The APETALA2 homolog CaFFN regulates flowering time in pepper

Flowering time is an important agronomic trait that contributes to fitness in plants. However, the genetic basis of flowering time has not been extensively studied in pepper. To understand the genetics underlying flowering time, we constructed an F2 population by crossing a spontaneous early flowering mutant and a late-flowering pepper line. Using bulked segregant RNA-seq, a major locus controlling flowering time in this population was mapped to the end of chromosome 2. An APETALA2 (AP2) homolog (CaFFN) cosegregated with flowering time in 297 individuals of the F2 population. A comparison between the parents revealed a naturally occurring rare SNP (SNP2T > C) that resulted in the loss of a start codon in CaFFN in the early flowering mutant. Transgenic Nicotiana benthamiana plants with high CaFFN expression exhibited a delay in flowering time and floral patterning defects. On the other hand, pepper plants with CaFFN silencing flowered early. Therefore, the CaFFN gene acts as a flowering repressor in pepper. CaFFN may function as a transcriptional activator to activate the expression of CaAGL15 and miR156e and as a transcriptional repressor to repress the expression of CaAG, CaAP1, CaSEP3, CaSOC1, and miR172b based on a qRT-PCR assay. Direct activation of CaAGL15 by CaFFN was detected using yeast one-hybrid and dual-luciferase reporter assays, consistent with the hypothesis that CaFFN regulates flowering time. Moreover, the CaFFN gene association analysis revealed a significant association with flowering time in a natural pepper population, indicating that the CaFFN gene has a broad effect on flowering time in pepper. Finally, the phylogeny, evolutionary expansion and expression patterns of CaFFN/AP2 homologs were analyzed to provide valuable insight into CaFFN. This study increases our understanding of the involvement of CaFFN in controlling flowering time in pepper, thus making CaFFN a target gene for breeding early maturing pepper.


Introduction
The timing of the transition from vegetative growth to flowering is a major factor that ensures that plants produce enough progeny for their sustainable development [1][2][3] . Thus, flowering time is an important agronomic trait that contributes to fitness in annual plants. Genetic analysis of the model plant Arabidopsis thaliana has identified numerous genes involved in flowering-time control that may participate in the photoperiod, vernalization, ambient temperature, aging, autonomous or gibberellin pathways 4 , each of which plays either a floral inhibiting, neutral, or promoting function in response to exogenous or endogenous signals. Downstream of these pathways are a small number of floral integrator genes, including FLOWERING LOCUS T (FT) and SUPPRESSOR OF OVEREXPRESSION OF CONSTANS 1 (SOC1) [5][6][7][8][9] . 'Integrators' can activate floral meristem identity (FMI) genes, such as LEAFY (LFY), APETALA1 (AP1), APETALA2 (AP2), FRUITFULL (FUL), and CAULIFLOWER (CAL), which encode proteins that promote floral development 1 . Among these FMI genes, the transcription factor AP2 directly promotes the expression of the floral repressor AGAMOUS-LIKE15 (AGL15) and directly represses the expression of the flowering promoting transcripts SOC1, AP1, SEPALLATA3 (SEP3), AGAMOUS (AG), and FUL 10 . In addition, AP2 positively regulates miR156 and negatively regulates miR172 to form a complex direct feedback loop that influences AP2 expression 10 .
Solanaceae includes species such as potato (Solanum tuberosum), tomato (S. lycopersicum), pepper (Capsicum annuum), eggplant (S. melongena), and tobacco (Nicotiana tabacum) that are of agricultural interest or research interest. Among these species, tomato has been more extensively studied for the control of flowering time [11][12][13] . As a member of the Solanaceae family, pepper (C. annuum L.) has a sympodial shoot structure. The pepper shoot apical meristem (SAM) first produces stems and leaves alternately in the vegetative phase and then, upon the transition to the reproductive phase, terminates in an inflorescence meristem (IM) that subsequently develops into the first flower; plant growth continues from the upper most axillary meristem 14,15 . Flowering time in pepper can be best measured by counting the number of leaves (or nodes) on the primary stem (Nle) from the cotyledon until the first flower, which has also been described as the number of first flower nodes (FFNs) [16][17][18] .
In pepper, several genes controlling flowering time have been identified by screening an EMS-mutagenized population. Of these genes, mutants of CaJOINTLESS (CaJ), CaBLIND (CaBL) and CaS were late flowering 14,15,19 , and mutants of fasciculate (fa) and E-62 were early flowering 20,21 . These genes have a pleiotropic effect on plant growth in addition to affecting flowering time in pepper; for example, CaS is required for flower formation, and CaBL regulates axillary meristem initiation 15,19 . However, most of these mutant genes are unlikely to be utilized in pepper breeding; previous studies have reported that mutants of CaS were characterized by complete inhibition of flower formation 15,19 , and mutants of CaBL had no axillary shoot development in the vegetative phase and only two sympodial units in the reproductive phase 19 .
Pepper germplasm resources exhibit extensive natural variation in flowering time 22 . Genetic analysis showed that the flowering time trait in pepper was a quantitative trait commonly controlled by a few major genes and some minor genes, as well as environmental factors such as temperature and photoperiod 16,17,23,24 . Quantitative trait loci (QTLs) affecting flowering time have been identified on different chromosomes using different populations 17,18,[23][24][25][26] . For example, Tan et al. 17 mapped one major and five minor QTLs affecting FFNs on chromosomes P2, P7, P10 and P11 based on an interspecific genetic map, and 12 genes were recommended as important candidate genes for the major QTL Nle2.2. Zhang et al. 18 identified 23 candidate genes on chromosome 12 within an interval of 3.98 Mb using combined specific-locus amplified fragment sequencing (SLAF-seq) and bulked segregant analysis (BSA). Zhu et al. 26 detected three flowering time QTLs based on an interspecific F 2 population, in which an AP2 protein encoding gene Capana02g000700 was predicted to be the candidate gene for the major QTL. However, further studies to identify and validate the function of candidate genes controlling flowering time in pepper have been limited to date.
The early flowering pepper B 9431 , which can blossom and bear fruit at the first node on the main stem, is an ideal natural material for determining the genetic control of flowering time in pepper. In this study, an F 2 population with 297 plants derived from an intraspecific cross between inbred lines B 9431 (early flowering) and A 145 (late flowering) was developed. The locus controlling flowering time in this population was mapped by the bulked segregant RNA-Seq (BSR-Seq) method. The heterologous overexpression assay in N. benthamiana and the virusinduced gene silencing (VIGS) assay in pepper A 145 were performed to verify the function of the candidate gene CaFFN. Similarly, a real-time fluorescent quantitative PCR assay, a yeast one-hybrid assay and a dual-luciferase reporter assay were conducted to determine how the CaFFN gene influences the flowering time of pepper. Furthermore, CaFFN gene association analysis was performed on 167 pepper inbred lines to test the effect of this gene on the control of natural variations in flowering time. The phylogeny, evolutionary expansion and expression patterns of CaFFN/AP2 homologs were analyzed to provide useful insight into CaFFN.

Results
Genetic analysis of the flowering time trait in pepper A natural precocious variant was discovered in our breeding work, from which a stable early flowering inbred line named B 9431 was created after years of selfing and selective breeding (Fig. 1a). B 9431 can flower after the development of one leaf on the primary stem. To investigate the inheritance of flowering time in pepper, an F 2 population including 297 plants developed from a cross between B 9431 and A 145 was studied. Flowering time, in this study, was measured by counting the number of leaves on the primary stem between the cotyledon and the first flower (referred to as the number of FFNs) according to previous studies [16][17][18] . The average number of FFNs on parent B 9431 was 2.3 (ranging from 1 to 4), parent A 145 was 14 (ranging from 13 to 15), and the F 1 progeny was 8.5 (ranging from 8 to 10); that on the F 2 population ranged from 2 to 14. The F 2 population showed a bimodal distribution (Fig. 1b). Based on a distribution of the F 2 population at the low point (number of FFNs of 5), the proportion of late-flowering (number of FFNs ranging from 6 to 14) to early flowering (number of FFNs ranging from 2 to 4) peppers was~3:1 (χ 2 = 1.08; p = 0.30), thus indicating the presence of one strong major gene for this trait 27,28 .

BSR-Seq identified the major locus underlying flowering time in pepper
Thirty plants with the minimum number of FFNs from the F 2 population were chosen to construct an 'early flowering pool', and 30 plants with the maximum number of FFNs were chosen to construct a 'late-flowering pool'. The total RNA of the two pools was extracted for Illumina high-throughput sequencing, and the clean data were aligned to the C. annuum cv. CM334 genome chromosomes (release 1.55) (http://peppergenome.snu.ac.kr/). The frequencies for the single-nucleotide polymorphisms (SNPs) thus detected were calculated for each pool. The average difference in the SNP frequency (index) between the two pools was calculated across a 1.0 Mb genomic interval using a 3.0 Mb sliding window, and the results were plotted against all twelve chromosomes of the reference genome (Fig. 1c). The figure shows an obvious peak in the SNP delta values at the distal end of chromosome 2, which indicates one strong major locus governing flowering time in this region.
Identification of a candidate gene with a point mutation at the start codon in the early flowering parent B 9431 At the end of chromosome 2, the AP2 transcription factor-encoding gene CA02g14540 caught our attention, since its likely ortholog AP2 in Arabidopsis has numerous functions, including participation in floral transition and development. The sequences of this gene in the two parents of the F 2 population were determined by sequencing the PCR products amplified with the primers listed in Supplementary Table S1. A point mutation at the start codon ATG (referred to as SNP2T > C) was discovered in the early flowering parent B 9431 (Fig. 2a). This SNP led to the loss of the original start codon of CaFFN and resulted in a putative frameshift mutation of CaFFN in B 9431 . To investigate the potential of this gene to be a candidate gene controlling flowering time in this study, a cleaved amplified polymorphic sequence (CAPS) marker CSF2 (Fig. 2b, Supplementary Table S1) was developed based on SNP2T > C. The 297 individuals in the F 2 population of B 9431 × A 145 were genotyped using marker CSF2. The results showed that all early flowering (number of FFNs ranging from 2 to 4) plants had the same genotype as B 9431 , and all late-flowering (number of FFNs ranging from 6 to 14) plants had the same genotype as A 145 or the F 1 (Fig. 2b), indicating cosegregation of CSF2 with flowering time. These results demonstrated that CA02g14540 was a candidate gene (referred to as the CaFFN gene) underlying flowering time in the F 2 population of B 9431 × A 145 . CaFFN is an AP2-like gene with a putative conserved miR172 target site. It is thus speculated that a modification of the miR172 target site would increase the expression of the CaFFN gene. For this reason, 22 transgenic lines transformed with modified CaFFN cDNA containing six mismatches (synonymous mutations) to miR172 (miR172-resistant CaFFN mutant, 35S:: CaFFNm3) were generated (Fig. 2a). Among them, two transgenic lines (lines 10 and 16) exhibited a notable delay in flowering time (Fig. 3a).  Table S1) specific to CaFFN were designed to examine the expression level of CaFFN in N. benthamiana. The results revealed that the expression level of the CaFFN gene was higher in these two phenotypic lines than in other plants (Fig. 3c). These two transgenic lines produced numerous leaves in an extended period of vegetative growth and flowered two and a half months later than control plants. In addition to the delayed-flowering phenotype, sterility and floral patterning defects, which were characterized by shorter filaments and poorly developed petals with a reduced size, were observed in line 10 (L10) and line 16 (L16) (Fig. 3b). L16, with the highest level of CaFFN mRNA, exhibited the most dramatic change; it produced the smallest petals and an even shorter style (Fig. 3b). These results provide evidence that the CaFFN gene has a function in regulating flowering time and floral development.

Silencing of the CaFFN gene promotes early flowering in pepper
To study the functional roles of CaFFN in flowering time in pepper, loss-of-function experiments were performed on parental line A 145 , in which CaFFN was silenced by VIGS. Agrobacterium carrying the tobacco rattle virus (TRV)-based vector TRV2::CaFFN mixed with Agrobacterium carrying the TRV1 vector were infiltrated into pepper seedlings at the four-leaf stage. Seedlings injected with TRV2::PDS-and TRV1-containing Agrobacterium were used as controls to measure the effect of gene silencing, which induced a photobleaching phenotype. Seedlings injected with TRV2-and TRV1-containing Agrobacterium were used as negative controls. The infiltrated seedlings were grown at 20 ± 1°C and 90 ± 5% relative humidity to enhance silencing efficiency. Plants injected with TRV2::PDS showed photobleaching (

Gene expression changes in response to CaFFN silencing
The above results in this study indicated that CaFFN is a repressor of flowering. However, the mechanism underlying flowering time alteration caused by CaFFN remains to be determined. As the closest homolog to AP2, determining whether the network surrounding AP2 is conserved and applicable to CaFFN is worthy of our attention. For this purpose, the likely orthologs of AG, AGL15, AP1, SEP3, and SOC1 (CaAG [CA11g09730], and CaSOC1 [CA01g19500]) were searched by reciprocal best-hit analysis in the C. annuum cv. CM334 genome (release 1.55) (http://peppergenome.snu. ac.kr/). The expression levels of these flowering time and floral organ development genes were analyzed by qRT-PCR relative to the housekeeping gene CaGAPDH [CA03g24310]. The results showed that the expression levels of CaAG, CaAP1, CaSEP3, and CaSOC1 were significantly higher in CaFFN-silenced plants than in the control (P < 0.01). In contrast, the expression level of CaAGL15 in CaFFN-silenced plants was significantly lower than that in the control (P < 0.01) (Fig. 4b). The expression levels of microRNAs were determined relative to CaU6 [CA07g02150]. MiR172b exhibited an increased level and miR156e represented a reduced level in CaFFNsilenced plants (Fig. 4b). The results suggested the role of CaFFN as a transcriptional activator to promote the expression of CaAGL15 and miR156e and as a  The finding that the expression level of the floral repressor gene CaAGL15 was reduced in CaFFN genesilenced pepper plants prompted us to examine whether CaAGL15 is directly regulated by CaFFN. We checked the ability of CaFFN to bind to the cis-elements of CaAGL15 using a yeast one-hybrid assay. The full-length cDNA of CaFFN was cloned in frame with the GAL4 activation domain (AD) in the pGADT7 vector. The cis-elements of CaAGL15 were inserted into the pAbAi vector. The minimal inhibitory concentration of aureobasidin A (AbA) was 400 ng/ml for bait strains of CaAGL15. The results showed that CaFFN directly interacted with CaAGL15, since yeast transformed with pGADT7-CaFFN and the bait vector grew in SD/-Leu selection medium supplemented with the minimal inhibitory concentration of AbA, while yeast transformed with pGADT7-lam and pAbAi-P53, as a negative control, did not (Fig. 4c). In addition, a dual-luciferase assay was performed in N. benthamiana leaves. Recombinant pGREEN0800-CaAGL15 containing the promoter region of CaAGL15 (reporter) and the fused protein with CaFFN (effector) were coinfiltrated into leaves. Cotransformation of CaAGL15 and the CaFFN protein resulted in a twofold LUC/REN ratio compared with the control (Fig. 4d). The results indicated that CaFFN can bind to the CaAGL15 promoter in vivo and transactivate the expression of the reporter gene.
CaFFN was significantly associated with flowering time in the pepper population In this study, the CaFFN gene was shown to be a flowering repressor in pepper. To determine the distribution of the SNP2T > C mutation of the CaFFN gene in the pepper population, the CaFFN sequences of 383 pepper accessions were surveyed in PepperPan, a pepper pangenome browser 29 . The results showed that SNP2T > C was a rare SNP that occurred in only three additional accessions, CXJ21 (B33), CXJ22 (B35), and CXJ70 (A100) (Fig. 5a), and these three accessions were early flowering materials, with FFN numbers of 3, 3, and 5, respectively, suggesting that SNP2T > C could affect flowering time in multiple genetic backgrounds. The rare frequency of SNP2T > C also implies that the mutation occurred very recently and has not been widely used in breeding programs.
Due to the low number of accessions with nucleotide C at position 2, SNP2T > C could not be validated by association analysis. To further investigate whether CaFFN is an important regulator of flowering time in the pepper A sequence comparison between the CaFFN alleles of 53 pepper accessions did not reveal clear causative polymorphisms in the coding region. It is speculated that the 5′-UTR may influence the number of FFNs by affecting the expression of the CaFFN gene. To test this hypothesis, 28 randomly selected accessions were divided into two haplotypes according to SNP-50G > A, and the expression level of the CaFFN gene was determined by qRT-PCR (Fig. 5b). Statistical analysis showed that SNP-50G > A significantly correlated with the expression level of the CaFFN gene (P < 0.05), suggesting that the expression level of CaFFN might be influenced by its 5′-UTR.

Characterization of the CaFFN/AP2 homologs in pepper
CaFFN has two AP2 domains and is a member of the AP2 family gene. AP2 family genes with two or more AP2 domains play essential roles in plant growth and development [30][31][32] , as shown in this study. To systematically analyze the AP2 family genes in pepper, the HMM profile with the AP2 domain (PF00847) was used as a query, and a total of 24 AP2 family genes were obtained in the genome database of the cultivated pepper Zunla-1 (Fig. 6). Duplication events are known to be the major driving force in the expansion of gene families [33][34][35] . To explore the major driving force for the expansion of the AP2 family in pepper, the duplication types of each gene were detected by the MCScanX package, including singleton, dispersed, proximal, tandem and whole-genome duplication/triplication (WGD/T) or segmental duplication. Combined with the data of the retained triplicate genes from WGT in pepper 36 , 50.0% (12) of the pepper AP2 family genes were assigned to WGD/T or segmental duplication types (Fig. 6). This indicated that WGD/T or segmental duplication was the major force for the expansion of the AP2 family genes in pepper. Among them, the CaFFN (Capana02g003062) gene was located in a large collinear block containing 98 genes. The timing of this duplication event was estimated to be~47.31 million years ago (Mya) based on the mean Ks value of segmental duplicated gene pairs within this collinear block. Thus, the segmental gene pair CaFFN and Capana02g000700 may have arisen from a WGT before pepper-tomato divergence (20 Mya) and after pepper-grape divergence (89 Mya) with a copy of the triplicated gene lost 36 .
To compare the expression patterns of the CaFFNrelated genes in pepper, Arabidopsis and tomato, the expression patterns of these genes in SAM at three different developmental stages were plotted in a heatmap (Fig. 6). Similar to the CaFFN gene, numerous CaFFNrelated genes showed a relatively high expression level at the vegetative meristem stage and a low expression level at the transition meristem stage, except for Capa-na04g002188, Capana09g001880, Capana10g001776, Capana10g001789, Capana11g000645, and TOE3.

Discussion
In this study, the locus controlling flowering time in the Putative miR172 target sites in AP2-like genes are conserved in eudicots, monocots, and early land plants [37][38][39] , indicating that the expression of AP2-like genes could be regulated by miR172. Mlotshwa et al. 38   In Arabidopsis, AP2 is involved in various developmental processes at the shoot apex 40,41 . The expression patterns of AP2 and CaFFN in SAM were consistent with their function as flowering repressors, which showed a high expression level at the vegetative meristem stage and a low expression level at the transition meristem stage. Similar expression patterns were found for nine additional homologs in Arabidopsis and tomato, indicating that these genes may have similar/overlapping functions. Previous research has demonstrated that six AP2 domaincontaining genes, including AP2, showed functional redundancy of their role in flowering time in Arabidopsis; the early flowering phenotype of the ap2 mutant was largely obscured by the effects of the other five members of the euAP2 lineage 10 . In contrast, all of the CaFFN mutants in C. annuum, including E-62 20 , B 9431 , B33, B35, and A100, exhibited an obvious early flowering phenotype, which indicated that the flowering time control function of this gene had little redundancy with other genes in C. annuum. This result was further confirmed by the expression patterns of CaFFN/AP2 homologs in the SAM of C. annuum, which differed from that of most of the detected CaFFN/AP2 homologs (Fig. 6), representing a functional differentiation of these homologs. On the other hand, no floral patterning defects were discovered in E-62, B 9431 , B33, B35, and A100, indicating that the function of floral patterning regulation of the CaFFN gene was redundant and could be compensated by other gene (s) in pepper.
Capana02g000700, as a paralog of CaFFN that has arisen from an ancient WGT (~47.31 Mya), might have some functional redundancy with CaFFN since it presented a similar expression pattern as CaFFN in the SAM of pepper (Fig. 6). The possible function of Capa-na02g000700 in flower development and transcriptional repression of flowering time was also reported in a previous study 26 . However, Capana02g000700 may only have a weak effect on flowering repressor in C. annuum, since its expression level in the early flowering inbred line CA1 (C. annuum) was much lower than that in the late flowering inbred line 740 (C. chinense) 26 . This is a reasonable explanation for why the early flowering phenotype of the CaFFN mutants in C. annuum was not obscured by Capana02g000700.
To investigate how the CaFFN gene participates in the complicated network to affect flowering time and floral organ morphogenesis, the transcription profiles of several key flowering time and organ development genes were assayed by qRT-PCR. The results suggested that CaFFN controls flowering time and floral development by regulating genes that are involved in the aging pathway (Fig.   4b): CaFFN functions as a transcriptional repressor to suppress the expression of the floral activator genes CaAP1, CaSEP3, and CaSOC1 as well as the key floral identity gene CaAG; it also functions as a transcriptional activator to activate the expression of the floral repressor gene CaAGL15. In addition, CaFFN negatively regulated miR172 and positively regulated miR156, as indicated by their expression relationship. Moreover, yeast one-hybrid assays and dual-luciferase reporter assays revealed that CaFFN directly activates the floral repressor gene CaAGL15. These results provide the first insight into the mechanisms employed by CaFFN to regulate flowering time and floral patterning in pepper.
Genome-wide association analysis (GWAS) of 36 agronomic traits, including the FFN trait, in 287 pepper accessions was carried out based on SLAF-seq data, but no peak regions associated with the FFN trait were detected 22 . It is unclear whether the CaFFN gene has a broad effect on flowering time in the pepper population. In our study, CaFFN gene association analysis was performed in 167 pepper inbred lines. A SNP 50 bp upstream of the start codon was significantly associated with the number of FFNs, explaining 14.5% of the phenotypic variation. This result indicated that the CaFFN gene has a broad effect on flowering time in the pepper population. This broad effect is not related to the SNP2T > C mutation at the start codon of CaFFN, which is a rare mutation in the pepper population, but is most likely due to the variation(s) in the 5′-UTR of the CaFFN gene. In addition to the result that the SNP in the 5′-UTR (SNP-50G > A) of CaFFN was significantly associated with the FFN trait, the results that SNP-50G > A significantly correlated with the expression level of the CaFFN gene indicated by qRT-PCR also support the speculation that the 5′-UTR may influence the number of FFNs by affecting the expression of the CaFFN gene in the pepper population. Whether SNP-50G > A is the causative mutation that influences the expression of CaFFN or a SNP tightly linked with the causative mutation needs further research.
The control of flowering time in pepper is an attractive objective for breeding aimed at creating locally adapted cultivars. In pepper, several genes, including CaJ, CaBL, CaS, FA, and CaAP2 14,15,[19][20][21] , were reported to be flowering regulators. Only FA and CaAP2/CaFFN were described as flowering repressor genes, and the effect of the FA gene on flowering time was minor 20,21 . Despite a certain degree of reduction in shoot and fruit size and the extreme flowering time phenotype in the homozygous mutant of the CaFFN gene 20 , it is still an ideal breeding material for genetic improvement. Mutated CaFFN could significantly reduce the number of FFNs with minimal negative impact on yield in multiple genetic backgrounds when heterozygous, making it very useful for hybrid pepper breeding. This gene was used to select early maturing pepper varieties in our pepper breeding experiments.

Plant materials
The precocious pepper B 9431 was crossed with the lateflowering inbred line A 145 (from the Asian Vegetable Research and Development Center). The F 1 hybrid was selfed to generate a segregating F 2 population. A total of 297 individuals of the F 2 population were used for mapping. The flowering time of pepper was measured by counting the number of leaves on the primary stem between the cotyledon and first flower. All plants were grown in a greenhouse in Nanchang (28°33′N, 115°56′E, China).

BSR-seq
BSR-Seq combined bulked-segregant analysis and RNA-Seq and was performed to map loci for target traits. For the genetic analysis, two pools were constructed; each pool contained 30 individuals from the F 2 segregating population with extreme phenotypes. For the phenotyping of the FFN trait, total RNA of leaves at the blooming period was extracted using RNAiso Plus reagent (Takara, Dalian, China). RNA was quantified and assessed using a Qubit Fluorometer (Invitrogen, Carlsbad, USA), Nanodrop spectrophotometer (Thermo Fisher Scientific, Wilmington, USA) and Agilent 2100 bioanalyzer (Agilent Technologies, Palo Alto, USA). Paired-end sequencing was carried out on an Illumina HiSeq X-ten instrument. Clean RNA-seq data were mapped to the C. annuum cv. CM334 genome chromosomes (release 1.55, http:// peppergenome.snu.ac.kr/) 42 using BWA software 43 . SNPs were screened using SAM tools 44 . The differences in allele frequencies of the 12 pepper chromosomes between the two pools were calculated and plotted.

SNP marker development and genotyping
To develop a SNP marker for the CaFFN gene, which was speculated to be the candidate gene controlling flowering time in pepper, the genomic sequences of the CaFFN gene in B 9431 and A 145 were amplified (with primers listed in Supplementary Table S1) and sequenced. The polymorphic SNP (SNP2T > C) at position 2 of the CaFFN gene between the B 9431 and A 145 genotypes was converted into the CAPS marker CSF2 for genotyping. PCR was carried out in B 9431 , A 145 , the F 1 population, and the F 2 population (297 individuals), and each reaction had a total volume of 10 μl containing 50 ng genomic DNA, 0.2 μM primer FFN1-5UF4, 0.2 μM primer FFN1-e1R2 (Supplementary Table S1) and 5 μl 2×Taq Master Mix (Novoprotein Scientific Inc., Shanghai, China). PCR was performed with the following conditions: denaturing for 1.5 min at 94°C, followed by 38 cycles of at 94°C for 20 s, 55°C for 20 s, and 72°C for 1 min, and a final elongation at 72°C for 5 min. The amplified PCR products were digested with the restriction enzyme TaiI (Thermo Fisher Scientific, Waltham, USA) directly in a final volume of 10 μl consisting of 0.65 μl 10 × R Buffer, 3.5 μl PCR product, 0.2 μl TaiI and 5.65 μl ddH 2 O. The mixture was incubated at 65°C for 15 h and 80°C for 20 min and subjected to gel electrophoresis on a 2% agarose gel.

Plant transformation of N. benthamiana
CaFFN gene overexpression vectors were generated by cloning into the XhoI + XbaI digestion sites of the vector pHellsgate 8. In this study, two different overexpression vectors were constructed: one included the wild-type CaFFN cDNA of A 145 (35S::CaFFN), and the other included mutant CaFFN cDNA with six mismatches to miR172 that did not change the amino acid sequence (synonymous mutations) (35S::CaFFNm3). The recombinant plasmid was introduced into N. benthamiana by Agrobacterium-mediated transformation as described previously 45 . The plants were screened in medium supplemented with 75 mg/L kanamycin (the negative control could not regenerate) in a plant growth chamber at 24°C under 16 h light/8 h dark cycles. The transgenic seedlings were identified by the primers FFN-RT-F and HR-gate8-RV (Supplementary Table S1).

VIGS in pepper A 145
For CaFFN silencing analysis, the specific coding regions of the CaFFN gene were used for VIGS vector construction. The sequence specificity was confirmed by genome-wide homologous sequence searching by BLAST against sequences in the Zunla databases (http:// peppersequence.genomics.cn/page/species/blast.jsp). Two fragments of~200 bp in different regions of the CaFFN gene were inserted into the pTRV2 vector at the EcoRI site to construct two vectors, TRV2::CaFFN-1 and TRV2::CaFFN-2. In addition, a fragment of~100 bp of the CaPDS gene was inserted into pTRV2 to construct a positive control vector, TRV2::CaPDS. The resulting vectors TRV2::rec and TRV1 were transformed into A. tumefaciens (strain GV3101). A. tumefaciens cells harboring TRV1 and TRV2::CaFFN-1/TRV2::CaFFN-2, TRV2 (as a negative control) or TRV2::CaPDS, as a positive control (resuspended in induction medium at a 1:1 ratio, OD 600 = 0.6), were coinfiltrated into seedlings of pepper A 145 at the four-leaf stage. The infiltrated seedlings were grown in a phytotron in the dark for 24 h and then at 20 ± 1°C under 16 h light/8 h dark cycles with a relative humidity of~90 ± 5%.

Quantitative reverse transcription PCR
To determine the relative transcription levels of selected genes, total RNA was extracted from plant tissue using RNAiso Plus reagent (Takara, Dalian, China) according to the manufacturers' instructions. Single-stranded cDNA was synthesized using TransScript ® II One-Step gDNA Removal and cDNA Synthesis SuperMix (TransGen Biotech., Beijing, China) for quantitative reverse transcription PCR (qRT-PCR) analysis of target genes or using Trans-Script ® miRNA First-Strand cDNA Synthesis SuperMix (TransGen Biotech., Beijing, China) for qRT-PCR analysis of target microRNAs. Quantitative real-time PCR was performed on a Bio-Rad CFX Connect Real-Time System (Bio-Rad Laboratories, Hercules, USA) using NovoStart ® SYBR qPCR SuperMix Plus (Novoprotein Scientific Inc., Shanghai, China). Gene expression was calculated relative to CaGAPDH or CaU6 in pepper or relative to NbbTUB in tobacco by the 2 −ΔΔCT method 46 . The amplification program was performed at 95°C for 2 min, followed by 95°C for 10 s and 60°C for 30 s (45 cycles). The primers used for qRT-PCR are listed in Supplementary Table S1. In each case, three technical replications were performed for each of at least three independent biological replicates.

Yeast one-hybrid (Y1H)
To investigate the interaction between the CaFFN transcription factor and CaAGL15, the full-length CaFFN open reading frame (ORF) sequence was transformed into the pGADT7 vector to provide the prey protein, and the cis-elements (~2000 bp upstream of the start codon isolated from pepper A 145 ) of the CaAGL15 gene were inserted into the pAbAi vector as bait sequences. The linearized bait plasmid pAbAi-CaAGL15 was transformed into the Y1H Gold yeast strain, and the minimal inhibitory concentration of AbA was determined for the bait strain with the empty pGADT7 vector. Then, the prey plasmid pGADT7-CaFFN was transformed into the bait strain to detect the DNA-protein interaction by incubating them at 30°C for 3 days on SD/-Leu medium supplemented with the minimal inhibitory concentration of AbA.

Dual-luciferase reporter assay
The cis-element of CaAGL15 (~2000 bp upstream of the start codon isolated from pepper A 145 ) was inserted into pGREEN0800-LUC 47 to generate the reporter plasmid pGreen0800-CaAGL15. The full-length ORF of CaFFN was inserted into the pGreenII 62-SK vector to generate the effector vector. The effector and reporter Agrobacterium cultures were mixed together and infiltrated into N. benthamiana leaves 48 . Empty pGreenII 62-SK (EV) was cotransformed with the CaAGL15 reporter as a no-interaction control. Firefly luciferase and Renilla luciferase were quantified at 3 days after infiltration using Dual Luciferase Reporter Assay System reagents (Promega, Fitchburg, United States). Three biological replicates were conducted for each combination.

Association analysis of CaFFN genes
A forward primer FFN-solcs-5UF1 and a reverse primer FFN-solcs-e4R1 (Supplementary Table S1) were used to amplify a fragment of the CaFFN gene in 164 inbred lines from core pepper collections. Each amplified fragment was sequenced using the FFN-solcs-5UF1 primer, and the sequences obtained were aligned using MUSCLE 49 . SNPs and InDels were identified, and variants were filtered according to the MAF (MAF ≥ 5%). The variants from 164 pepper inbred lines and FFN phenotypes were used in the association analysis, which was conducted using the cMLM in TASSEL v3.0 50 . The population structure (Q matrix) and familial kinship (K matrix) derived from our previous studies 51 were taken into account to avoid spurious associations. The associated variant(s) with the FFN trait were identified at the probability level of 0.01.

Analysis of AP2 family genes
The genome database of the cultivated pepper Zunla-1 36 was used for the identification of AP2 family genes. The hidden Markov model (HMM) profile of the AP2 domain (Pfam Accession No. PF00847) was used as a query to BLAST against the Zunla-1 protein database. AP2 family members were identified by checking gene structures in Pfam 52 to confirm the presence of two or more AP2 domains 31,53 . BLASTP alignment (E < 1e−5, top five matches) was carried out across the wholegenome protein sequences, and the output results were loaded into MCScanX (MATCH_SIZE: 5) 54 to identify collinear blocks and to distinguish the singleton, dispersed, proximal, tandem, and WGD/segmental duplication types of these genes. The coding sequences of the AP2 family genes and four additional genes (Capa-na01g000159, Capana10g001789, Capana03g000560, and Capana10g001776) were used to create a codon alignment by DAMEB6 55 . A phylogenetic tree of CaFFN/AP2 homologs was constructed using the maximum likelihood (ML) method in MEGA6 56 , and branch reliability was assessed with 1000 bootstrap replications. The Ks (synonymous substitutions per site) values were calculated in PAML 57 . The formula T = Ks/2 R was used to estimate the divergence time, where R is 1.5 × 10 −8 synonymous substitutions per site per year 58 . The expression data of 19 CaFFN/AP2 homologs were obtained from this study (primers are listed in Supplementary Table S1) and previous studies 20, 59,60 . The expression data were scaled to construct the heatmap by the Pheatmap package (https://CRAN.R-project.org/ package=pheatmap) in R.