Introduction

Citrus is cultivated throughout the world, from tropical to temperate climates where there are suitable soils, sufficient moisture, and optimal temperatures (Soost and Roose, 1996). The importance of the citrus industry to agriculture and the economy is demonstrated by this large-scale production. Citrus is known to have gametophytic self-incompatibility (SI) governed by a single gene (S) in pummelo [Citrus maxima (Burm.) Merr.], mandarin (Citrus spp.), and their hybrids (Soost, 1987; Yamamoto et al., 2006; Kim et al., 2011; Ngo et al., 2019). SI is an important trait for breeding and seedless fruit production in Citrus cultivars with a high degree of parthenocarpy (Iwamasa and Oba, 1980; Vardi et al., 2000). For Citrus, however, there is very limited information about the self-incompatibility S gene sequence (Honsho et al., 2019; Liang et al., 2020), S allele variation, S allele frequencies, and the S genotypes of cultivars (Kim et al., 2020).

Ngo et al. (2019) studied the patterns and strength of self-incompatibility in 121 Citrus accessions, including 77 self-incompatible pummelos. They found that the degree of SI in the stigmas, upper styles, middle styles, and basal styles of the self-incompatible pistils differed and could be divided into three categories: high, moderate, and low. Eventually, there were no or only a few pollen tubes in the base of the self-pollinated styles of these self-incompatible accessions. These results suggested that self-incompatibility in a given Citrus accession could be determined only by observation of the pollen tubes in the basal style (Ngo et al., 2019).

To elucidate the self-incompatibility genotypes of Citrus cultivars, Ngo et al. (2011) used Got-3 allozyme markers associated with the self-incompatibility S gene and proposed the S₁S₂ genotype for ‘Banpeiyu’ (Citrus maxima), S₁S₃ for ‘Tosa Buntan’ (Citrus maxima), S₄S₅ for ‘Hassaku’ (Citrus hassaku), S₆S₇ for ‘Yuge Hyokan’ (Citrus yuge-hyokan), and S₁S₈ for ‘Hyuganatsu’ (Citrus tamurana). For the S₁ and S₂ alleles proposed by Ngo et al. (2011), Kim et al. (2011) determined the S genotypes in seven full- or semi-self-incompatible Citrus accessions by pollination with two homozygous S₁ seedlings (S₁S₁ and S₂S₂), each derived from self-pollination of the self-incompatible ‘Banpeiyu’ (S₁S₂) pummelo. For the S₄ and S₅ alleles proposed by Ngo et al. (2011), Zhou et al. (2018) determined 12 full- and semi-self-incompatibility S genotypes and one full-self-compatibility S genotype (S_fS_f) in Citrus accessions by pollination with two homozygous S₁ seedlings (S₄S₄ and S₅S₅), each from self-pollination of the self-incompatible ‘Hassaku’ (S₄S₅). For the S₃ allele proposed by Ngo et al. (2011) and the S₁₁ allele discovered by Kim et al. (2020), Kim et al. (2020) determined the S genotypes in 28 full- or semi-self-incompatible Citrus accessions by pollination with two homozygous S₁ seedlings (S₃S₃ and S₁₁S₁₁), each from self-pollination of the self-incompatible Clementine (S₃S₁₁). In these S-genotyping studies with more than 100 full- and semi-self-incompatible Citrus accessions, the S allele frequency was calculated for S₁ and S₂ (Kim et al., 2011), S₃ (Kim et al., 2020), S₄and S₅ (Zhou et al., 2018), and S₁₁ (Kim et al., 2020) alleles.

In the previous study for S₉ and S₁₀alleles (Kim et al., 2010), homozygous S₁ seedlings (S_aS_a and S_bS_b) of ‘Hirado Buntan’ (S_aS_b) were produced through bud self-pollination and allelic symbols of S₉ and S₁₀ were determined, respectively, for S_a and S_b, and then using the S₉ and S₁₀alleles, pollen detection of Citrus accessions with S₉ and S₁₀ alleles was carried out focusing on a small number of pummelos growing on Hirado Island adjacent to the northernmost part of Kyushu Island, Japan. A total of only 18 Citrus accessions was used for the determination and detection of S₉and S₁₀alleles with the homozygous S₁ seedlings (S₉S₉ and S₁₀S₁₀) of ‘Hirado Buntan’ in the previous study. Because of the small number of accessions, we could not determine the origin of the S₉ and S₁₀ alleles, the rates of Citrus accessions with S₉ or S₁₀ alleles, and S₉ and S₁₀ allele frequencies in pummelo (a source of S alleles) and Citrus accessions (introgression of S alleles from pummelo). The present study was carried out to clarify the origin of the S₉ and S₁₀ alleles and to calculate the rates of Citrus accessions and pummelos with S₉ or S₁₀ alleles and the S₉ and S₁₀ allele frequencies in pummelo and Citrus accessions through S-genotyping of 146 Citrus accessions (not including artificial hybrid cultivars) with S₉ or S₁₀ alleles. When comparing the results of S-genotyping between pummelo (source of S alleles) and Citrus accessions (introgression of S alleles from pummelo), the genetic diversity in pummelo and its relation to the development and evolution of full-self-incompatible and semi-self-incompatible Citrus species and cultivars determined using S alleles as markers. Here, we aimed to uncover the genetic diversity in pummelo and the development and evolution of Citrus species and cultivars by monitoring their S alleles.

Materials and Methods

Plant Materials

We collected 146 Citrus accessions, including 82 pummelo accessions, to use for pollination with two homozygous S₁ seedlings (S₉S₉ and S₁₀S₁₀) of ‘Hirado Buntan’ [Citrus maxima (Burm.) Merr.]. Most Citrus cultivars included in these accessions were provided by the National Institute of Fruit Tree Science (NIFTS), the Kumamoto Prefectural Fruit Tree Experiment Station (KMPFTES), and the Kagoshima Prefectural Fruit Tree Experiment Station (KSPFTES). All Citrus accessions and homozygous S₁ seedlings (S₉S₉ and S₁₀S₁₀) were trees more than ten years old, grown in the Experimental Farm of Kyushu University, Sasaguri, Fukuoka. Self-incompatibility in these accessions had been determined through the observation of pollen tube arrest in their style base eight days after self-pollination (Ngo et al., 2019).

Production of Homozygous S₁ Seedlings of ‘Hirado Buntan’ for the S Gene

Self-incompatible ‘Hirado Buntan’ was used to produce S₁ seedlings homozygous for the S gene (Wakana et al., 2004). Seeds were collected and planted from fruits obtained by bud pollination on ‘Hirado Buntan’, and these S₁ seedlings were top-worked on twenty-year-old ‘Hashimoto Wase’ satsuma mandarin (Citrus unshiu) trees and allowed to grow for more than ten years, until their flowers were available for pollination examination. The S genotype of the top-worked S₁ seedlings homozygous for the S gene (S₉S₉ or S₁₀S₁₀) was determined by Kim et al. (2010). Of the S₁ seedlings homozygous for the S gene, two S₁ seedlings (HBS3 and HBS4) with the S₉S₉ genotype and two S₁ seedlings (Hirado Buntan Seedling No.2; HBS2 and Hirado Buntan Seedling No.11; HBS11) with the S₁₀S₁₀ genotype were used to pollinate Citrus accessions. In addition to the four S₁ seedlings, HBS19 and HBS25 seedlings with the S₉S₉ genotype and an HBS20 seedling with the S₁₀S₁₀ genotype, which were determined in the present study by a self-incompatibility reaction in the lower styles eight days after pollination with S₉ and S₁₀ pollen of HBS3 and HBS2, respectively, were used for the present study because they produced enough pollen to carry out the pollination experiment.

Pollination of Citrus Accessions by Homozygous S₁ Seedlings (S₉S₉ or S₁₀S₁₀)

To examine whether the 146 Citrus accessions share one or both of the S alleles (S₉ and/or S₁₀) with ‘Hirado Buntan’ (S₉S₁₀), they were pollinated with the homozygous S₁ seedlings of ‘Hirado Buntan’, i.e., HBS3, HBS4, HBS19, and HBS25 seedlings with the S₉S₉ genotype and HBS2, HBS11, and HBS20 seedlings with the S₁₀S₁₀ genotype. Normal flowers from which petals and stamens were removed just before flowering were selected, and the stigmas of the normal pistils were pollinated with the S₉ or S₁₀ pollen of ‘Hirado Buntan’ homozygous S₁ seedlings until they showed a yellow color; they were then bagged to prevent further pollination. Five emasculated flowers were pollinated per Citrus accession in each year. Eight days after pollination, the styles were collected and fixed with acetic acid:alcohol (1:3 v/v). These were stored in 70% ethanol at 4 ℃ until use and were softened with 0.8 N sodium hydroxide solution for 24 hours when used. After that, the softened styles were stained overnight with a solution of 0.1% aniline blue. The stained pistils were divided into five parts (stigma, upper style, middle style, lower style, and ovary). The lower styles were observed under a fluorescence microscope (Nikon 50i). When no or fewer than ten pollen tubes were observed in the lower styles, their upper styles or stigmas were observed to ascertain the presence of a large number of pollen tubes. The pollen-tube observation and the determination of S alleles were carried out according to the report of Zhou et al. (2018). The pollen-tube observation was carried out using three pistils per accession and was repeated for two to four years, especially in the accessions in which the incompatibility reaction is not clear. The mean number of pollen tubes in the lower styles observed in each year was averaged per year in each accession.

In the previous study, we determined the S genotype of ‘Kabusu’ sour orange (S_fS₉) by pollination with a homozygous S₁ seedling (S₉S₉) of ‘Hirado Buntan’ (Kim et al., 2010). However, further certification of the ‘Kabusu’ S genotype was carried out in the present study, since it was determined using two pistils and since ‘Zadaidai’ sour orange had a different S genotype (S_fS₁₁) (Kim et al., 2020) from ‘Kabusu’ despite that they are considered as clones (personal communication).

Variation of the Number of Pollen Tubes in Lower Styles of Back-cross-incompatible Pistils (S₉S₁₀) When Pollinated with ‘Hirado Buntan’ (S₉S₁₀)

Five heterozygous S₁ seedlings (HBS5, HBS15, HBS18, HBS25 and HBS27) with the S₉S₁₀ genotype were pollinated with ‘Hirado Buntan’. The number of pollen tubes was examined in lower styles. It was also examined in their upper styles to check the existence of many pollen tubes. The procedure of pollination and pollen tube observation was similar to that mentioned above.

Variation of the Number of Pollen Tubes in the Lower Styles of Incompatible Pistils of Citrus Plants Carrying S₉ and/or S₁₀ Alleles When Pollinated with S₉ or S₁₀ Pollen

In the present study, when ‘Hirado Buntan’ was pollinated with S₉ pollen of ‘Hirado Buntan’ S₁ seedlings, almost no S₉ pollen tubes were observed in the lower styles, whereas when pollinated with S₁₀ pollen of another ‘Hirado Buntan’ S₁ seedling, nine S₉ pollen tubes on average were observed per lower style. To examine the difference in the degree of SI in the lower part of styles between S₉ and S₁₀ pollen tubes, seven S₁ seedlings [HBS2 (S₁₀S₁₀), HBS3 (S₉S₉), HBS4 (S₉S₉), HBS7 (S₉S₉), HBS18 (S₉S₁₀), HBS19 (S₉S₉) and HBS25 (S₉S₉)] and six Citrus accessions [‘Hirado Buntan’ (S₉S₁₀), ‘Kabusu’ (S_fS₉), Amakusa No.6 (S₉S₁₀), Yatsushiro No.4 (S₁₀S₁₁) , ‘Bangkok Buntan’ (S₁₀S_?) and Kawanabe No.2 (S₅S₁₀) ] carrying S₉ and/or S₁₀ alleles were used. HBS3 (S₉S₉), HBS4 (S₉S₉) and HBS25 (S₉S₉) were used as pollen parents generating S₉ pollen, while HBS2 (S₁₀S₁₀) was used as a pollen parent generating S₁₀ pollen. The number of pollen tubes was examined in lower styles, and also in their upper styles to check the existence of many pollen tubes. The procedure of pollination with S₉ or S₁₀ pollen and pollen tube observation was similar to that mentioned above.

Results

Determination of Citrus Accessions with S₉ Pollen

When the S₉ pollen of ‘Hirado Buntan’ pummelo homozygous S₁ seedlings (S₉S₉) was pollinated to 126 Citrus accessions, many S₉ pollen tubes penetrated into the lower styles of 122 accessions (Table 1 and Fig. 1). The average number of pollen tubes observed in the lower styles of the 122 accessions ranged from 38.0 to 488.7. This result suggested that these accessions show cross compatibility with S₉ pollen and do not have an S₉ allele. In four of the 126 accessions, no or only a few pollen tubes penetrated into the lower styles: 0.5 for the Amakusa No. 6 pummelo, 1.0 for the ‘Hirado Buntan’ pummelo, 0.0 for the ‘Kabusu’ sour orange, and 0.0 for the ‘Kikudaidai’ sour orange relative. This result suggested that these accessions show cross incompatibility with S₉ pollen and have an S₉ allele.

Determination of Citrus Accessions with S₁₀ Pollen

When the S₁₀ pollen of ‘Hirado Buntan’ pummelo homozygous S₁ seedlings (S₉S₉) was pollinated to 127 Citrus accessions, many S₉ pollen tubes penetrated into the lower styles of 122 accessions (Table 1 and Fig. 1). The average number of pollen tubes observed in the lower styles of the 122 accessions ranged from 45.0 to 382.7. This result suggested that these accessions show cross compatibility with S₁₀ pollen and do not have an S₁₀ allele. In five of the 127 accessions, no or only a few pollen tubes were observed in the lower styles: 0.0 for Amakusa No. 6 pummelo (Citrus maxima), 9.0 for the ‘Hirado Buntan’ pummelo, 0.0 for the ‘Bangkok Buntan’ (KPFTES) pummelo (Citrus maxima), 1.7 for the Kawanabe No. 2 pummelo (Citrus maxima), and 1.5 for the Yatsushiro No. 4 pummelo (Citrus maxima). This result suggested that these pummelo accessions show cross incompatibility with S₁₀ pollen and have an S₁₀ allele. Except for accessions belonging to the pummelo group, those that have an S₁₀ allele were not detected in the other groups of Citrus accessions examined.

Fig. 1.

Pollen tube arrest in the pistil of ‘Hirado Buntan’ (S₉S₁₀) pollinated with HBS2 (S₁₀S₁₀) (A-D), and normal pollen tube growth in the pistil of ‘Hassaku’ (S₄S₅) pollinated with HBS3 (S₉S₉) (E-H). A and E: Stigma; B and F: upper part of style; C and G: middle part of style; D and H: lower part of style. Note the multiple pollen tubes in each section of the style from the stigma to the style base in right photos and the reduction of the number of pollen tubes from multiple (A) to zero (D) in the left photos. In photo D, no pollen tube is seen; the pollen tube-like structure emitting strong fluorescence is the phloem of conductive tissue in the style base. Bars indicate 0.1 mm.

Variation of the Number of Pollen Tubes in Lower Styles of Back-cross-incompatible Pistils (S₉S₁₀) Pollinated with ‘Hirado Buntan’ (S₉S₁₀)

In the upper styles of back-crossed pistils (S₉S₁₀) with ‘Hirado Buntan’ (S₉S₁₀), more than 200 pollen tubes per style were observed, while in the lower styles the average number of pollen tubes ranged from 2.5 for HBS27 to 8.0 for HBS5 with the average number of 5.8 per style (Table 2). The average number of 5.8 pollen tubes per style was almost the same as that in self-pollination of ‘Hirado Buntan’ reported by Ngo et al. (2019) (Table 2).

Variation of the Number of Pollen Tubes in the Lower Styles of Incompatible Pistils of Citrus Plants Carrying S₉ and/or S₁₀ Alleles When Pollinated with S₉ or S₁₀ Pollen

In the upper styles of all cross-incompatible pistils pollinated with S₉ pollen, more than 72 pollen tubes per style were observed (Table 3). In the lower styles of all pistils pollinated with S₉ pollen, the average number of pollen tubes observed ranged from 0 for Kabusu (S_fS₉) × HBS3 (S₉S₉) to 8.5 for HBS4 (S₉S₉) × HBS3 (S₉S₉) with the average of 2.3 per style (Table 3). In the lower styles of all pistils pollinated with S₉ pollen from HBS3 (S₉S₉), the average number of pollen tubes was 8.5, 0.7 and 5.0 per lower style of HBS4, HBS7 and HBS18, respectively, and 0.8, 0.7 and 0.6 for ‘Hirado Buntan’, ‘Kabusu’ and Amakusa No.6 pummelos (Table 3). HBS19 (S₉S₉) × HBS25 (S₉S₉) and Hirado Buntan (S₉S₁₀) × HBS4 (S₉S₉) showed 1.0 and 2.0 pollen tubes per lower style.

In the upper styles of all cross-incompatible pistils pollinated with S₁₀ pollen from HBS2 (S₁₀S₁₀), more than 180.7 pollen tubes per style were observed (Table 3). In the lower styles of pistils pollinated with S₁₀ pollen, the average number of pollen tubes observed was the highest (9.0) in ‘Hirado Buntan’, followed by 1.7 in Kawanabe No. 2 (S₅S₁₀) and 0 in Amakusa No.6 (S₉S₁₀), Yatsushiro No.4 (S₁₀S₁₁) and ‘Bangkok Buntan’ (S₁₀S_?) pummelos with the average number of 2.1 pollen tubes per lower style (Table 3). The average number of 2.1 per lower style was almost the same as that of 2.3 in cross-incompatible pistils pollinated with S₉ pollen.

Discussion

Distribution of Citrus Accessions with S₉ and/or S₁₀ Alleles

In this study, S gene alleles were found in accessions belonging to pummelo and a pummelo relative for the S₉ allele and those belonging to pummelos for the S₁₀ allele. This suggests that S₉ and S₁₀ alleles have their origin in pummelo and not in mandarin. In the previous study with S₉ and S₁₀ pollens, neither the S₉ nor the S₁₀ allele was found in seven Hirado Island pummelo accessions estimated to be relatives of the ‘Hirado Buntan’ pummelo (Kim et al., 2011). This suggests that the S₉ and S₁₀alleles of ‘Hirado Buntan’ are unique to the Hirado Island pummelo group. In the present study, we found that S₉ and S₁₀alleles exist in the Amakusa No. 6 pummelo, which was collected from an old tree located at Yunuki in the Amakusa Islands, Kumamoto Prefecture. The Amakusa No. 6 pummelo has a morphology similar to that of ‘Hirado Buntan’ in the tree, flowers, and fruit. The old tree was seeded and likely was not grafted, and the self-incompatibility reaction in the pistils to S₁₀ pollen differed from that observed in ‘Hirado Buntan,’ i.e., 0.0 for the Amakusa No. 6 pummelo and 9.0 for ‘Hirado Buntan’ (Table 1). It may be said that the Amakusa No. 6 pummelo and ‘Hirado Buntan’ are not the same plant but different genotypes.

‘Kabusu,’ one of four sour orange accessions examined, has an S₉allele (Table 1) as suggested by Kim et al. (2011), while three of the four accessions have no S₉allele. Kim et al. (2020) reported that the ‘Zadaidai’ sour orange, one of the four sour orange accessions examined in this study, has an S₁₁ allele. On the other hand, two of the four sour orange accessions examined have neither an S₉allele nor an S₁₀allele. Although it has been reported that the sour orange originated from an F₁ hybrid between a pure pummelo and mandarin (Wu et al., 2014) or is an offspring of a cross between lemon and the F₁ of a pummelo and mandarin cross (Shimizu et al., 2016), it has also been reported that sour orange accessions show various differences not only in morphology but also in molecular markers (Siragusa et al., 2006). Thus, the S allele difference among ‘Kabusu’ (S_fS₉), ‘Zadaidai’ (S_fS₁₁), and two of the four sour orange accessions (‘Bonotsu-daidai’ and ‘Bouquet des Fleurs’) may suggest that they are genetically different cultivars, not clones in which S allele mutation occurred during the long history of cultivation. To verify the genotype of the original or pure sour orange, further assessment of common or standard sour oranges is necessary. In the present study, we demonstrated that ‘Kikudaidai,’ a very close relative of sour orange, has an S₉ allele, while Kim et al. (2020) reported that ‘Kikudaidai’ has an S₁₁ allele. This suggests the possibility that ‘Kikudaidai’ (S₉S₁₁) was produced by direct crossing between ‘Kabusu’ and ‘Zadaidai’ or by crossing between a plant with ‘Kabusu’ and one with ‘Zadaidai’ in their pedigrees.

In addition to Amakusa No. 6 (S₉S₁₀) and ‘Hirado Buntan’ (S₉S₁₀), ‘Bangkok Buntan’ (KPFTES), the Kawanabe No. 2 pummelo, and Yatsushiro No. 4 pummelo have an S₁₀allele. Although the S₁₀allele is distributed in pummelo accessions at a higher rate than the S₉allele, it does not appear in the Citrus accessions except for pummelo and ‘Kikudaidai’. These results suggest that the distribution of Citrus accessions with S₉ and/or S₁₀alleles is limited to pummelo and sour orange, probably because of the low fruit quality of these accessions, resulting in limited expansion of Citrus accessions with S₉ and/or S₁₀alleles during the long history of cultivation of pummelo and sour orange—about 500 years (estimation by authors) and about 2000 years in Japan (Tanaka, 1948b) respectively, and for several thousand years in other parts of the world. A ‘Hirado Buntan’ pummelo with high fruit quality was introduced in Japan in 1846 (Tanaka, 1948a), and thus, the introgression of S₉ and/or S₁₀alleles into the Citrus accessions is thought to be very limited in Japan.

Characteristics of the S Allele Frequency for S₉ and S₁₀ Alleles

The rates of Citrus accessions with S_n alleles except for S_f alleles have been reported for the S₁ and S₂ (Kim et al., 2011), S₃ (Kim et al., 2020), S₄ and S₅ (Zhou et al., 2018), and S₁₁ (Kim et al., 2020) alleles of pummelo origin. According to these reports, the rates of Citrus accessions with S_n alleles except for the S_f allele are 29.9% for S₁, 21.3% for S₂, 17.7% for S₃, 9.4% for S₄, 6.3% for S₅, and 13.2% for S₁₁. The present result for the rates of Citrus accessions with S₉ and S₁₀ alleles are 3.1% (4 of 132 accessions) and 3.9% (5 of 129 accessions), respectively. When six sweet orange accessions and two Hyuganatsu accessions (Table 1) are counted as one genotype each, the rates of Citrus genotypes with S₉ and S₁₀ alleles are 3.2% (4 of 126 accessions) and 3.9% (5 of 127 accessions), respectively. The pummelo accessions had S₉ and S₁₀ alleles at higher rates of 2.9% (2 of 70 accessions examined) and 7.0% (5 of 71 accessions examined), respectively. The rates for S₉ and S₁₀ alleles are the lowest and second lowest among the eight rates determined. The rates of pummelo accessions with S_n alleles except for the S_f allele are 34.4% for S₁, 26.4% for S₂, 12.2% for S₃, 1.4% for S₄, 11.8% for S₅, and 33.3% for S₁₁. These S allele frequencies are 17.2% for S₁, 13.2% for S₂, 6.1% for S₃, 0.7% for S₄, 5.9% for S₅, and 16.7% for S₁₁. The rates of pummelo accessions with S₉ and S₁₀ alleles are 2.9% and 7.0%, respectively. Therefore, the S₉ and S₁₀ allele frequencies are 1.8% and 2.3%, respectively. The S₉ and S₁₀ allele frequencies are second and third lowest among the eight S allele frequencies in pummelo. The total of the eight S allele frequencies in pummelo is 64.6%, from which it is roughly estimated that more than twelve S alleles are estimated to exist in pummelo.

Differences in the Self-incompatibility Reaction in Lower Styles between S₉ and S₁₀ Alleles

Ngo et al. (2019) classified the self-incompatibility reaction into several types by the degree of pollen tube arrest in the upper, middle, and lower parts of the styles. They classified self-incompatibility in ‘Hirado Buntan’ into an MMH type (Moderate in stigma, Moderate in upper style, High in middle style). It exhibited, however, a relatively slow or mild self-incompatibility reaction in the lower styles with an average penetration of 5.7 pollen tubes per style and in ovaries with an average penetration of 3.7 pollen tubes per ovary, although the self-pollinated fruit is seedless (Ngo et al., 2019). Yamamoto et al. (2006) observed a similar phenomenon in the self-pollination of ‘Hirado Buntan’ carrying S₉ and S₁₀ pollen. The present data for back-cross-incompatibility in pollinations with ‘Hirado Buntan’ pummelo showed a slow self-incompatibility reaction in the lower part of styles with an average penetration of 5.8 pollen tubes per lower style, similar to the value of self-pollination of ‘Hirado Buntan’ (Table 2). The result suggests that the same extent of mild SI reaction in the lower part of styles occurs in all five back-crosses. This also suggests the possibility that the mild SI reaction in the lower part of styles of ‘Hirado Buntan’ is an inherited trait.

On the other hand, we postulated that the extent of mild SI reaction in the lower part of styles will be different between pollinations with S₉ pollen and S₁₀ pollen, since when ‘Hirado Buntan’ (S₉S₁₀) was pollinated with S₉ pollen of ‘Hirado Buntan’ S₁ seedlings (S₉S₉), 1.0 S₉ pollen tube on average per lower style was observed, whereas when pollinated with S₁₀ pollen of another ‘Hirado Buntan’ S₁ seedling (S₁₀S₁₀), 9.0 S₁₀ pollen tubes on average per lower style were observed (Table 1). The average number of S₉ pollen tubes of HBS3 and HBS4 penetrating into the lower style of ‘Hirado Buntan’ was 1.1 and 2.0, respectively, while the average number of S₁₀ pollen tubes of HBS2 penetrating into the lower style of ‘Hirado Buntan’ was 9.0 (Table 3). The number of S₉ and S₁₀ pollen tubes may suggest that in the lower style of ‘Hirado Buntan’, pollen tubes carrying the S₁₀ allele show a slower incompatibility reaction than those carrying the S₉ allele. Hiratsuka et al. (2012) reported that in the Japanese pear [Pyrus pyrifolia (Burm.f.) Nakai], the degree of self-incompatibility is different between the pollen grains carrying different S alleles. In all crosses with S₉ or S₁₀ pollen, however, no difference was detected in the number of pollen tubes in the lower styles between S₉ and S₁₀ pollen with the average of 1.4 S₉ pollen tubes per lower style and 2.2 S₁₀ pollen tubes per lower style (Table 3). The slow or mild self-incompatibility reaction in the lower styles of self-pollinated ‘Hirado Buntan’ and the back-crossed S₁ seems to be specific to the pollination with ‘Hirado Buntan’. On the other hand, it may be said generally that there is no clear difference in the mild self-incompatibility reaction in the lower styles between S₉ and S₁₀ pollens; the extent of the mild self-incompatibility reaction in the lower styles will be different in pollinations with S₉ or S₁₀ pollen.

Conclusion

Of the 146 Citrus accessions including 82 pummelo accessions examined, self-incompatibility genotypes have been completely ascertained or determined in ‘Kabusu’ sour orange (S_fS₉), a ‘Kikudaidai’ sour orange relative (S₉S₁₁) and two local pummelo plants. The S allele frequencies for S₉ and S₁₀ alleles of pummelo origin are very low in pummelo accessions and eventually in Citrus accessions compared to those of the other S alleles of pummelo origin reported, suggesting pummelo plants with S₉ and S₁₀ alleles did not contribute to the evolution and development of pummelos and eventually Citrus cultivars, except for ‘Hirado Buntan’ pummelo and sour oranges that are used in the commercial production of fresh fruit and root stocks. On the other hand, the S₉ allele may be a good marker to detect cultivars and hybrids with ‘Hirado Buntan’ and sour oranges in their pedigrees. The mild self-incompatibility reaction seen in the lower style of self-pollinated ‘Hirado Buntan’ was also confirmed in the lower styles of the S₁ plants pollinated with ‘Hirado Buntan’. However, there is no clear difference in the mild self-incompatibility reaction in the lower styles between S₉ and S₁₀ pollen. Finally, we examined the degree of self-incompatibility between S₉and S₁₀alleles in the lower part of styles of S₁ seedlings of ‘Hirado Buntan’ and Citrus accessions with S₉and/or S₁₀ alleles. The result indicated no difference in the reaction between the two alleles.