiya 2006) from the SNP markers obtained by ddRAD-Seq analysis.The SNP markers were selected by removing low-quality loci (> 5 progeny have missing values at each locus).For EC1, 103 additional SSR markers were used.QTL analysis for each texture trait was performed by composite interval mapping in Windows QTL Cartographer v. 2.5 software (Wang et al. 2007) with the following parameter settings: model 6, forward and backward stepwise regression model, 5 maximum background marker loci, window size 10, and 1 cM walking speed along chromosomes.The logarithm of odds (LOD) thresholds for QTL detection in both mapping methods were determined by 1000 permutations.

Results
The continuous distributions of fruit texture traits in the three F 2 populations suggest polygenic control of these traits (Fig. 2).As expected, the high-crisp parent ('Encore 10') had high crispness-related scores (FCI, FFDR, FFDK) and the low-crisp parents (CS-PMR1 and 'Atar') had low scores.On the other hand, placenta, skin, and esh rmness showed different features: both 'Encore 10' and 'Atar' had similar scores for PFI and SFI; and both parents of all populations had similar scores for FFI (Fig. 2; Table S1).
ddRAD-Seq analysis returned an average of 0.5 million high-quality reads per sample with an average Q30 quality score (indicating a 0.1% chance of error and 99.9% con dence) of 94.38 at 279 out of 2664 SNP sites in EC1, 1.7 million reads with Q30 = 90.26 at 607/3571 SNPs in EC2, and 1.3 million reads with Q30 = 94.54 at 496/2093 SNPs in EA.With the 103 additional SSR markers in EC1, we used 382 (EC1), 607 (EC2), and 496 (EA) markers for constructing linkage maps.Each map has seven linkage groups (LGs), and spans 946.31 cM (with an average distance between markers of 2.48 cM) in EC1, 661.81 cM (1.09 cM) in EC2, and 706.54 cM (1.42 cM) in EA.All sets of seven LGs correspond to the cucumber 'Chinese Long' v3 reference genome (Li et al. 2019).The order of markers was almost entirely conserved throughout the three maps (Fig. 3).EA had long regions with no markers (> 10 cM) of 14.13 cM on Chr. 2, 12.25 cM on Chr. 4, 11.15 cM on Chr. 5, and 16.95 cM on Chr. 6. EC1 had only one large gap, of 10.11 cM on Chr. 5. EC2 had no large gaps.
Many QTLs for fruit texture traits detected were signi cant at LOD thresholds of 3.5-3.8(Fig. 4).The LOD peaks of these QTLs ranged from 3.53 to 34.12, and the proportion of phenotypic variation explained (R 2 ) ranged from 0.04 to 0.55 (Table 1).QTLs for each texture trait were detected in the same or near regions of EC1 and EC2 if regions where their LOD scores exceeded the threshold overlapped.On the other hand, the detection of some QTLs in different regions between EC and EA was considered to be caused by the different parental line combinations.In total, 56 QTLs for texture traits were detected: 11 for PFI, 12 for SFI, 8 for FFI, 10 for FCI, 8 for FFDR, and 7 for FFDK (Fig. 4; Table 1).c R 2 is the proportion of phenotypic variance explained by the QTL.
All QTLs are named with the trait abbreviation followed by the LG number and the sequential number of the QTL for this trait (1, 2, 3, or 4) along the chromosome (Table 1).The QTLs for SFI on Chrs. 1 and 3 were detected in the same region among the three populations, but the contributing parents were different.These QTL names were appended with the initial letter of their parent name in superscript.Some major QTLs for different texture traits were detected in almost the same region (Fig. 4; Table 1): p -1.1, s -1.1, -1.1, fci-1.1,ffdr-1.1, and ffdk-1.1 on one part of Chr.1; s -1.2 E and -1.2 on another part of Chr.1; p -2.1 and s -2.1 on Chr.2; fci-3.1,ffdr-3.1, and ffdk-3.1 on Chr.3; p -6.1, s -6.1, and fci-6.1 on Chr. 6.

Discussion
In cucumber breeding, fruit texture traits are primary targets for improving fruit quality.Breeders have hitherto assessed fruit texture traits by sensory evaluation.Genotypic effects and heritability of fruit rmness have been investigated in crossing experiments, and it has been suggested that rmness is controlled genetically (Peterson et al. 1978;Kanno and Kamimura 1980;Yoshioka et al. 2009Yoshioka et al. , 2010)).However, few genetic studies of fruit crispness in cucumber have been performed (Yoshioka et al. 2009(Yoshioka et al. , 2010)).Evaluating fruit texture traits quantitatively by using mechanical measurement in segregating populations allows their genetic analysis.As all texture traits of our three populations of F 2 progeny were distributed widely, these traits are controlled quantitatively (Fig. 2 Linkage maps of three F 2 populations were constructed by using ddRAD-Seq analysis and compared (Fig. 3).The order of common markers is almost identical among the three populations and the cucumber reference genome (Li et al. 2019).Polymorphic loci were detected more or less evenly throughout each LG in EC1 and EC2.However, in EA, there were relatively large gaps, of > 10 cM, between markers.This result suggests that the 'Encore 10' inbred line and the 'Atar' inbred line have similar structures in part of the genome because they were both derived from current F 1 cultivars, and thus polymorphic sites cannot be detected in these parts.
Fifty-six QTLs for rmness (PFI, SFI, FFI) and crispness-related scores (FCI, FFDR, FFDK) were detected (Fig. 4; Table 1).Some of the QTLs with large effect for PFI and SFI were detected in the same regions in EC1 and EC2 but in different regions in EA.The esh is the most important part because it accounts for most of the cucumber fruit.Composite interval mapping detected 8 QTLs for FFI, 10 for FCI, 8 for FFDR, and 7 for FFDK.Those on Chr. 1 were detected in all populations (Fig. 4).However, their effects differed between EC and EA: In EC1 and EC2, -1.2 (in different locations from those for crispness-related scores), fci-1.1,ffdr-1.1, and ffdk-1.1 had large effects.In EA, on the other hand, -1.1 had a large effect, but fci-1.1 and ffdr-1.2had very small effects.Although there were some overlapping regions, crispness and rmness appear to be controlled by different mechanisms.Furthermore, QTLs for all crispness-related scores on Chr. 3 were detected in all populations, and some had high LOD scores (Fig. 4).Those on Chr. 6 were detected in both EC1 and EA (Fig. 4).Yoshioka et al. (2009) reported that fruit texture parameters differed signi cantly among cultivation periods.However, these QTLs, especially those on Chr. 3, were detected in multiple populations and cultivation periods, indicating their importance in controlling esh crispness traits.
There are no reports of QTLs for rmness and crispness in cucumber, but some reports describe QTL analysis and gene identi cation of fruit texture traits in other species.Several QTLs and candidate genes for fruit rmness are reported in melon (Moreno et al. 2008) Poles et al. 2020).Md-PG1 is related to cell wall structure and its modi cation.In the cucumber reference genome, a few polygalacturonase-like genes are located within the estimated QTL range for esh crispnessrelated scores detected in Chr. 3.However, esh crispness in cucumber still diffed among cultivars before harvesting (Shimomura et al. 2016).The results indicate that a variety of genetic mechanisms may govern fruit crispness.
CS-PMR1 can be used as breeding material because of its strong resistance to powdery mildew and downy mildew (Sakata et al. 2006;Fukino et al. 2013;Yoshioka et al. 2014).Some large-effect QTLs for resistance to these diseases are located near the QTLs detected here.In particular, the QTLs with large effects on SFI, FFI, and crispness-related scores on Chr. 1 identi ed here (Fig. 4) are located near some genes for resistance to powdery mildew (pm1.1 and pm1.2) and downy mildew (dm1.2 and dm1.).The locations of QTLs for fruit texture, morphological traits, and disease resistance on these chromosomes are highly complex.In developing new cultivars with introgressed traits, breeders repeat backcrossing and selecting over many years.However, the apparent linkage between fruit texture traits and resistance would reduce breeding e ciency in gene pyramiding.Our result will contribute to the development of new DNA markers tightly linked to QTLs for fruit texture traits that can be used in breeding.
Fruit rmness and crispness differ greatly among cultivars around the world, and consumers have different preferences.In cucumber, fruit texture is considered to be the most important deliciousness factor in Japan, because cucumber has little distinctive taste-related characteristics, such as sweetness, as are found in other fruit crops and vegetables.For example, esh crispness scores were signi cantly higher in Japanese F 1 hybrid cultivars than in other cultivars such as the Beit Alpha type and the European greenhouse type (Yoshioka et al. 2009;Sakata et al. 2011;Shimomura et al. 2016).Therefore, fruit crispness has been an important target in cucumber breeding.The newly identi ed QTLs for fruit rmness and crispness will allow us to develop DNA markers with the traits.In particular, the QTLs on Chr. 3, where QTL for esh crispness-related scores were detected among all populations, would be crucial to selecting for high crispness.We anticipate that our results will advance the development of informative DNA markers for fruit texture traits, and of new cucumber cultivars with desired fruit qualities.Frequency distributions of fruit texture traits in F2 populations derived from 'Encore 10' ("E") × CS-PMR1 ("C") crosses (EC1, EC2) and E × 'Atar' ("A") cross (EA).PFI, placenta rmness; SFI, skin rmness; FFI, esh rmness; FCI, esh crispness index; FFDR, esh apparent fractal dimensions based on Richardson plot; FFDK, esh apparent fractal dimension based on box-counting (Kolmogorov's dimension).Figure 3 Linkage map of 'Encore 10' × CS-PMR1 crosses (EC1, EC2) and 'Encore 10' × 'Atar' cross (EA).Linkage groups were assigned to cucumber reference genome chromosomes ('Chinese Long v3').Left position is EC1, center is EC2, and right is EA in each linkage group.Ladder steps in each linkage group indicate loci; black lines between linkage groups indicate common loci among (or between) populations.
Apr 2012 and 21 Apr 2014 and transplanted into a plastic greenhouse on 2 May 2012 and 9 May 2014.A population of 163 F 2 progeny (EA) was derived from a cross between 'Encore 10' and low-crisp line 'Atar' ("A"; Beit Alpha type), an inbred line derived from a commercial F 1 hybrid cultivar.The EA population was sown on 25 Apr 2016 and transplanted on 10 May 2016.Plants set fruit 1 month after transplanting.Fruits for puncture tests were harvested on 26 May to 15 June 2012, 11 June to 15 July 2014, and 13 June to 22 July 2016.These fruits weighed around 100 g, the Japanese market standard.

Roudaut et al. 2002 ;
Horie et al. 2004;Yoshioka et al. 2009).For the evaluation of skin rmness (SFI) with a 30° wedge plunger, the samples were punctured from outside to inside to a distance of 4 mm at three locations.The rst peak force in each forcedeformation curve was recorded and used as SFI.The data were analyzed in R software v. 4.0.2(R Development Core Team, 2005).
2.08 software(Lo and Chain 2014).Then the data were mapped to the cucumber reference genome ('Chinese Long' v3) sequence(Li et al. 2019) in the Cucurbit Genomics Database of the International Cucurbit Genomics Initiative (http://www.icugi.org) in BWA-MEM v. 0.7.17 software(Li 2013).The resulting sequence alignment/map (SAM) format les were converted to binary sequence alignment/map (BAM) format les.SNP calling and genotyping were performed with the mpileup tool in SAMtools v. 0.1.15and BCFtools v. 1.9 software (Li 2011).SNPs with > 8 reads in > 80% of progeny were used for analysis.
a b c R2is the proportion of phenotypic variance explained by the QTL. a p , placenta rmness; s , skin rmness; , esh rmness; fci, esh crispness index; ffdr, esh apparent fractal dimension based on Richardson plot; ffdk, esh apparent fractal dimension based on box-counting (Kolmogorov's dimension).bs -1.2 E is contributed by 'Encore 10' inbred line, s -3.1 A is contributed by 'Atar' inbred line, s -1.2 C and s -3.1 C are contributed by CS-PMR1.cR 2 is the proportion of phenotypic variance explained by the QTL. a p , placenta rmness; s , skin rmness; , esh rmness; fci, esh crispness index; ffdr, esh apparent fractal dimension based on Richardson plot; ffdk, esh apparent fractal dimension based on box-counting (Kolmogorov's dimension).bs -1.2 E is contributed by 'Encore 10' inbred line, s -3.1 A is contributed by 'Atar' inbred line, s -1.2 C and s -3.1 C are contributed by CS-PMR1.c R 2 is the proportion of phenotypic variance explained by the QTL.