The Unstable Restorer-of-fertility locus in pepper (Capsicum annuum. L) is delimited to a genomic region containing PPR genes

Unstable Restorer-of-fertility (Rfu), conferring unstable fertility restoration in the pepper CGMS system, was delimited to a genomic region near Rf and is syntenic to the PPR-like gene-rich region in tomato. The use of cytoplasmic-genic male sterility (CGMS) systems greatly increases the efficiency of hybrid seed production. Although marker development and candidate gene isolation have been performed for the Restorer-of-fertility (Rf) gene in pepper (Capsicum annuum L.), the broad use of CGMS systems has been hampered by the instability of fertility restoration among pepper accessions, especially sweet peppers, due to the widespread presence of the Unstable Restorer-of-fertility (Rfu) locus. Therefore, to investigate the genetic factors controlling unstable fertility restoration in sweet peppers, we developed a segregation population (BC4F5) from crosses using a male-sterile line and an Rfu-containing line. Segregation did not significantly deviate from a 3:1 ratio for unstable fertility restoration to sterility, indicating single dominant locus control for unstable fertility restoration in this population. Genetic mapping delimited the Rfu locus to a 398 kb genomic region on chromosome 6, which is close to but different from the previously identified Rf-containing region. The Rfu-containing region harbors a pentatricopeptide repeat (PPR) gene, along with 10 other candidate genes. In addition, this region is syntenic to the genomic region containing the largest number of Rf-like PPR genes in tomato. Therefore, the dynamic evolution of PPR genes might be responsible for both the restoration and instability of fertility in pepper. During genetic mapping, we developed various molecular markers, including one that co-segregated with Rfu. These markers showed higher accuracy for genotyping than previously developed markers, pointing to their possible use in marker-assisted breeding of sweet peppers.


Introduction
Cytoplasmic male sterility (CMS) is a maternally inherited trait characterized by the inability to produce functional pollen (Chen et al. 2017). The CMS phenotype is caused by the presence of aberrant mitochondrial genes in various plant species (Hanson 1991). In the cytoplasmic-genic male sterility (CGMS) system, the Restorer-of-fertility (Rf) gene from the nuclear genome suppresses the CMS phenotype. Rf genes often alter the expression of proteins from CMS-associated genes and are therefore thought to reduce or remove the deleterious effects caused by CMS-associated genes. The CGMS system eliminates the need for the laborious emasculation process and thus can be used for hybrid seed production in diverse, agriculturally important crops.
PPR proteins constitute one of the largest protein families in land plants, with more than 400 members in most species (Barkan and Small 2014). PPR proteins contain a number of PPR motifs, including a degenerate 35 -amino acid motif repeated in tandem (Manna 2015). The combination of the 5th and 35th amino acids from the respective PPR motifs can determine the specificity of recognition of one nucleotide of the target RNA molecule (Shen et al. 2016). PPR proteins function in various processes in plants, including RNA editing, splicing, cleavage, and degradation. In CMS-BT rice, the proteins encoded by PPR type Rf genes Rf1a and Rf1b bind to and cleave dicistronic atp6-orf79 transcripts (Wang et al. 2006). Unlike other PPR genes, Rf-like PPR (RFL) genes are located close to each other in a cluster. The Rf4 locus in rice is located in a typical PPR cluster region that also includes Rf1a and Rf1b (Tang et al. 2014). The PPR family in Capsicum has been identified, along with the cluster of RFLs on the distal end of chromosome 6 ( Barchenger et al. 2018).
The stability of Rf genes in various environments is critical for the use of CGMS systems in producing F 1 hybrid seeds. However, various reports indicate that the stability of Rf genes is strongly affected by the external environment. In CMS-S maize, Rf9 shows consistent restorer ability at temperatures below 28 ℃. However, when the daily maximum temperature is higher than 30 ℃, the male fertility phenotype changes to male sterility (Gabay-Laughnan et al. 2009). In rice CMS-BT, the Rf5 or Rf6 gene can restore normal seed setting, but lines heterozygous for Rf6 show more stable seed setting rates than lines heterozygous for Rf5 under heat-stress conditions (Zhang et al. 2017).
Although numerous past efforts to identify the Rf gene in pepper have led to the development of several molecular markers that co-segregate with Rf and the isolation of Rf candidate genes in the co-segregation regions, the identity of the Rf gene has not yet been confirmed through functional complementation experiments. Jo et al. (2016) delimited the genomic region of Rf, and identified potential candidate genes within it (CDS1, CDS2, CDS3, and CaPPR6), and chose CaPPR6 as the candidate pepper Rf gene. Similarly, Cheng et al. (2020) and Zhang et al. (2020) performed fine mapping for pepper Rf and identified several PPR genes in this region that co-segregated with Rf. All co-segregation regions defined in these three studies were located in a narrow area of chromosome 6, although each study identified different PPR genes as the Rf candidate gene. In contrast, Wei et al. (2020) chose a gene encoding a NEDD8-conjugating enzyme E2 as the candidate Rf gene based on bulk segregant RNA sequencing. Further functional confirmation based on transformation or genome editing is required to confirm the identity of Rf genes in pepper.
In addition to complete sterility or fertility, pepper plants can exhibit unstable fertility at certain temperatures, as described in the first report of CMS and fertility restoration in peppers in 1958 (Peterson 1958). Kim et al. (2013) reported that a pepper line with thermo-sensitive cytoplasmic male sterility (TCMS) was sterile at temperatures above 15 °C but became fertile when the night temperature dropped below 13 °C. Lee (2001) suggested that besides the major fertility restorer gene, there is an unstable dominant Rf (St u ) locus that affects male fertility. Lee et al. (2008a) observed the partial restoration phenotype, which refers to the generation of only a small quantity of pollen, in CMS cytoplasm-containing individuals and determined that the nuclear gene partial restoration (pr) is tightly linked to Rf or is the third allele of Rf based on genetic mapping. On the other hand, Min et al. (2009) demonstrated that the locus of the gene responsible for unstable male sterility is not linked to the Rf locus. Ortega et al. (2020) suggested the presence of a minor gene that involves in fertility restoration interacting with the environment as well as clustering of RFL genes that may confer genotype-specific restoration as possible reasons for imperfect prediction ability of Rf-linked markers in the application of them to diverse pepper accessions. The partial or unstable restoration phenotype is widespread among pepper accessions, especially among sweet peppers. The lack of stable restorer lines in sweet pepper has been an obstacle to using the CGMS system in this important crop. To overcome this obstacle, the stable Rf gene from hot pepper was successfully introduced into sweet pepper through marker-assisted backcrossing (Lin et al. 2015).
In the current study, we investigated the unstable restoration trait in sweet pepper and identified the genetic and genomic positions of the instability gene using previously and newly developed markers. Based on this information, we chose candidate genes underlying the unstable restoration trait. Our findings should facilitate the use of the CGMS system for the improvement of sweet pepper.

Plant materials
The hot pepper lines C. annuum 'Chungyang A' and 'Chungyang C' and the sweet pepper lines C. annuum 'MSGR-A' and 'SPR03' were used for inheritance analysis and to develop molecular markers linked to the unstable restoration trait (Table 1). C. annuum 'MSGR-A', with small blocky fruit and stable male sterility, was used as the stable rf allele donor parent, and 'SPR03', with blocky fruit and Unstable Restorer-of-fertility (hereafter designated as Rf u ), was used as the recurrent parent (Fig. 1). A BC 4 F 5 population was developed by crossing MSGR-A and SPR03 and self-pollinating the progenies (Fig. 2). Heterozygotes were selected using two Rf-linked markers: PR-CAPS, which is linked to the partial restoration locus, and PPR6_3'UTR-SCAR, which is tightly linked to the Rf locus. The BC 4 F 5 individuals were grown in the greenhouse at the Seoul National University farm (Suwon, Korea). Leaf samples of 12 stable male sterile lines and 11 unstable male sterile lines, which were provided by Eco-Seed (Gimje, South Korea), were used to test the newly developed markers.

Evaluation of the male fertility phenotype
The male fertility phenotype was investigated four to six times per individual by observing pollen production on anthers in early summer (June) and autumn (October). In these seasons, the temperature inside the greenhouse was maintained between 20 and 30 °C during day and between 15 and 25 °C at night. The presence of pollen was determined by observing anthers with the naked eye and was confirmed by dusting the pollen from anthers onto black paper. If pollen was detected at least once in an individual during the four or more observations, the individual was considered to be partially fertile. To further confirm the presence of viable pollen in the plants, pollen was collected from dehiscent flowers and stained with fluorescein diacetate (FDA, Sigma-Aldrich, USA). The stained pollen was observed under an optical microscope (Carl Zeiss, Germany). After 2 min incubation at room temperature, pollen viability was examined under a fluorescence microscope. Pollen grains were harvested from at least five flowers and stained with FDA. This process was repeated three times.

Genomic DNA extraction
Young leaves were collected from each plant and used for genomic DNA (gDNA) extraction using the cetyltrimethylammonium bromide (CTAB) method (Park et al. 2009). The gDNA was dissolved in 1 × TE buffer and diluted in triple-distilled water to a final concentration of 50 ng/µl for molecular marker analysis.

Analysis of mitotypes
The mitotypes of the lines were analyzed using a multiplex-PCR marker developed by Jo et al. (2019). This marker can be used to determine the genotypes of four mitochondrial loci in a single PCR; five mitotypes of C. annuum were identified when this marker was applied  to diverse germplasm. The PCR was performed using four pairs of primers (5 pmol each) described by Jo et al. (2019) in a total volume of 50 μl containing 100 ng DNA, 200 μM dNTP mix, 5 μl 10 × buffer, and 1.5 units Taq DNA polymerase (Takara Bio, Inc., Kusatsu, Japan). The PCR conditions were 32 cycles of 98 °C (10 s), 59 °C (30 s), and 72 °C (100 s), followed by a final extension at 98 °C (5 min).

Molecular marker analysis
For molecular marker analysis, PCR was conducted in 20 µl reaction mixture containing 100 ng of gDNA template, 2 µl of 10 mM dNTPs, 2 µl of 10 × PCR buffer (Takara, Seoul, Korea), 0.5 µl of 10 pmol primers, 0.3 µl of Takara R Taq™ polymerase (Takara, Seoul, Korea), and triple-distilled water (to 20 µl). The cycling conditions were an initial cycle of 94 °C for 5 min; 35 cycles of 94 °C for 30 s, the annealing temperature of each marker (Table 2) for 30 s, and 72 °C for 20 s; and then a final extension of 72 °C for 5 min. For Cleaved Amplified Polymorphic Sequence (CAPS) markers, the PCR product was digested with 5 units of restriction enzymes ( Table 2). The digested PCR products were separated on 2% agarose gels.
For SNP markers, direct Sanger sequencing of amplicons was performed.

Development of Rf-linked markers
To identify the polymorphism between MSGR-A and SPR03, direct sequencing was carried out (Macrogen, Seoul, Korea). Basic local alignment search tool (BLAST) was carried out using the pepper reference genome 'UCD10X (v1.0)', which was obtained from an F 1 line of a cross between CM334 and a blocky nonpungent pepper (Hulse-Kemp et al. 2018) to identify the physical locations of CAPS markers. Additionally, we used markers that were previously shown to be polymorphic between Chungyang A and Chungyang C (Jo et al. 2016) or was developed to map the partial restoration locus (Lee et al. 2008a, b). Information about these markers was retrieved from the literature: Co1Mod1-CAPS, 4162-SCAR, and G16-CAPS (Jo et al. 2016), PR-CAPS (Lee et al. 2008a, b), and 13T7-SCAR (Jo et al. 2010).

Linkage analysis
Linkage analysis was performed using molecular markers and 300 BC 4 F 5 individuals from an introgression between MSGR-A and SPR03. The linkage map was constructed using CarthaGene software with a LOD threshold of 3.0  Lee et al. (2008a, b) and distance threshold of 30 cM (De Givry et al. 2005). Genetic distance in the linkage map was determined using the Kosambi mapping function. A linkage map was created based on the genetic distance between the markers using Map Chart 2.3 software.

Gene annotation
Annotation of the target region was performed by BLAST analysis of the DNA sequence in this region [from UCD10X (v1.0) reference genome] against coding DNA sequences (CDSs) previously annotated to the CM334 v1.55 genome (Kim et al. 2014). If the similarity was higher than 98% in the BLAST results, the sequence block in the target region was annotated based on the matched CDS.

Screening of Rf-like PPR genes in tomato
The deduced protein sequences of the petunia Rf gene (Rf-PPR592) and a candidate pepper Rf gene (CaPPR6) were used as query sequences for BLAST against tomato (Solanum lycopersicum) protein sequences (ITAG release 2.40; http:// solge nomics. net/ tools/ blast/). The e-value threshold was set to 10 -10 . Proteins with sequences showing identity to Rf-PPR592 > 50% and longer than 100 amino acids were selected as Rf-like PPRs. The genomic locations of the genes encoding these proteins were determined based on the tomato genome ITAG release 2.40.

Inheritance study in an Rf u -segregating population
We obtained a BC 4 F 5 population during the course of introgression of the stable rf allele from a small blocky sweet pepper line (MSGR-A) to a large blocky sweet paper line (SPR03) that was assumed to contain Unstable restorerof-fertility (Rf u ) (Fig. 1). To reveal the inheritance pattern of the unstable male sterility phenotype, we evaluated the male sterility phenotype in a segregating population of 300 BC 4 F 5 individuals. In the BC 4 F 5 population, 228 individuals were partially fertile and 72 individuals were completely sterile. The anthers of partially-fertile individuals were larger and lighter in color than those of sterile individuals (Fig. 2). Microscopic observation of FDA-stained pollen showed that partially-fertile individuals produced a small number of viable pollen grains (Fig. S1). Partially fertile individuals were also clearly distinguished from the fully fertile recurrent parent line (SPR03) in that they produced much less pollen, but did not show observable difference from SPR03 in fruit number and size. The amount of pollen produced by partially fertile individuals was not consistent during the four or more observations and appeared to be affected by environmental conditions. Thus, we described this phenotype as unstable fertility restoration rather than partial restoration. The segregation ratio between individuals showing unstable fertility restoration and complete sterility phenotypes in the BC 4 F 5 population did not significantly deviate from an expected ratio of 3:1 (χ 2 = 0.160; P = 0.689), suggesting that the instability is controlled by a single dominant locus in this population (Table 3).

Analysis of mitotypes
Since the subgenomic structure of the mitochondrial genome may affect the stability or fertility of plants (Janska et al., 1998), we investigated the mitotypes of pepper lines with unstable fertility using a multiplex-PCR marker that can classify five mitotypes in C. annuum. The amplification pattern in lines with unstable fertility could not be distinguished from that of lines with stable sterility (mitotype 4; Fig. 3). Therefore, the unstable lines contained mtDNA structures that are characteristic of typical CMS lines, implying that mtDNA rearrangements are not involve in the unstable fertility of these lines. Among parental lines used for population development, MSGR-A (with stable rf) carried mitotype 4, as did other sterile lines. The parental line SPR03 (with Rf u ) carried mitotype 3, which is predominantly detected in domesticated pepper lines from Asia. Because BC 4 F 5 individuals contain cytoplasm from MSGR-A, unstable fertility is determined by the presence or absence of nuclear Rf u . In addition, Rf u induces instability only in the CMS cytoplasm background, as SPR03, which contains Rf u in the non-CMS cytoplasm background, was fully fertile (Fig. 2, Fig. 3).

Polymorphism survey of previously developed markers
To test previously developed markers in the segregating population used in this study, we first surveyed the polymorphism of markers between hot and sweet pepper parental lines, including 4162-SCAR, 120 kb-end-SCAR, PPR12-SCAR, 3336-last-SCAR, Co1mod1-CAPS, and 4940-CAPS. These markers showed polymorphisms in hot pepper lines Chungyang C (RfRf) and Chungyang A (rfrf), as described previously (Jo et al. 2016), but they did not distinguish the  (Fig. S2a). Therefore, we developed the new molecular markers 4162-CAPS and G16-SCAR based on the previously detected SNPs in 4162-SCAR and G16-CAPS, respectively ( Fig. S2b; Table 2). PPR12-SCAR and 4940-CAPS showed polymorphisms between SPR03 and MSGR-A, but inconsistent PCR products were produced. We therefore surveyed polymorphisms in 13T7-SCAR, G16-SCAR, PR-CAPS, PPR6_3'UTR-SCAR, and 4162-CAPS between the sweet pepper lines. When 31 BC 4 F 4 individuals were tested, 4162-CAPS showed two recombinants, G16-SCAR and PR-CAPS showed one recombinant, and 13T7-SCAR co-segregated with the male sterility phenotype (Fig. 4). These results indicate that the Rf u locus is not located between 4162-CAPS and G16-SCAR and is located close to 13T7-SCAR. In a previous study, the Rf gene was mapped between 4162-SCAR and G16-CAPS (Jo et al. 2016). Therefore, the location of the Rf u locus revealed by genetic mapping in the current study is different from that of the Rf locus.

Development of additional markers linked to unstable male sterility
Since 13T7-SCAR was located closer to Rf u compared to the other markers examined, we performed further mapping by developing additional markers around 13T7-SCAR. We developed four CAPS markers within the 1.  S2b). Additionally, we developed three more markers (G30-SNP, G40-SNP, G110-SNP) to further narrow down the Rf u -containing region delimited by initial mapping (Fig. S2b).  Fig. 4 Genotypes of recombinants found among BC 4 F 4 individuals determined using Rf u -like markers

Genetic mapping and genomic analysis of the Rf u locus
We constructed a local linkage map using the BC 4 F 5 population and seven molecular markers developed in this study (Fig. 5a). The PPR6_3'UTR-SCAR and G16-SCAR markers were linked to that Rf u locus at distances of 2.4 and 0.9 cM, respectively. Jo et al. (2010) reported that the 13T7-SCAR marker was linked to the Rf locus at a distance of 1.4 cM; however, the genetic distance of 13T7-SCAR from the unstable male sterility locus was 0.4 cM. The markers 213.3 MB-CAPS, 212.9 MB-CAPS, and 213.85 MB-CAPS were linked to the unstable male sterility locus at a genetic distance of 2.2, 4.3, and 8.9 cM, respectively. When we tested the 214.14 MB-CAPS marker in this population, we did not observe any recombinants, demonstrating that the 214.14 MB-CAPS marker is tightly linked to the Rf u locus. G16-SCAR and 13T7-SCAR were closest to 214.14 MB-CAPS on two different sides; thus, the Rf u locus was delimited by these two flanking markers. Integrative analysis by combining our results with genetic map information obtained from a previous study (Jo et al., 2016) clearly showed that the Rf u locus is separated from the Rf locus. The genetic distance between the Rf u and Rf loci was determined to be 1.2 cM based on a genetic map developed using the Chungyang F 2 population (Jo et al., 2016;Fig. 5a). These results indicate that Rf u is not an allele of Rf but is a separate gene that is genetically linked to Rf.

Annotation and characterization of genes in the Rf u -containing DNA region
The DNA region between two markers (G16-SCAR and 13T7-SCAR) flanking the Rf u locus corresponded to the 4,575-5,054 kb region of chromosome 6 in the UCD10X (v1.0) pepper reference genome (Fig. 5b). Further mapping   using markers developed from this region reduced the Rfucontaining region to be the 4,656-5,054 kb region. Therefore, we were able to delimit the Rf u -containing DNA region within a 398 kb region. A total of 11 genes were annotated in this region based on CDSs previously annotated in the CM334 reference genome (v1.55; Kim et al. 2014). These genes encode three NBS-LRR resistance proteins, two adenylate isopentenyltransferases, an ER lumen retaining receptor family-like protein, a Gypsy/Ty-3 retroelement polyprotein, a PPR protein, a RING finger protein, a glycosyl group transferase, and a protein of unknown function ( Table 4). The PPR gene (CA00g30080) was positioned 25 kb away from the 214.14 MB-CAPS marker. This gene encodes a 248 amino acid protein. We searched for the gene sequence that matched CA00g30080 among the CDSs of the blocky-type nonpungent pepper accession 'Early Calwonder' (ECW), for which de novo genome assembly information is available (Kim et al. 2021). The sequence of the protein encoded by the best-matched CDS of ECW (ECW. scaffold897.110) shows 97% identity and an insertion of one amino acid compared to the sequence of the protein encoded by CA00g30080 (Fig. S3). We aligned the protein sequences of CA00g30080 and ECW.scaffold897.110 on an array of the PPR motifs in Rf-PPR592, the petunia Rf, and CaPPR6 (a candidate Rf of pepper). Inside this alignment, three gaps were detected in the protein sequences of CA00g30080 and ECW.scaffold897.110. Notably, two of these gaps were 35 amino acids in size, which is identical to the length of a PPR motif (Fig. S3).

Characterization of a tomato genomic region syntenic to the pepper Rf u -containing region
We previously developed the G05G1-HRM marker from an expressed sequence tag (EST) in pepper with high similarity to a tomato gene near a PPR gene, which was mapped to the same position as 13T7-SCAR (Jo et al. 2010). The G50G1-HRM sequence is located inside the Rf u -containing DNA region identified in the present study and only 4.6 kb away from the 214.14 MB-CAPS marker that co-segregated with Rf u . We developed four markers (including G05G1-HRM) by comparative mapping using the tomato genome sequence, which could be used to link the DNA region surrounding Rf and Rf u to the corresponding genomic region of tomato. The corresponding tomato DNA region locates on 1.7-1.9 Mb region on chromosome 6 (Fig. 5a). More detailed analysis using the annotated genes showed that five genes in the Rf u -containing DNA region between G16-CAPS and 13T7-SCAR were orthologous to tomato genes dispersed along 1.74-1.79 Mb region on chromosome 6 (Fig. 5b). Interestingly, three PPR genes were detected in this region, and two more were identified in a surrounding region. We examined the Rf-like PPR (RFL) genes in tomato to determine whether they are specifically clustered in this region. Among the 249 and 248 tomato PPR proteins (> 100 amino acids long) that showed similarity to petunia Rf-PPR592 and pepper CaPPR6, respectively, only nine shared > 50% identity with Rf-PPR592 or CaPPR6 (Table 5). The identity detected for these nine PPR proteins was clearly higher than that of other PPR proteins (~ 20-30%; Fig. 6). These results indicate that the nine tomato PPR proteins are RFL proteins. Five of these protein are encoded by five genes inside the 1.72-1.77 Mb (46 kb in size) region on chromosome 6. However, the four other genes are scattered on different regions of chromosome 6 or on different chromosomes. Therefore, we conclude that the tomato DNA region syntenic to the pepper Rf u locus is the region with the highest density of RFL genes in the tomato genome.

Application of Rf-linked markers to stable and unstable breeding lines
We used six markers, including two newly developed markers (212.9 MB-CAPS, 214.14 MB-CAPS), two markers modified from previously developed markers (G16-SCAR, 4162-CAPS), and two previously developed markers (13T7-SCAR, Co1Mod1-CAPS), to genotype breeding lines containing CMS cytoplasm (Fig. 3) that showed stable sterility or unstable fertility based on the presence of a nuclear unstable restorer (Table 6). Five markers showed polymorphism between SPR03 and MSGR-A. However, the Co1Mod1-CAPS marker that co-segregated with Rf in a previous study (Jo et al. 2016) did not show polymorphism in any of the sweet pepper lines examined. The 212.9 MB-CAPS and 4162-CAPS markers, which are more distant from Rf u than the other markers, failed to correctly genotype most pepper lines harboring Rf u . However, 214.14 MB-CAPS, which co-segregated with Rf u , and G16-SCAR, which is located on a boarder of an Rf u -containing DNA region, showed high accuracy in predicting the genotypes of pepper lines carrying Rf u (10 out of 11 lines for 214.14 MB-CAPS, 9 out of 11 lines for G16-SCAR). The rate of successful genotyping in all lines examined was the highest using G16-SCAR (82.7%; Table 6). Finally, we generated a phylogenetic tree based on the marker genotypes from all breeding lines. Most lines showing stable sterility were grouped together in the tree, as were most lines showing unstable fertility (Fig. 7). These results indicate that several markers can be used for phylogenetic analysis of breeding lines based on Rf or Rf u genotypes.

Localization of the Rf u locus in the pepper genome
The CGMS system is useful for producing F 1 hybrid seeds in pepper and other crops. Even though sweet pepper is an economically important crop, the CGMS system is not being used efficiently in this species. The presence of an unstable restorer that partially restores fertility or acts inconsistently depending on environmental conditions poses an obstacle to the use of the CGMS system in sweet peppers. Although unstable fertility or sterility in peppers has been repeatedly reported since Peterson (1958) first described this trait, there have been few genetic or genomic analyses of unstable fertility restoration, presumably due to the difficulty of phenotyping this trait, even though the pepper reference genome has been available since 2014 (Kim et al. 2014). Lee et al. (2008aLee et al. ( , 2008b reported the first genetic mapping of the partial restoration (pr) locus. Although the authors identified markers for pr close to Rf, they could not determine whether pr was an allele of Rf or another gene that is tightly linked to Rf. On the other hand, Wang et al. (2004) performed quantitative trait locus (QTL) analysis of fertility restoration in   1 3 pepper and identified one major QTL on chromosome 6 and additional minor QTLs located on other chromosomes. The unstable Rf was considered to be related to the minor QTLs. Min et al. (2009) determined, based on segregation analysis in a large F 2 population, that the locus for unstable sterility was independent from the Rf locus. However, none of these studies identified the genomic location of the gene causing instability.
Here, we delimited the location of Rf u in the pepper reference genome. Based on this analysis, we demonstrated that the Rf u gene is located close to Rf. The Rf u gene identified in the current study might be identical to pr based on a comparison of the phenotypes caused by Rf u and pr and the genetic locations of these two genes (Lee et al. 2008a;2008b). Lee et al. (2008b) developed a segregation population by crossing lines containing pr and CMS cytoplasm with a line containing Rf. The authors obtained partially restored lines only when pr was homozygous in the CMS cytoplasm background, demonstrating that pr is a recessive gene. By contrast, in the current study, we used a line containing rf as one of the parental lines and determined that Rf u acts as a dominant gene over the rf allele. If we assume that pr and Rf u are the same gene, the results of these two studies indicate that Rf is epistatic to Rf u (or pr). Except for partial restoration, temperature-sensitive instability of fertility have been reported Peterson 1958;Shifriss 1997). For example, Kim et al. (2013) reported that the fertility of a TCMS line can be restored to almost normal levels below a critical temperature (13 °C). Although minor fluctuations in pollen levels, presumably depending on environmental conditions, were detected in the current study, the observed phenotype was different from that in z The genotypes of markers that were mapped in this study are represented as Rf u or rf u , while the genotypes of Co1Mod1-CAPS, which co-segregated with Rf (Jo et al. 2016), are designated as Rf or rf  Kim et al. (2013) since the amount of pollen did not reach normal levels. Therefore, it appears that additional genes affect the stability of fertility restoration in pepper besides Rf u characterized in the current study.

Implications for the evolution of PPR genes in the Rf and Rf u loci in pepper
The Rf u locus has been delimited to a DNA region containing a PPR gene on chromosome 6. All three fine-mappingbased approaches to isolate pepper Rf genes localized the Rf loci to PPR-gene-containing DNA regions on chromosome 6 (Cheng et al. 2020;Jo et al. 2016;Zhang et al. 2020). The Rf u locus and three Rf loci are located near each other on chromosome 6, and thus this region has been hypothesized to be the genomic region where RFL genes are clustered together. Among PPR genes, which number more than 400 in most plant species, the RFL genes, which show high similarity to Rf and share a very close evolutionary relationship, are clustered in narrow genomic regions . For example, 19 RFL genes are clustered within an ~ 1 Mb region on chromosome 1 in Arabidopsis (Lurin et al. 2004). In pepper, Barchenger et al (2018) showed that 11 RFL genes clustered on the distil end of chromosome 6. Jo et al. (2010) screened a pepper BAC library using an EST with high similarity to the petunia Rf gene as a probe. The markers developed from three major groups of the 74 selected BAC clones, including 13T7 SCAR, which is located 0.4 cM away from Rf u (identified in the current study), were shown to be linked to Rf. This finding confirms that RFL genes are densely located on the DNA region surrounding the Rf and Rf u loci. Among Rf u and the three Rf loci, only the Rf u locus is syntenic to the tomato DNA region where the RFL genes are clustered. This finding suggests that the pepper Rf u locus and the tomato genomic region containing the RFL gene cluster might have evolved from a common ancestral sequence, whereas the Rf loci might have evolved later in pepper independently. The RFL genes evolved under diversifying selection, presumably to help plants cope with the evolution of new CMS genes ). Since CMS cytoplasm has not been found in tomato, the co-evolution of Rf loci following the emergence of the pepper CMS gene might had led to the generation of the pepper-specific Rf locus. The sequences in the Rf u locus might have been used as ancestral sequences during this process, considering that PPR genes likely spread by retroposition from the original gene (O'Toole et al. 2008).
The PPR protein encoded by the PPR gene in the Rf ucontaining region is much shorter (248 amino acids) than the petunia Rf protein Rf-PPR592 (592 amino acids) and the pepper Rf-candidate CaPPR6 (589 amino acids). We suggest three possibilities for the relationship between this short PPR protein and unstable fertility restoration. First, the incomplete structure might provide this protein with partially impaired function, leading to incomplete restoration. The evolution of Rf with higher functionality might have led to the accumulation of mutations in the Rf u -candidate PPR gene, thus decreasing the functionality of the protein product. Second, there might be another functional PPR gene in an unassembled sequence in the reference genome used in this study. The sequence around CaPPR6 is highly complex and thus was not successfully assembled in the main chromosome of the CM334 reference genome. Jo et al. (2016) obtained the target region only by chromosome walking. The dynamic nature of the DNA region containing RFL genes might have hampered its precise assembly in the reference genome. Finally, there might be substantial differences in the structure of the Rf u -containing region or sequence of the Rf u -candidate PPR gene between the reference genome and the genome of SPR03, the Rf u donor. The short PPR gene in the reference genome could be a nonfunctional recessive allele of Rf u rather than the dominant allele, as is present in the genome of SPR03. Although the PPR gene is a strong candidate gene of Rf u considering its putative function, we cannot exclude the possibility that Rf u is one of the other ten genes in the delimited Rf u -containing region. Further studies are needed to reduce the number of candidate genes by fine mapping with a larger population and to characterize candidate genes functionally.

Feasibility of marker-assisted breeding using developed markers
We used previously and newly developed markers to examine breeding lines with or without unstable restorers to evaluate the feasibility of marker-assisted breeding using these markers. Although three markers developed inside or on the flanking sequences of the defined Rf u -containing DNA region showed relatively high accuracy for genotyping of unstable restorer-containing lines, there were still substantial differences in their accuracy. For example, G16-SCAR, which is located 0.9 cM from Rf u , showed ~ 20% higher accuracy than 214.14 MB-CAPS, which co-segregated with Rf u . This may reflect the existence of unstable restorers other than Rf u (as discussed above) and/or subdivided haplotypes of markers. The markers developed from the DNA regions surrounding the Rf locus often contained haplotypes that were not classified as Rf or rf, presumably due to the structural dynamics in these regions (Min et al. 2009;Lee et al. 2008a;Jo et al. 2010). Therefore, perhaps additional polymorphisms that could not be detected by cleavage with a restriction enzyme or based on amplicon length are also present in the Rf u -linked marker sequence. The haplotypes of the markers might be further classified based on additional polymorphisms.
To breed stable maintainer lines, lines with rfrf without Rf u should be selected. In normal cytoplasm background, the fertility of the lines with rfrf/RfuRfu genotype, which should not be used as maintainer lines, is not phenotypically different from that of those with rfrf/rf u rf u genotype, which can be used as stable maintainer. This is because unstable restoration is only detectable in CMS cytoplasm background. Since lines with latent unstable restorer are widely spread among sweet pepper lines, marker-assisted selection of lines with rfrf/rf u rf u genotype has significance in the development of sweet pepper maintainer lines. Therefore, it is important to be able to discriminate genotypes of unstable restorer from those of stable restorer using markers. G16-SCAR showed high accuracy for determining the Rf u genotype in the current study. On the other hand, Co1Mod1-CAPS co-segregated with Rf and showed the highest accuracy for predicting the Rf genotype among Rf-linked markers in a previous study (Jo et al. 2016). In this study, all breeding lines containing Rf u were genotyped as rfrf using Co1Mod1-CAPS. Since the Rf u -containing breeding lines in our panel showed unstable fertility, and Rf is likely epistatic to Rf u , as discussed above, it is highly likely that this genotyping result is correct and that Co1Mod1-CAPS is useful for discriminating the Rf genotype from the Rf u genotype. Therefore, the combined use of G16-SCAR and Co1Mod1-CAPS is likely to provide high accuracy in marker-assisted breeding of stable maintainer or restore lines.