Genetic diversity revealed by assessment of molecular markers
Genetic diversity evaluated by molecular markers provides useful and fundamental information for crop improvement. Because of next-generation sequencing technologies, high-throughput SNPs have become a powerful and popular means to assess genetic relatedness at the genomic level. Nevertheless, SSRs exhibiting relatively high polymorphism level per locus provide rich allelic information for genetic diversity analysis. A currently-preferred phylogenetic relationship of rice germplasm was established by using SSRs that discerned 5 major groups – specifically, aus, indica, aromatic, temperate japonica and tropical japonica (Garris et al. 2005). Phylogenetic trees of rice germplasm using genome-wide SNPs or structural variants were in accordance with these 5 groups (Wang et al. 2018; McNally et al. 2009; Fuentes et al. 2019). In the present study, a total of 953 alleles were detected by 75 polymorphic markers, varying from 3 to 37 alleles per locus with an average of 12.7 (Additional file 2: figure S1, Additional file 1: Table S2), which was higher than other studies (Chakhonkaen et al. 2012; Jin et al. 2010; Nachimuthu et al. 2015). PIC values, which are good indicators of marker polymorphism levels, were in the range of 0.18 to 0.95, with a mean of 0.72, higher than those reported in European (0.49), Chinese (0.42) and Indonesian (0.66) rice germplasm, respectively (Courtois et al. 2013; Jin et al. 2010; Thomson et al. 2007). Moreover, 66 of 75 markers (88%) were considered highly informative with PIC values > 0.5 (Additional file 2: figure S1). Thus, these 75 markers provided rich allelic information for genetic diversity analysis.
The genetic diversity and differentiation in the collected rice germplasm
The diversity panel of 148 accessions could be separated into two subpopulations according to STRUCTURE analysis, clearly corresponding to indica and japonica groups (Additional file 2: Figure S3a). Further division into five subpopulations, indica cultivars, indica landraces, japonica cultivars, japonica landraces and wild rice, were supported by K = 5 (Table 1, Additional file 2: Figure S3b). Most accessions were classified into the expected groups according to the records of the National Plant Genetic Resource Center (NPGRC) Taiwan, however some were incongruent due to admixed genetic background. For example, landraces, Sinceyauo from Japan, Hung-K’o-No from China and Burieuraozu, Nabohai, Paotsupagaiahon, and Tangengenrankatsu from Taiwan, were grouped with japonica cultivars (Pop III); an Aus cultivar (Dular) and O. rufipogon-21 were assigned to the indica landrace group (Popn IV); and three indica cultivars, Basmati 385, Taichung Sen 2 and Taichung Sen 3, were placed in the wild rice group (PopV) (Table 1). These accessions might still share identical by descent segments since derivation from common ancestors. One possible factor contributing to such incongruous findings that cannot be neglected is introgression owing to gene flow among wild species, landraces, and cultivars (Ishikawa et al. 2006, Wang et al. 2018). For example, a mega variety TC65 inherited photoperiod-insensitive alleles of Heading date 1 (Hd1) and Early heading date 1 (Ehd1) from two landraces, Muteke and Nakabo, by spontaneous introgression of natural gene flow during modern breeding (Wei et al. 2016). Indeed, landraces have been commonly used in breeding programs especially in the early purification breeding stage, and wild relatives are becoming more widely used to enlarge gene pools of breeding germplasm. Two old Taiwanese indica cultivars, Taichung Sen 2 and Taichung Sen 3, were derived from landraces based on records of their breeding pedigrees. Thus, admixed accessions are not necessarily rare outcomes of natural introgression, but derive from intentional cross hybridization in at least some cases.
All subpopulations showed high gene diversity (0.75) and PIC (0.72) with mean allele numbers of 12.71 per locus (Table 2). The wild rice group (PopV) had the highest mean allele number (6.91), Nei’s gene diversity (0.74) and PIC value (0.71) despite having fewer accessions (14) than the other subpopulations. Wild rice was also more diverse than landraces and cultivars in other genetic studies (Kovach and McCouch 2008; Londo et al. 2006). As expected, the indica cultivars (PopI) and japonica cultivars (PopIII) exhibited the least diversity, being closely-related to each other within the same subpopulation in this study.
Morphology and genetic background are quite different between indica and japonica rice through independent origins, long-term adaptation to diverse environments and selection for various human preferences. The extent of genetic differentiation between these two subspecies was revealed by FST analysis (Ikehashi 2009, Zhang et al. 2007). High genetic differentiation (FST = 0.3084) was observed between indica and japonica groups in our rice diversity panel (Table 3), in agreement with several studies (Thomson et al. 2007, Lin et al. 2012, Cui et al. 2017). The level of differentiation between indica and japonica landraces (FST = 0.3040) was lower than that between indica and japonica cultivars (FST = 0.4200). Landraces were selected by farmers for adaptation to local environments and various preferences; while modern cultivars result from intense directional selection for specific traits. Less differentiation in landraces than in cultivars was associated with different selection intensity.
The gene diversity of indica accessions was higher than that of japonica accessions since the bottleneck effect was more severe in japonica rice during early domestication (Kovach and McCouch 2008, Wang et al. 2018, Zhu et al. 2007). In the present study, genetic diversity was much lower in japonica than indica populations as well (Table 3). The finding that indica accessions exhibited more diversity than japonica accessions was also supported by PCoA analysis in which indica accessions were sparser on the 2-D plot (Fig. 1), the same tendency as in previous studies using Taiwan breeding germplasm and a collection from Borneo Island ( Lin et al. 2012; Thomson et al. 2009). Nevertheless, in our collection the level of diversity of japonica landraces was higher than that of indica landraces (Table 3) because the former included both upland and lowland accessions.
Unveiling Taiwanese Rice Germplasm
Today, indigenous peoples still cultivate their own landraces with unique traits, such as large grain and aroma, on upland fields in Taiwan. The cultivation of rice, accompanied by foxtail millet, can be dated back to 5000 years ago by unearthed grain remains from some archaeological sites in Tainan Science Park, southern Taiwan (Tsang 2012). Approximately 98% and 83% of the excavated carbonized rice grains from the Tapenkeng Culture period (4800 − 4200 B.P.) and Niuchoutzu Culture period (3800 − 3300 B.P), respectively, were classified as japonica rice according to grain morphology, and the rest were intermediate types with larger grain size than japonica (Tsang 2012; Wang 2007). Quasi-temperate japonica has been proposed (Fuller 2011; Hsu et al. 2019) to have experienced Austronesian spread from Eastern China but there is no strong evidence of prehistoric cultural links or affinity between Fujian (southeast China) and Taiwan (Fuller 2011).
In the present study, 17 landraces labeled with ‘#’ in the Additional file 1: Table S1, were grouped in Clusters VIII, IX, X and XII which belong to the japonica clade (Fig. 2). These indigenous landraces were genetically distinct from modern temperate japonica cultivars, Cluster XI (Fig. 2), and presumed to belong to tropical japonica rice (javanica). The upland landrace, Tangengenrankatsu, has admixed genetic background and is genetically close to O. rufipogon-18. Only few indigenous landraces were clustered in indica clades, albeit some were classified as japonica rice by morphology according to NPGRC records, such as Pakaikauneku, Kaisentetsuchitsu, Napatsupa S3, and Baridon (Additional file 1: Table S1, Fig. 2). Tropical japonica, diverged from temperate japonica, is thought to have originated in the upper Thai-Malay Peninsula and might have moved from the Malay Archipelago northward through Indonesia, the Philippines, Taiwan, Ryukyus, and Japan (Castillo 2017; Chang 1976; Gutaker et al. 2020). Thus, Taiwan was on the dispersal route of tropical japonica and 2/3 of carbonized rice grains unearthed from remains of Niaosung Culture (1400 − 1000 B.P.) had grain length larger than 4 mm which resembled tropical japonica (Wang 2007). In accordance with archaeobotanical evidence, phylogenetic analysis of SSR genotypes classified indigenous upland landraces as tropical japonica (Fig. 2). The tropical japonica landraces preserved by indigenous people experiencing human-mediated movement from Southeast Asia thousands of years ago had been selected for various preferences and for adaptation to hot and humid environments in tropical and subtropical regions. The landraces collected from indigenous Taiwanese tribes about a century ago exhibited much variation in plant architecture, seeds and allele richness of genes conferring domesticated syndrome traits (Hsieh et al. 2011).
The other 4 landraces were genetically related to indica cultivars by PCoA (Fig. 1) and grouped in the same indica clade by phylogenetic analysis (Fig. 2). All except Pai-Mei-Fan were assigned to cluster I-IV in the unrooted neighbor-joining tree, distinct from indica cultivars grouped in cluster V (Fig. 2). Only 6 accessions were recorded with indigenous language pronunciations, including 5 (Baridon, Napatsupai S3, Pakaikameku, Kaisentetsuchitsu, Hopot Utatyaru) in cluster I and Parahainakoru in cluster III. These 6 indica landraces might have been preserved and cultivated by indigenous people for thousands of years, however archaeobotanical evidence is lacking. We cannot rule out that these indica landraces were adopted by indigenous people only hundreds of years ago, after Chinese introduced much indica rice. About 400 years ago, immigrants from Guangdong and Fujian, Southeast China brought indica rice prevailing in these areas to Taiwan, and 1,197 indica accessions were identified officially in the early 20th century (Iso 1964). The landraces in cluster II – IV came from Taiwan and China and showed no significant isolation-by-distance (Fig. 2). However, the indica landraces were divided into two large clades, Cluster I & II and Cluster III & IV, which might reflect two origins, Guangdong and Fujian. The genetic diversity of indica landraces in Taiwan is relatively high (Table 3, 4) which might result from intrinsic high variation in indica rice and multiple origins as well. Taiwanese indica cultivars are closer to IR64 than Dular, an Aus cultivar in India (Fig. 2). Based on the pedigree information, 14 of 17 Taiwanese indica cultivars can be traced back to IRRI accessions or DGWG as their breeding parents (Lu and Lu 2010).
Landraces are selected by farmers and cultivated regionally by traditional practices, employing various preferences with loose selection and adaptation to diverse environments. Landraces, intermediate between wild relatives and cultivars, exhibit higher genetic diversity and morphological variation than cultivars. Landraces are an important genetic reservoir for crop improvement to cope with climate changes and increase sustainability. In Taiwan, 16 officially acknowledged indigenous peoples have their own cultures and diet preferences, including diversified crop germplasm. Four Waxy alleles were identified in Taiwanese foxtail millet landraces, and Waxy genotypes were associated with various culinary and cultural preferences, leading to genetic variation of physicochemical properties and digestibility (Kuo et al. 2018; Yin et al. 2019). Taiwanese rice landraces compromised of tropical japonica and indica rice revealed diverse genetic variation in plant architecture and seeds (Hsieh et al. 2011) and herein showed much SSR diversity (Table 3, 4). This high genetic variation indicates that Taiwanese landraces are a reservoir of genetic diversity and beneficial genes/alleles for rice breeding and improvement.
Taiwanese landraces have had great impact on modern rice breeding not only in Taiwan but also elsewhere in the world. According to the database of rice breeding pedigrees (Taiwan Rice Information System, TRIS), Taiwanese landraces were commonly used to introgress useful genes for rice improvement, especially in the early breeding programs a half-century ago. The most prominent varieties, japonica TC65 with photoperiod insensitivity and indica variety Taichung Native 1 (TCN1) with semi-dwarf stature, have had great impact on rice breeding and research. TC65 inherited the null function allele of hd1 with a 1901-bp insertion on exon 1 (Yano et al. 2000) and ehd1 with one nonsynonymous substitution in the GARP domain (Doi et al. 2004) via natural introgression from two upland landraces, Nakabo and Muteka (Wei et al. 2016). Because photoperiod insensitivity was a highly desired trait, TC65 had been extensively applied in modern rice breeding programs, leading to all current Taiwanese temperate japonica cultivars inheriting the ehd1 and hd1 alleles. A very popular new japonica variety, Tainan 16 released in 2012, was introgressed with ehd1 and hd1 alleles from Taiwanese cultivar Tainung 67 to Japanese elite cultivar Koshihikari by using marker-assisted backcross selection (Chen et al. 2010; Chen et al. 2012). Taiwanese temperate japonica cultivars can be cultivated in two crop seasons under tropical and subtropical environments, making Taiwan the southmost region of temperate japonica cultivation. The indica variety TCN1 inherited null function of sd1 with a 383-bp deletion from the landrace DGWG (Sasaki et al. 2002). In addition to IR8, the miracle rice with high yield to which is attributed the Green Revolution in Asia in the 1970s (Evenson and Gollin 2003), this DGWG allele has been widely applied to improve grain yield of both indica and japonica varieties in the past 50 years (Asano et al. 2011; Zhao et al. 2010). Yet, there are still numerous useful genes/alleles existing in the genetic reservoir of Taiwanese landraces, for example conferring large grain size, aroma, and biotic and abiotic resistance. Untapped beneficial genes from landraces can help to breed new varieties for resilient and sustainable agriculture.
Modern cultivars are a result of intensive directional selection for specific traits which are frequently determined by government policy and demands of markets. In Taiwan, the major dining staple was indica rice before War World II but changed to temperate japonica rice because of government policy during Japanese occupation. Now, japonica rice is for dining; while indica rice is used for various food processing needs, such rice noodles, pudding, and cakes. Thus, japonica and indica improvement have different breeding goals. For indica rice, high yield with resistances to biotic and abiotic stresses are breeding goals; thus, diverse germplasm from landraces or introduced from other countries are commonly utilized as donor parents (Lin et al. 2012). Therefore, there was no obvious difference in genetic diversity and differentiation between early and late indica cultivars (FST=0.0045, Table 4). On the other hand, the government policy for the breeding goal of japonica rice was changed from high yield to premium grain quality in 1981 because rice demand had been decreasing due to miscellaneous factors and westernization of diets. The germplasm used for improving different traits seemed to be associated with high differentiation between early and late japonica cultivars, FST=0.3751 (Table 4). In order to improve grain quality, a few Japanese elite temperate japonica cultivars were introduced and used extensively in recurrent breeding crosses (Lin et al. 2012; Wu and Lin 2008). This led to modern Taiwanese japonica cultivars being grouped at the same clade, Cluster XI with the Japanese elite cultivar, Nipponbare (Fig. 2), as japonica varieties from Taiwan and Japan did not differ significantly in the pattern of genetic diversity (Lin et al. 2012). The genetic distances between any two Taiwanese japonica cultivars were in the range of 0.43–0.58 (Fig. 2); consequently, the gene pool of japonica cultivars is relative narrow as compared to either japonica landraces or modern indica cultivars (Table 3, 4), resulting in genetic vulnerability in rice cultivation and management.
To overcome severe genetic vulnerability of temperate japonica cultivars, wild relatives and indica rice were introduced to breeding programs. For example, japonica Tainung 67 was the descent of a cross of japonica Tainung 61 and O. rufipogon, and japonica Taichung 192 was an indica/japonica-crossed variety (Lu and Lu 2010). In the past two decades, aromatic rice varieties have become popular in Taiwan, such as japonica cv Kaohsiung 147, Taichung 194, Taikeng 4, Tainung 71, and Taoyuan 3, for which Basmati from India and Taishousen from Japan were the important donor parents of aroma genes. Recently, numerous advanced breeding lines introduced from IRRI and wild relatives have been used in breeding programs to improve biotic and abiotic stresses for sustainable agriculture, e. g. IRBB66 pyramided with 5 bacterial blight resistant genes (Yap et al. 2016). Thus, current rice breeding goals in Taiwan emphasize grain quality first, followed by other traits such as resistances and multi-dimensional utilizations (forage and landscape). To achieve various goals, germplasm for breeding are not limited to the domestics but also exotics.