Candidate genes for first flower node identified in pepper using combined SLAF-seq and BSA

First flower node (FFN) is an important trait for evaluating fruit earliness in pepper (Capsicum annuum L.), but the genetic mechanisms that control FFN are still poorly understood. In the present study, we developed 249 F2 plants derived from an intraspecific cross between the inbred pepper lines Z4 and Z5. Thirty plants with the highest FFN and 30 plants with the lowest FFN were chosen and their DNAs were pooled according to phenotype to construct two bulked DNA pools. Specific-locus amplified fragment sequencing (SLAF-seq) was combined with bulked segregant analysis (BSA) to identify candidate regions related to FFN. According to our genetic analysis, the FFN trait is quantitatively inherited. A total of 106,848 high-quality single nucleotide polymorphism (SNP) markers were obtained, and 393 high-quality SNP markers associated with FFN were detected. Ten candidate regions within an interval of 3.98 Mb on chromosome 12 harboring 23 candidate genes were identified as closely correlated with FFN. Five genes (CA12g15130, CA12g15160, CA12g15370, CA12g15360, and CA12g15390) are predicted based on their annotations to be associated with expression of the FFN trait. The present study demonstrates an efficient genetic mapping strategy and lays a good foundation for molecular marker-assisted breeding using SNP markers linked to FFN and for cloning and functional analysis of the key genes controlling FFN.


Introduction
Pepper (Capsicum annuum L.), Solanaceae, has a sympodial shoot structure with a solitary flower known as the first flower. During vegetative growth, the pepper shoot apical meristem (SAM) produces stems and leaves that are arranged in an alternate spiral pattern. The SAM can later undergo a transition to an inflorescence meristem that subsequently develops into the first flower at start of the transition from vegetative to reproductive growth [1,2]. Therefore, the formation of the first flower is a crucial phase in plant growth that is regulated by a among others. Integrated BSA and SLAF-seq strategies also have been successfully used for SNP discovery in pepper to refine the region containing a gene for resistance to Phytophthora root rot to a 2.57-Mb region [31]. This combined approach has also been used to identify one major QTL associated with resistance to Cucumber mosaic virus on pepper chromosome 02 [32]. These studies have demonstrated the efficiency of combined BSA and SLAF-seq as a strategy for the identification of genes or QTLs linked to a specific trait in plants.
Thus, in the present study we have used a combined BSA and SLAF-seq strategy to identify genomic regions linked to the FFN trait in DNAs pooled from distinct FFN phenotypes in an F 2 population derived from a cross between inbred parental lines Z4 and Z5 in pepper. Our objectives were to: 1) investigate the mode of inheritance of the FFN trait in pepper; 2) identify the genomic regions correlated with variation in FFN; and 3) identify candidate genes and SNP markers linked to the FFN trait in pepper.

Plant materials
The first flower node was measured as the number of nodes on the primary axis from the node of the cotyledon to that of the first flower. An F 2 mapping population comprising 249 plants was derived from a cross between the Capsicum annuum pepper lines Z4 and Z5, which were bred in Beijing Vegetable Research Center of the Beijing Academy of Agriculture and Forestry Sciences. Plants from line Z5 tend to be tall and have a high FFN (i.e., 19 nodes), while those from line Z4 tend to be of medium height with a low FFN (i.e., 9 nodes). Parental and F 2 plants were grown in the greenhouse at the Beijing Vegetable Research Center of the Beijing Academy of Agriculture and Forestry Sciences in Beijing, China.

Genetic analysis of the FFN trait
Phenotypic data were statistically analyzed using Microsoft Excel (Microsoft Office, Microsoft, 2003) and data were plotted using SigmaPlot 10.0 (SPSS Inc., Chicago, IL).

DNA extraction and construction of DNA pools
Total genomic DNA was isolated from young leaves of the both parental lines and F 2 plants using the cetyl trimethyl ammonium bromide (CTAB) method [33]. Two DNA pools were constructed by separately pooling an equal amount of DNA from each of 30 extreme high FFN plants (H-pool) or 30 extreme low FFN plants (L-pool) identified in the F 2 population.

Construction of SLAF libraries and high-throughput sequencing
The genome of Capsicum annuum cv. 'Criollo de Morelos 334' (CM334) (http://peppergenome. snu.ac.kr/download.php, version 1.55) was used as the reference genome in the present study. Genomic DNAs from both parents and the H-pool and the L-pool were digested with the restriction enzyme HaeIII (New England BioLabs, Ipswich, USA) after optimizing restriction enzyme digestion completeness to obtain even genome coverage. Single-nucleotide A overhangs were polished from these DNA fragments using Klenow fragment (New England Bio-Labs), and fragments were then ligated to dual-index sequencing adaptors [34]. Adaptor-ligated fragments were then amplified by PCR, purified, pooled, and screened to construct the SLAF library. Details of the processes for SLAF library construction and screening were performed as described in Sun et al. (2013) [26]. Target DNA fragments of sizes in the range of 414-514 bp were selected as SLAFs from the quality-tested library and prepared for paired-end sequencing on an Illumina HighSeq 2500 platform (Illumina, Inc., San Diego, CA, USA) at Beijing Biomarker Technologies Corporation in Beijing, China (http://www.biomarker.com.cn). To check the reliability and validity of sequencing and screening processes, the genome of rice (Oryza sativa L. japonica, http://rice.plantbiology.msu.edu/, version 7.0) was selected as a control to undergo in parallel the same library construction and sequencing processes as performed for the pepper mapping population.

Data analysis for SLAF-seq
Raw reads were filtered for quality and trimmed to remove adaptors, and then sequence quality was assessed based on sequencing quality scores and guanine-cytosine (GC) content [26]. The proportion of sequencing quality scores !Q30 in the four libraries was >80% (A quality score of Q30 indicates a 0.1% error rate or 99.9% sequence accuracy.). High-quality reads were mapped onto the pepper reference genome using BWA software [35]. We clustered all pairedend reads that had perfect index reads according to sequence similarity among both parents and the two pooled libraries using blast [36]. Sequences with >90% identity were grouped into a single SLAF locus (or SLAF tag).

Identification of high-quality SNPs
Single-nucleotide polymorphisms (SNPs) were detected primarily using GATK software [37]. Using clean reads mapped onto the reference genome, local realignments were conducted, and SNPs were detected using GATK software as described by https://www.broadinstitute.org/ gatk/guide/best-practices.php. To ensure the accuracy of SNPs identified using GATK, SAMtools software also was used to detect SNPs [38]. The intersection of SNPs that were detected using both GATK and SAMtools software was designated as final SNPs for further analysis. In addition, the localization (e.g., upstream, downstream, or intergenic regions) of SNPs, and the coding effects (e.g., synonymous or non-synonymous mutation) of SNPs were annotated using SnpEff software [39] based on gene model annotations at the pepper reference genome databases (http://peppergenome.snu.ac.kr/download.php, version 1.55). Before performing association analysis, SNPs were filtered using the following criteria: SNPs with multiple alleles were filtered out; SNPs with sequencing depths of less than 4× in each pool or parent were excluded; SNPs with the same genotypes among pools were removed; and SNPs with recessive alleles that were not inherited from parents with recessive genotypes in pools were filtered out. Ultimately, a collection of high-quality SNP markers was obtained for use in association analysis.

Association analysis
Association mapping was performed using either Euclidean distance (ED) [37] or the SNPindex algorithm [40,41] separately. ED between the allele frequencies at each SNP in the Lpool and H-pool was calculated as in Hill et al. (2013) using the equation [37] ED ¼ where each letter (A, C, G, T) indicates the frequency of its corresponding DNA nucleotide. ED values were then squared to decrease the effects of noise and increase the effects of large ED measurements based on distance measurements raised to a power (ED x ). Data were then fitted using Loess regression [37]. In addition, the threshold for the significance of markertrait associations was set at 1% of the biggest Loess-fitted values. The genomic regions at which the Loess-fitted values exceeded the threshold were then designated as candidate regions related to FFN in pepper.
The SNP-index algorithm is valuable for finding significant differences in the frequencies of genotypes between DNA pools. The frequency of genotypes in the SNP-index algorithm is denoted by SLAF depth [42]. The SNP-index is calculated as the proportion of the depth of Lpool or H-pool derived from the female parent relative to the two parental depths, and then the ΔSNP-index is defined by subtracting the H-pool SNP-index from the L-pool SNP-index. Therefore, the ΔSNP-index is equal to 0 if the SNP-index of the L-pool is equal to that of the H-pool. If the ΔSNP-index value equals 1, one genotype is associated almost entirely with the high FFN pool, and the associated SNPs are therefore linked closely to the high FFN phenotype. On the contrary, if the ΔSNP-index value equals -1, the associated SNPs are linked to the low FFN phenotype. The confidence coefficient of the ΔSNP-index was calculated, and all ΔSNP-index values were fitted using SNPNUM [38]. FFN-related regions were identified when the fitted values of markers were above the threshold at the 99% of confidence interval.
Finally, ED, SNP-index, and ΔSNP-index values were plotted and the intersections between the candidate regions for FFN that were identified using the ED and SNP-index methods were designated as the final candidate FFN-related regions. A circular graph representing the distribution of chromosomes, genes, SNPs, ED values, and ΔSNP-index values was then plotted using CIRCOS 0.66 software (http://circos.ca/).

Annotation of candidate genes
To verify the predicted genes in the target region, we compared these candidate genes to the CM334 reference genome using blast. Putative functions of candidate genes were predicted based on sequence alignments with annotated genes in the databases at Swiss-Prot (http:// www.uniprot.org/), the Gene Ontology (GO, http://www.geneontology.org/), the Cluster of Orthologous Groups of proteins (COG, http://www.ncbi.nlm.nih.gov/COG/), the Kyoto Encyclopedia of Genes and Genomes (KEGG, http://www.genome.jp/kegg/), and the NCBI nonredundant protein database (NR, ftp://ftp.ncbi.nih.gov/blast/db/) using blastp with default parameters.

Genetic analysis of the FFN trait
An F 2 population comprising 249 plants developed from the Z4 × Z5 cross was used to investigate the inheritance of FFN, which ranged in value from 8 to 20 (Fig 1). The continuous phenotypic distribution and transgressive segregation of FFN in the F 2 population in Fig 1 suggests that FFN is quantitatively inherited, and predicts that the FFN phenotype is likely controlled bysome major genes and some minor genes.

Evaluation of the SLAF library
HaeIII was chosen as the restriction enzyme for SLAF library construction based on preliminary experiments to identify a restriction enzyme that produced SLAFs evenly distributed on the pepper reference genome (S1A Fig). The SLAF libraries were then evaluated to verify the accuracy of sequencing compared with the rice genome control. First, the sequence reads of the rice genome control were compared with those of the pepper reference genome using BWA software [35]. The percentage of mapped paired-end reads in the pepper genome was 80.12% (Table 1), which is a typical efficiency for mapped paired-end reads from a SLAF library. Second, the efficiency of enzyme digestion is an important index for revaluating the likelihood of successful SLAF experiments. More efficient restriction enzyme digestion results in more successful SLAF experiments. In the present study, the efficiency of restriction enzyme digestion was 91.45% (Table 1), indicating that restriction enzyme digestion was adequate for constructing the SLAF libraries. Finally, average read length of SLAF was determined based on the insert size distribution of mapped paired-end reads in the rice genome. When the distance between reads at the ends of SLAFs was within 1 Kb, the integrity of SLAFs was greater than 0.1 and sequencing depth was greater than 3×, SLAFs of 414-514 bp in length were used for further analyses (S2 Fig).

Sequence data analysis and SNP identification
After constructing SLAF libraries and performing high-throughput sequencing, a total of 153,821,146 raw reads were obtained with an average read length of 100 bp ( Table 2). After   [35]. The proportion of clean reads that could be mapped to the reference genome was >88.81%, which also reflects sequencing accuracy ( Table 2). In total, cluster analysis identified 492,259 SLAFs (S1 Table) that were distributed evenly on the chromosomes of the pepper reference genome (S1B  (Table 2). A total of 1,001,405 SNPs were identified using GATK [37] and SAMtools software [38], and the properties of total SNPs were showed in S2 Table. The number of SNPs on each chromosome (shown in S1 Table) ranged from 21,371 on chromosome 08 to 150,965 on chromosome 11. The sequence coverage of the pepper genome was more than 0.042 when calculated across all markers. The distribution of SNPs on each pepper chromosome is shown in S1C Fig. After filtering, a total of 106,848 of high-quality SNPs remained as useful markers for association analysis to identify candidate regions associated with FFN in pepper.

Association analysis based on Euclidean distance
ED values were calculated for total of 106,848 high-quality SNP markers to identify genomic regions and markers associated with FFN. The ED value for each pair of SNPs was calculated. ED values were squared to decrease the effects of noise, and association values were then fitted by Loess regression [37]. The marker-trait association threshold was set to 0.31based on the 1% of the biggest Loess-fitted values. A graph was generated for the association values based on ED for each chromosome (Fig 2). Only one candidate region related to FFN for which the Loess-fitted value was above the threshold was detected on chromosome 12. In this region, we identified an interval from 196,328,926 to 210,751,601 bp containing 125 genes, among which two genes contained non-synonymous SNPs (Table 3). There were also 1,069 high-quality SNP markers within these candidate regions.

Association analysis based on SNP-indices
The SNP-indices for 106,848 SNP markers were calculated, and graphs of SNP-index values were drawn for the H-pool ( Fig 3A) and L-pool ( Fig 3B) as plots of average SNP-index values against each sliding-window position in the CM334 genome assembly. The ΔSNP-index was determined by subtracting the SNP-index of the L-pool from that of the H-pool, and was Table 2. Summary of sequence data from parental line DNAs and bulked DNA pools.  (Table 3). However, none of the identified genes carried non-synonymous SNPs. Final candidate regions tightly associated with the FFN trait in pepper were identified using a combination of ED and SNP-index association analysis. The same candidate regions were identified by ED and SNP-index association analyses within the interval from 0.00091 to 2.77 Mb on chromosome 12 that harbors 23 genes.

Visualization of the combined results of SLAF-seq and BSA
We used a strategy combining SLAF-seq and BSA to identify genomic regions, SNPs, and genes associated with FFN. All of the results of these approaches are visualized in the circular graph shown in S3  Candidate genes for first flower node in pepper represent the 12 pepper chromosomes, the distribution of genes on the pepper chromosomes, SNP density, ED values, and ΔSNP-index values, respectively.

Annotated SNP markers and genes within the candidate region
We identified a total of 393 high-quality SNP markers within the candidate regions (Table 3) Table 3). The localizations and coding effects of SNPs in these candidate regions were annotated using SnpEff software [39]. There were 3, 588, and 3 SNPs between genes in the Z4 and Z5 genomes in their upstream, intergenic, and downstream regions, respectively. We also identified 4, 418, and 1 SNPs between the H-pool and the L-pool in the upstream, intergenic, and downstream regions of genes, respectively (S3 Table). However, no non-synonymous SNPs were identified in the candidate regions among parental lines and both pools (S3 Table). Thus, more than 98% of SNP markers were located in intergenic regions.
We predicted that 23 protein-coding genes (Table 4), including nine genes with no annotations in the public databases, are be located within the 10 candidate regions for FFN on pepper chromosome 12, based on the current annotation of the CM334 pepper reference genome. We found annotations for 12, 5, 1, and 12 genes in the candidate regions in the GO, COG, KEGG,  Fig, 11 genes in the 'cellular component' domain are associated with the GO term 'cell part' (GO:0044464); four genes in the 'molecular function' domain are associated with the GO term 'catalytic activity' (GO:0003824), four genes in the 'molecular function' domain are associated with 'transporter activity' (GO:0005215), four genes in the 'molecular function' domain are associated with 'binding' (GO:0005488); and 10 genes are associated with the GO term 'cellular process' (GO:0009987) in the 'biological process' domain. COG analysis predicted five genes, among which the functions of CA12g15090, CA12g15100, and CA12g15320 were related to 'transcription', 'general function prediction only', and 'inorganic ion transport and metabolism', respectively. The function of CA12g15130 was related to 'intracellular trafficking, secretion, and vesicular transport', and that of CA12g15370 was associated with 'posttranslational modification, protein turnover, and chaperones' ( Table 4; Fig 4). KEGG analysis identified a match for only one gene (CA12g15130), a homolog of SEC22 in Arabidopsis thaliana that encodes a 25.3 kDa vesicle transport protein that takes part in both the 'phagosome' pathway (S5 Fig) and the 'SNARE interactions in the vesicular transport' pathway (S6 Fig). Five of these 23 candidate genes might be related to first flower node development based on our current annotation, but their functions must be further studied. CA12g15130 plays an important role in vesicle trafficking from the endoplasmic reticulum to the Golgi complex, and is highly expressed in Arabidopsis flowers [43]. CA12g15360, which encodes the pentatricopeptide repeat-containing protein At2g40720, was predicted to take part in xylem and phloem pattern formation Candidate genes for first flower node in pepper according to its GO annotation. The RING-H2 finger protein ATL52 (CA12g15370) might be involved in the protein 'ubiquitination' pathway. Hu et al. (2003) reported that ATL52 is expressed in Arabidopsis flowers [44]. Additionally, CA12g15390 matched a purine permease 1, which is highly expressed in leaves, stems, and flowers, but not in roots [45,46]. This protein has been predicted to function as a transporter for nucleotides, nucleosides, and their derivatives [45]. CA12g15160 is homologous to At1g48120 from Arabidopsis, which encodes a serine/ threonine-protein phosphatase 7 long form homolog that is expressed in the SAM, root tips, hydathodes, leaf vasculature, and mature flowers [47]. Because the stem and leaves arise from the SAM, this gene might be important for the development of the first flower node, as it might be required to maintain cell division activity in meristematic cells. These five predicted genes that are expressed in flowers, xylem and phloem, stem, and the SAM [43][44][45][46][47] participate in many important biological processes in relevant tissues and organs and might play important roles in the FFN trait in pepper. The molecular functions, if any, of these candidate genes in the control of FFN need to be further examined.

SLAF-seq and BSA has been successfully used in many gene mining research
SLAF-seq, a recently developed strategy for SNP discovery and large-scale genotyping [26], has been applied successfully for construction of linkage maps and analysis of QTL in many species [48][49][50][51][52]. Compared with earlier methods of marker development (e.g., restriction fragment length polymorphism (RFLP), random amplified polymorphic DNA (RAPD), amplified fragment length polymorphism (AFLP), and simple sequence repeat (SSR), SLAF-seq has several advantages including high accuracy, high throughput, efficiency, and cost effectiveness for SNP discovery and large-scale genotyping [26]. SLAF-seq takes advantage of large amounts of sequence data to develop the SLAF markers, improve marker development, and increase coverage of entire genomes [48]. In addition, BSA is an efficient way to identify markers that are specific for a trait of interest. We combined the SLAF-seq and BSA techniques to identify the genomic regions associated with the FFN trait in pepper. Previous studies have combined the SLAF-seq and BSA approaches successfully in many organisms [27][28][29][30][31][32]. For example, Xu et al.
(2016) combined SLAF-seq and BSA to map resistance to Phytophthora root rot to pepper chromosome 10 [31]. Guo et al. (2017) have also identified one major QTL linked to Cucumber mosaic virus resistance in the physical interval from 152.87 to 153.20 Mb on chromosome 02 in pepper using this integrated strategy [32].

We detected 10 FFN candidate regions efficiently by using SAF-seq combined with BSA
We obtained a total of 76.24 Mb of clean reads from 153.82 Mb of raw reads with an average read length of 100 bp, and evaluated sequencing depth and sequencing quality scores. For successful SLAF-seq, Sun et al. (2013) suggested that sequencing depth should be greater than 6× and quality scores should not be lower than Q30 [26]. The restriction enzyme digestion efficiency, sequencing depths, Q30 scores, and percentages of mapped paired-end reads achieved in our study demonstrate that the construction and sequencing of our SLAF library were sufficiently efficient, accurate, and high-quality. SNPs are important tools for molecular genetic analysis, due to their high frequency, wide distribution [53,54], high polymorphism, and ability to reveal fine-scale genetic variation [55]. Chen et al. (2015) showed that SNP markers are much more densely distributed than are SSR or other markers [54]. In total 1,001,405 SNPs were obtained in this study, and more than 21,371 SNPs were mapped to the chromosomes of pepper, covering the entire pepper genome more densely than in the map published by Cheng et al. (2016) [56]. The SNP coverage (more than 0.042) in our study was much higher than that in Chen et al. (2015) (~0.00208) [54], and provides adequate marker density and increases the accuracy of candidate gene identification. A total of 106,848 high-quality SNP markers were used for association analysis to identify candidate regions closely associated with FFN in pepper. To improve the accuracy of FFN candidate region identification, we overlapped results from ED association analysis and SNP-index association analysis, which accurately and quantitatively evaluated parental allele frequencies and the inheritance of parental alleles by F 2 progenies [57]. Although the criterions for the threshold set in ED analysis and SNP-index analysis were different, candidate regions identified in ED analysis were almost in coincidence with that in SNP-index analysis, revealing the accuracy for FFN candidate region identification. As the threshold at the 99% of confidence interval was used in SNP-index analysis, the results obtained in SNP-index analysis would be more accuracy than that in ED analysis. Therefore, the final candidate regions were then narrowed down to a 3.98-Mb interval on chromosome 12 harboring 393 high-quality SNP markers. Annotations for SNP markers revealed that these SNP markers, 98% of which were located in intergenic regions, were useful for refining the mapping of a gene related to FFN to a smaller region.
Since the CM334 reference genome was released in 2014, it has been available for comparisons of high-throughput sequencing results from other crosses to determine the distribution of DNA polymorphisms across the pepper genome [21]. FFN, an important criterion of fruit earliness in pepper, is quantitatively inherited [11]. When FFN is lower and plants are shorter, fruits develop earlier [9,10]. We mapped the FFN trait to 10  In previous studies, QTL for length of the primary axis and leaf number on the primary axis have also been mapped to chromosome 12 [13,19]. Additionally, FFN is an important factor related to flowering time [16], which has also been mapped to chromosome 12 [13]. Because FFN is positively correlated with plant height, primary axis length, the number of leaves, and the number of branches [10] in pepper, loci in addition to FFN on chromosome 12 are likely important for early vegetative development, which is consistent with the results of Alimi et al. (2013) [19]. However, Alimi et al. (2013) mapped leaf number on the primary axis to a region different than that related to FFN at 23.1 cM on chromosome 12 [19]. We conclude that the 3.98-Mb interval associated with FFN on pepper chromosome 12 is a strong candidate region for FFN in pepper that could contain the gene(s) controlling this trait.

We identified 5 candidate genes correlated to FFN in pepper
Among the 23 candidate genes for which annotations were found in the GO, COG, KEGG, and Swiss-Prot databases, some might be related to FFN, and should be useful for future gene isolation and functional testing studies. S4, S5 and S6 Figs, Fig 4 and Table 4 show details of the annotations from the GO, COG, KEGG and Swiss-Prot databases of the 23 candidate genes identified in this study that can be related to FFN. The transition of the SAM from vegetative to reproductive growth [1,2] reveals its importance in the development of the first flower. In previous studies, FA, CaJ, Ca-ANANTH, Ca-LEAFY, Capsicum annuum S, and CaBLIND were found to promote flower development [1][2][3][4][5], while the CaRNA-binding protein and Ca-APETALA2 were found to suppress flower formation [7,8]. Tan et al. (2015) annotated three homologs of Arabidopsis APETALA2 and CLF [14] as related to leaf number on the primary axis in pepper, which might also be related to FFN. FFN is not only relevant to flowering, but also to plant growth and development, including plant height, primary axis length, lateral branch number, and leaf number on the primary axis [9,10]. Based on annotations in our study, we have identified five candidate genes correlated to FFN in pepper that are homologous to flower or stem development-related genes in Arabidopsis. CA12g15160, which is homologous to Arabidopsis At1g48120 and encodes a serine/threonine-protein phosphatase 7 long form homolog, is particularly interesting because it is closely related to development of the SAM [47]. Due to the importance of the SAM in the development of the first flower, the gene CA12g15160 is an important candidate gene for FFN in pepper. Additionally, CA12g15390, CA12g15370, and CA12g15130 are also highly expressed in Arabidopsis flowers [43][44][45][46], and the GO annotation for CA12g15360 indicates that it could be related to xylem and phloem development. The protein encoded by CA12g15130 might take part in both the 'phagosome' pathway and the 'SNARE interactions in the vesicular transport' pathway. Although these genes that were annotated based on the reference genome are homologous to genes related to flower or stem development in Arabidopsis, none of previously known flowering genes have been found within the candidate regions. On the one hand, the constructed reference genome is still not perfect enough to find all the genes controlling pepper flowering due to the large pepper genome (3.48 Gb) [21]. On the other hand, BSA used in the study can be extended to identify a few major genes that controlling quantitative traits, but it is not useful to the analysis of minor genes [58]. Additionally, we still lack direct evidence that these genes control the FFN trait of pepper. Future functional analyses of these candidate genes could reveal whether they play any parts in the control of the FFN trait in pepper.

Conclusions
In the present study, we combined the SLAF-seq technology and BSA to successfully identify genomic regions involved in control of the FFN trait in pepper (Capsicum annuum lines Z4 and Z5). This method proved efficient for mapping genes related to the FFN trait in the reference genome (Capsicum annuum line CM334). Ten candidate regions within a 3.98-Mb interval on chromosome 12 were detected and associated with expression of the FFN trait. Five out of 23 candidate genes, in particular CA12g15160, were chosen for further analysis based on their annotations. The functions of the genes correlated to FFN will be further examined in future studies using transformation and mutation approaches.