Association of functional nucleotide polymorphisms at DTH2 with the northward expansion of rice cultivation in Asia

Abstract

Flowering time (i.e., heading date in crops) is an important ecological trait that determines growing seasons and regional adaptability of plants to specific natural environments. Rice (Oryza sativa L.) is a short-day plant that originated in the tropics. Increasing evidence suggests that the northward expansion of cultivated rice was accompanied by human selection of the heading date under noninductive long-day (LD) conditions. We report here the molecular cloning and characterization of DTH2 (for Days to heading on chromosome 2), a minor-effect quantitative trait locus that promotes heading under LD conditions. We show that DTH2 encodes a CONSTANS-like protein that promotes heading by inducing the florigen genes Heading date 3a and RICE FLOWERING LOCUS T 1, and it acts independently of the known floral integrators Heading date 1 and Early heading date 1. Moreover, association analysis and transgenic experiments identified two functional nucleotide polymorphisms in DTH2 that correlated with early heading and increased reproductive fitness under natural LD conditions in northern Asia. Our combined population genetics and network analyses suggest that DTH2 likely represents a target of human selection for adaptation to LD conditions during rice domestication and/or improvement, demonstrating an important role of minor-effect quantitative trait loci in crop adaptation and breeding.

Success of modern agriculture relies on the better adaptation of crops to local environments and the expansion of their cultivation area, which is largely determined by optimal flowering time (i.e., heading date in rice) (1). Photoperiodic sensitivity that tailors vegetative and reproductive growth to local climates is considered to be the most important factor in determining flowering time and thus ensures crop adapt to specific ecological conditions and environmental variation (1, 2). In optimal growing conditions, late flowering allows a longer vegetative growth period that promotes the accumulation and allocation of more resources to seed production, whereas early flowering is favored in environments with a short or unpredictable growing season (3). The trade-off between resource allocation and stress avoidance is particularly important for crop yield and quality (3). Deciphering the molecular mechanisms underlying local adaptation of plants is of great interest not only to plant biologists and evolutionary biologists, but also to crop breeders because it could provide useful guidelines for identifying target genes for selection. In recent decades, several flowering genes involved in local adaptation have been identified, including FRIGIDA and FLOWERING LOCUS C in Arabidopsis; Dwarf8 in maize; Ppd-D1 in wheat (Triticum aestivum); EARLY MATURITY 8 in barley (Hordeum vulgare); vernalization genes VRN1-3 in wheat and barley; and FLOWERING LOCUS T (FT) in sunflower (Helianthus annuus) (4–13). However, elucidating the adaptive genetic variation in flowering-time control and the underlying molecular mechanisms has remained a challenging task.

As a primary staple for billions of people worldwide, rice has been a major target of natural or artificial selection during domestication and breeding (14). Recently, several genes associated with the domestication of cultivated rice have been cloned, including those controlling shattering (shattering4 and QTL of seed shattering in chromosome 1), pericarp color (Rc), plant architecture (PROSTRATE GROWTH 1 and Semi-dwarf1), and seed dormancy (Seed dormancy 4) (15–21). Rice is a short-day (SD) plant but is cultivated in diverse environments at latitudes ranging from 55°N in China to 36°S in Chile, far beyond the geographical range of its wild progenitor, Oryza rufipogon (22). The northward expansion of rice cultivation to long-day (LD) environments relative to its wild progenitor and the development of different cultivars adapted to local environments imply a strong impact of artificial selection on photoperiodic genes during rice domestication and breeding (22, 23).

Previous studies have revealed that rice heading is promoted by Heading date 3a (Hd3a, the major SD florigen) and RICE FT 1 (RFT1, the closest paralogue of Hd3a that works as an LD florigen) (24). Upstream of these florigen genes, Heading date 1 (Hd1) and Early heading date 1 (Ehd1) function as two major floral signal integrators that receive multiple signals from other genes, including OsGIGANTEA (OsGI), Rice Indeterminate 1 (RID1; also known as OsID1/Ehd2), Early heading date 3 (Ehd3), OsMADS50, OsMADS51, Ghd7 (for Grain number, plant height, and heading date 7), DTH8 (for days to heading on chromosome 8), PHYB (PHYTOCHROME B), and SE5 (PHOTOPERIOD SENSITIVITY 5), to regulate the expression of florigen genes (2, 25–31). Analyses of these flowering-time genes have given several hints about local adaptation in rice. For example, it has been suggested that combination of Hd1 and Ehd1 alleles is important for flowering-time traits in rice (1). Sequence analysis of the Ghd7 allelic variants demonstrated an important role of this locus in the adaptability of cultivated rice on a global scale (28). However, the association of flowering-time traits with most flowering-time gene haplotypes in rice cultivars awaits in-depth investigations in terms of local adaptation of these genes.

In this study, we cloned a minor-effect quantitative trait locus (QTL), DTH2 (for Days to heading on chromosome 2), which induces heading in LD conditions. We identified two functional nucleotide polymorphisms (FNPs) in DTH2 that are associated with the changes in flowering time and increased reproductive fitness during northward expansion of rice cultivation. This study also provides strong evidence that minor-effect QTLs play important roles in crop adaptation and diversification, and thus might be major targets of artificial selection during crop domestication and breeding.

Results

QTL Analysis for Heading Date and Map-Based Cloning of DTH2.

We have identified a minor-effect locus named qDTH-2 (QTL for DTH2, designated as DTH2 in this study) by using a recombinant inbred line population derived from a cross between Asominori (Aso, japonica) and IR24 (indica) (32). To study the genetic effect of DTH2 on heading date, a nearly isogenic line (NIL) carrying a 9.6-cM chromosomal segment of IR24 containing DTH2 (between markers RM318 and RM240; SI Appendix, Fig. S1) in the Aso background was developed (Methods). Compared with Aso, the NIL plants exhibited a 7.4-d delay in heading date under natural LD (NLD) conditions in Beijing (northern China, 116.4° E/39.9° N, day length > 15 h), but not in natural SD (NSD) conditions in Hainan (southern China, 110.0° E/18.5° N, day length < 12 h; Fig. 1 A and B). NIL also showed delayed heading in controlled LD (CLD) conditions (14 h light/10 h dark) but not in controlled SD (CSD) conditions (10 h light/14 h dark; Fig. 1B). Interestingly, we found that, under NLD conditions, approximately 81.5% of Aso grains, but only 28.6% of NIL grains, reached maturity (judged by yellow pigmentation of the seed coat). In contrast, all grains from Aso and NIL matured under NSD conditions (Fig. 1 C and D). The NIL showed a leaf emergence rate similar to Aso under CLD and CSD conditions (SI Appendix, Fig. S2), suggesting that DTH2 is an LD-specific regulator of rice heading. As these lines are ∼97% isogenic with Aso (Methods), these results suggest that in the Aso genetic background the Aso allele of DTH2 is better adapted to the NLD conditions in terms of reproductive fitness.

Fig. 1.

Phenotypes of the parental lines, map-based cloning, complementation, and RNAi of DTH2. (A) The phenotypes of Aso and NIL under NLD conditions in Beijing. Photo was taken at the Aso heading stage. (B) Days to heading of the parental lines under four different conditions. The open and shaded bars represent Aso and NIL, respectively. P values were obtained in two-tailed t tests. (C) The brown rice of Aso and NIL from panicles sampled 60 d after Aso heading under NLD conditions. (D) Seed maturity percentages of Aso and NIL under NLD and NSD conditions. (E) Map-based cloning of DTH2. (Lower) Coding region structure of DTH2 and natural variation between the Aso (DTH2-a) and NIL (DTH2-i) alleles. Rectangles and lines represent exons and introns, respectively. (F and G) Days to heading (F) and seed maturity percentages (G) of T₃ transgenic lines including empty vector–NIL (EV-N), complementation (nos. 18-7-2 and 26-1-4), empty vector–Aso (EV-A), and RNAi (nos. 5-4-2 and 38-37-4) under NLD conditions. P¹ and P² indicate two-tailed t test and Wilcoxon signed-rank test results between complementation and EV-N plants and between RNAi and EV-A plants, respectively. Numbers of plants used in each study are indicated in the columns. Data are presented as mean ± SEM.

Genetic analyses of an F₂ population derived from a cross of Aso and NIL showed 3:1 segregation of early- and late-heading plants (3,213:1,099; χ² = 0.55, P = 0.46), indicating that DTH2 behaves as a single Mendelian factor. For positional cloning of DTH2, we used the average phenotypic values of F₂-derived families (F_2:3) to measure the heading date (Methods). Using 1,099 late-heading F_2:3 families (SI Appendix, Fig. S3), we narrowed DTH2 down to a 37.6-kb genomic region between the markers dcaps3 and dcaps5, co-segregating with caps1 (Fig. 1E). Nine putative ORFs were predicted in this region (Rice Genome Annotation Project, http://rice.plantbiology.msu.edu/index.shtml;Fig. 1E). Of these, ORF5 (LOC_Os02g49230) is predicted to encode a CONSTANS (CO)-like protein that contains one zinc finger domain (5–90 aa) at its N-terminal region and one conserved CCT (CO, CO-like, TOC1) domain (350–392 aa) at its C-terminal region (Plant Transcription Factor Database, http://plntfdb.bio.uni-potsdam.de/v3.0/; SI Appendix, Figs. S4 and S5). Four SNPs were identified in the coding sequence (CDS) of ORF5 between the Aso allele (designated DTH2-a) and IR24 allele (designated DTH2-i; Fig. 1E). There is a rice gene closely related to ORF5 (LOC_Os06g19444) that shares 76% amino acid identity with ORF5 (SI Appendix, Figs. S4 and S5). A 7,867-bp genomic fragment (SI Appendix, SI Methods) from Aso was able to fully rescue the late-heading and low seed maturity phenotypes in transgenic NIL plants under NLD conditions (Fig. 1 F and G). Further, transgenic RNAi Aso plants with reduced expression of DTH2-a showed late-heading and low seed maturity percentages under NLD conditions compared with the Aso plants transformed with the empty vector (Fig. 1 F and G and SI Appendix, Fig. S6). We thus concluded that ORF5 is DTH2.

DTH2 Protein Is Targeted to Nucleus and Has Transcriptional Activation Activity.

To determine the subcellular localization of DTH2 protein, we fused the full-length CDS of DTH2-a with that of GFP. Transient expression of DTH2-a–GFP fusion protein was exclusively colocalized with a nuclear marker, OsMADS3-mCherry fusion protein, in rice leaf protoplasts (SI Appendix, Fig. S7 A–D). Further, transactivation activity assays showed that DTH2 had a strong transcriptional activation activity in yeast (SI Appendix, Fig. S7E). Deletion analysis showed that neither the zinc finger domain nor the CCT domain alone, but the middle region between them, was essential for its transactivational activity (SI Appendix, Fig. S7E), which is similar to CO (33). These results support a role of DTH2 in transcriptional regulation of gene expression in the nucleus.

DTH2 Expression Is Regulated by the Circadian Clock.

To examine the temporal and spatial expression patterns of DTH2, we examined its expression levels in various Aso tissues collected from plants grown in NLD conditions by quantitative real-time PCR (qRT-PCR). DTH2 transcripts were detected in all tissues examined, but its expression was most abundant in leaf blade and sheath (SI Appendix, Fig. S8A). Consistent with our qRT-PCR results, GUS signals were detected in leaf blade, sheath, panicle, culm, and root of the pDTH2-a::GUS transgenic plants, with the strongest accumulation in the vascular tissues of the various organs (SI Appendix, Fig. S8 B–J). To investigate whether DTH2 expression is regulated by the circadian clock, Aso and NIL plants were entrained under CLD/CSD conditions for 2 d and then transferred to continuous dark (DD)/continuous light (LL) conditions for 2 d (Fig. 2). Remarkably, qRT-PCR assays showed that the DTH2-a and DTH2-i exhibited similar expression levels and patterns. Their expression started to accumulate 8 h after dawn, peaked 2 h after dusk, and started to decrease quickly thereafter under CLD and CSD conditions (Fig. 2). Notably, the rhythmic amplitude of DTH2 expression drastically reduced when released into DD, but continued to oscillate when released into LL following 48-h CLD and CSD clock entrainment (Fig. 2), suggesting that DTH2 expression is controlled by the circadian clock under these free-running conditions.

Fig. 2.

qRT-PCR analysis of diurnal expression patterns of DTH2-a and DTH2-i alleles from CLD to DD conditions (A), from CSD to DD conditions (B), from CLD to LL conditions (C), and from CSD to LL conditions (D). The white and black bars represent light and dark periods, respectively. The dark gray bars indicate subjective light during DD conditions; the light gray bars indicate subjective dark during LL conditions. The expression level is shown as a ratio with the rice Ubiquitin (UBQ) gene. Mean values ± SD were obtained from three technical repeats and two biological repeats.

DTH2 Promotes Heading by Up-Regulating Hd3a and RFT1.

The diurnal expression pattern of DTH2 suggests that it participates in photoperiodic regulation of flowering. To test this, we compared the expression of several photoperiod-related genes in Aso and NIL plants under CLD and CSD conditions. We found that the mRNA levels of two florigen genes, Hd3a and RFT1, were higher in Aso than NIL under CLD conditions (Fig. 3 A and B), but almost the same under CSD conditions during the 24-h period (Fig. 3 E and F). Consistent with this, expression of the downstream floral meristem identity genes, OsMADS14 and OsMADS15 (24), was also altered in NIL only under CLD conditions (SI Appendix, Fig. S9 A and B).

Fig. 3.

qRT-PCR analysis of Hd3a, RFT1, Hd1, and Ehd1 diurnal expression patterns in Aso and NIL under CLD (A–D) and CSD conditions (E–H). The white and black bars represent the light and dark periods, respectively. The expression level is shown as the ratio with the rice UBQ gene. Mean values ± SD were obtained from three technical repeats and two biological repeats.

In rice, floral signals are transduced through two major floral integrators, Hd1 and Ehd1 (1, 2). We next examined whether DTH2 regulation of Hd3a and RFT1 is mediated through Hd1, Ehd1, or both. To our surprise, no significant difference in the expression of Hd1 and Ehd1 was found between Aso and NIL (Fig. 3 C, D, G, and H). Moreover, the expression of DTH2 in NILs carrying a deficient hd1 or ehd1 allele was comparable to that in their corresponding WT plants (SI Appendix, Fig. S10 A and B). These results suggest that the DTH2-dependent effect on the florigen genes is independent of Hd1 and Ehd1. In support of this notion, the expression of other regulators of Hd1 and Ehd1, such as RID1/OsID1/Ehd2, Ehd3, OsMADS50, OsMADS51, Ghd7, DTH8, PHYB, and SE5, was also not affected by DTH2 (SI Appendix, Fig. S9 A and B). Conversely, expression of DTH2 was also not affected by these floral regulators (SI Appendix, Fig. S10 C–J). Further, we generated a T65/NIL F₂ population with 2,000 individuals (SI Appendix, SI Methods), and isolated 16 hd1/ehd1/DTH2-i triple mutants from the population. We found that all 16 triple mutants flowered later than the T65 parental line [T65 contains nonfunctional alleles at both Hd1 and Ehd1 (26); P = 6.3 × 10⁻⁵; SI Appendix, Fig. S11]. Together, these results support the notion that DTH2 represents a floral integrator that induces heading under LD conditions, acting independently of Hd1 and Ehd1.

Identification of Two FNPs at DTH2.

To explore the molecular basis of the functional differences between DTH2-a and DTH2-i in regulating heading, we compared the CDS of DTH2-a and DTH2-i and found four SNPs at positions 25 (S1: A/G), 461 (S2: C/T), 1645 (S3: T/C), and 1721 (S4: T/G) (Fig. 1E). S3 is a synonymous SNP, whereas the rest are nonsynonymous, corresponding to Arg/Gly (R9G), Ala/Val (A154V), and Tyr/Asp (Y319D) amino acid replacements, respectively (Fig. 1E). We then compared the DTH2 coding region sequences from 80 rice landraces and did not find additional polymorphisms (SI Appendix, Table S1). We next analyzed the correlation between each of three nonsynonymous SNPs (S1, S2, and S4) and the heading date of the 80 landraces grown in NLD conditions (SI Appendix, Table S1). Significant correlations were found for S1 (r = 0.30, P = 0.02) and S4 (r = 0.47, P = 9.03 × 10⁻⁵), but not for S2 (r = 0.21, P = 0.10).

To confirm the effects of S1 and S4, we examined the heading phenotypes of transgenic plants expressing each of the following four transgenes: A1, A2, A3, and A4. A1 (S1G/S4G) and A4 (S1A/S4T) constructs are identical to NIL (DTH2-i) and Aso (DTH2-a), respectively, whereas A2 and A3 contain S1A/S4G and S1G/S4T recombinant CDS, respectively (Fig. 4A). All transgenes were driven by the DTH2-a promoter and were individually transformed into the NIL plants. We found that, except for the A1 transgene, A2, A3, and A4 transgenes all promoted heading and seed maturity, with A4 having the strongest promoting effect in NLD conditions (Fig. 4 B and C). Taken together, our association studies and transgenic experiments suggest that S1 and S4 are FNPs in terms of heading date regulation.

Fig. 4.

Natural variation and network analysis for DTH2. (A) Diagram showing the four transgenic constructs representing four allele types based on the two polymorphic sites (S1 and S4). The black and white bars denote exons and introns, respectively. (B and C) Days to heading (B) and seed maturity percentages (C) of the transgenic plants harboring the four different transgenes under NLD conditions. Two independent T₁ transgenic families for each transgene (nos. 27 and 43 for A1, nos. 10 and 16 for A2, nos. 29 and 30 for A3, and nos. 14 and 17 for A4) were grown under NLD conditions in Beijing. Plus and minus symbols indicate transgene-positive and transgene-negative segregants, respectively. P¹ and P² are P values produced by two-tailed t tests and Wilcoxon signed-rank tests between the positive and negative plants, respectively. Numbers of plants are indicated in the columns. Data are presented as mean ± SEM. (D) Allele network of the DTH2 for 127 cultivated and wild rice. Three groups (A1, A2, and A4) of alleles were identified, with each group consisting of multiple allele subtypes based on the SNPs uncovered in the introns of DTH2. Allele frequencies are proportional to the size of the circles. The length of the lines between the circles indicates the evolutionary distance between the two alleles. The proportions of the four rice groups (indica, japonica, javanica, and wild rice) are represented by different color codes. (E) Distribution map of 80 rice landraces carrying different DTH2 alleles.

Network and Genetic Diversity Analyses for the DTH2 Alleles in Cultivated and Wild Rice.

To examine the evolutionary origin of the DTH2 alleles, we sequenced the DTH2 coding regions of 47 wild rice accessions (O. rufipogon; SI Appendix, Table S1) and performed a network analysis (34) together with the 80 sequenced landraces. Based on the combination of two FNPs, all alleles identified could be classified into three groups (A1, A2, and A4), with no A3 allele found so far (Fig. 4D). Of them, the A1 (including DTH2-i) and A2 groups were found in wild rice and landraces growing in tropical and subtropical areas (indica and javanica), whereas the A4 group (including DTH2-a) was found exclusively in japonica landraces growing in temperate areas (SI Appendix, Table S1). This result suggests that the A4 group represents a derived allele group that arose and was fixed during the domestication or improvement of japonica rice. Interestingly, the distribution of these DTH2 alleles in cultivated rice is closely correlated with the latitude, with the A4 allele arising in parallel with the domestication and/or migration of japonica rice (Fig. 4E).

Together with the transgenic and association studies, we inferred that the A4 allele may have undergone intensive selection during the northward expansion of japonica rice to NLD conditions in Asia. To test for this inference, we compared the diversity between cultivated and wild rice and conducted multiple neutrality tests. The diversity test showed that, on average, the japonica (samples with A4 allele) exhibited a much lower diversity (π = 0.0002) than other cultivated rice groups (π = 0.0006 for indica, π = 0.0011 for javanica) and wild rice (π = 0.0014; SI Appendix, Table S2). The diversity ratio between japonica and wild rice (0.08) is far below that (0.2) observed for random gene fragments across the japonica and wild rice genomes (35). We further examined whether the nucleotide polymorphism data fit the neutral model by using the Tajima's D (36) and Hudson–Kreitman–Aguade (37) tests with a set of known neutral genes (Adh1, GBSSII, Ks1, Lhs1, Os0053, SSII1, and TFIIAγ-1) (35) as controls. Neither Tajima's D nor Hudson–Kreitman–Aguade values significantly deviated from the neutral expectation (SI Appendix, Table S2). However, this result may be caused by the extremely low diversity of DTH2 in cultivated and wild rice. As an alternative, we sequenced and calculated the diversity for the 20 genes surrounding DTH2, as selection might lead to a selective sweep in the flanking region of the target gene (20). As expected, the average nucleotide diversity of the 20 genes in japonica (π = 0.000263) is much lower than those of other rice groups (π = 0.0022 for indica, π = 0.0033 for javanica) and wild rice (π = 0.0043) (SI Appendix, Fig. S12).

To examine whether the low genetic diversity of japonica could be caused by a domestication bottleneck (35) rather than selection, we performed a coalescent simulation with two derived populations that underwent a bottleneck in reference to seven neutrally evolving genes (SI Appendix, SI Methods and Fig. S13). Relative to the K value of 0.2 for neutral genes (35) (SI Appendix, SI Methods), we obtained a much lower K value (0.005) for the genes surrounding DTH2, a significant departure from neutral genes (likelihood ratio of 9.41, P = 0.0022). This result indicates that the reduced diversity at the genes surrounding DTH2 in japonica cannot be explained by a domestication bottleneck alone, and is most likely the result of selective sweep around DTH2, providing additional evidence that DTH2 has been subjected to artificial selection during domestication or improvement of japonica rice.

Discussion

Functional Conservation and Divergence of CO-Like Genes Among Species.

We cloned the DTH2 gene by using a map-based cloning approach and showed that DTH2 encodes a CO-like protein. qRT-PCR analysis showed that, similar to CO, COL9, and Hd1 (38–40), DTH2 expression is regulated by the circadian clock, with a peak of expression at night and lowest expression during the light period under CLD and CSD conditions (Fig. 2). Notably, members of the CO-like gene family have been shown to play important roles in the photoperiodic regulation of flowering in many plant species and are conserved between monocots and dicots (41, 42). There are 17 and 19 CO-like genes in the Arabidopsis and rice genomes, respectively (41) (SI Appendix, Fig. S4). Interestingly, previous studies showed that two CO-like genes, Hd1 and Ghd7, negatively regulate flowering under LD conditions in rice, whereas Hd1 has another role in promoting flowering under SD conditions (25, 28). In addition, the wheat and barley homologues of rice Hd1/Ghd7 (TaHd1/VRN2 and HvCO1/HvVRN2, respectively) are also involved in regulating flowering (10, 43–45). Phylogenetic analysis revealed that DTH2 shares the highest sequence homology with the Arabidopsis COL9 gene (39) (SI Appendix, Figs. S4 and S5). However, it has been proposed that Arabidopsis COL9 possibly functions as an LD-specific repressor of flowering by suppressing the expression of CO and FT (39), and we show that DTH2 acts as promoter of rice heading under LD conditions. Thus, DTH2 and COL9 may regulate photoperiodic flowering via different mechanisms in rice and Arabidopsis, respectively. It will be interesting to investigate whether posttranscriptional regulation [e.g., protein stability (38)] is involved in the regulation of DTH2 function in future studies.

DTH2 Is Involved in Flowering-Time Local Adaptation and Association of Two FNPs in DTH2 with Northward Expansion of Rice Cultivation.

Flowering time is one of the best-studied ecologically significant traits under natural or artificial selection for adaptation of plants to specific natural environments (1, 2). However, only a few flowering-time genes that function in adaptation have been identified, and their underlying molecular mechanisms have been elucidated (4–13). Among the flowering genes identified in Arabidopsis, only FRIGIDA and FLOWERING LOCUS C, two pivotal regulators of the vernalization pathway, show molecular hallmarks of natural selection, and the polymorphisms in these genes appear to underlie the extensive natural variation of flowering time for local adaptation (4, 5). It is believed that the northward expansion of rice into the LD environments during the past several thousand years is largely facilitated by early flowering and reduced photoperiod sensitivity, because the optimum flowering time in adaptation to agroenvironments and crop seasons can secure a harvest before cold weather approaches (1). However, molecular and ecological evidence supporting such a claim has not yet been forthcoming.

Based on multiple lines of evidence, we argue that DTH2 is a likely target of artificial selection during rice domestication under the LD environments in northern Asia. First, Aso (carrying the DTH2-a allele) flowered 1 wk earlier than NIL (carrying the DTH2-i allele) under NLD and CLD conditions, conferring increased reproductive fitness and better adaptation to NLD conditions (i.e., higher percentages of mature grains; Fig.1C). Second, our transgenic experiments showed that the A4 transgene (containing S1 and S4) had the strongest promoting effect in NLD conditions, demonstrating that S1 and S4 are two FNPs (Fig. 4 B and C). Third, our sequence analysis and association studies found that the distribution of DTH2 alleles is closely correlated with latitude, and the A4 (DTH2-a) allele was derived recently and predominantly existed in the japonica landraces growing in temperate areas with low temperatures and short growing seasons, including Japan, Korea, and northern China (Fig. 4 D and E). In addition to a much lower nucleotide diversity at DTH2 in japonica than in other cultivated and wild rice, we found that the average nucleotide diversity of the 20 genes surrounding DTH2 is significantly reduced in japonica relative to other cultivated and wild rice groups. Particularly, our simulation indicates that this substantial reduction of nucleotide diversity surrounding the DTH2 region cannot be explained by a domestication bottleneck alone (SI Appendix, Fig. S12), providing a signature of selective sweep in the DTH2 region of japonica rice. Together, these observations suggest that the two FNPs (S1 and S4) may contribute significantly to the distribution and ecological adaptation of rice varieties to different local climates, and that the A4 allele (DTH2-a) might have been selected during northward expansion of rice cultivation.

Importance of Minor-Effect QTLs in Crop Adaptation and Breeding.

Recent theoretical studies proposed that, during adaptation and speciation, fewer mutations of large effect could be fixed at the early stage of adaptation, and additional mutations of decreasing effects fixed subsequently (46). This notion is supported by recent molecular genetic studies and sequence analyses (47, 48). For example, the rice shattering4 and QTL of seed shattering in chromosome 1 that control shattering in rice are among the major loci of large effect selected early during rice domestication (15, 16, 49); however, accumulating evidence suggests that there are other shatter-controlling genes differentially fixed within different subpopulations of rice (48). Here, we found that a relatively minor alteration (∼1 wk) in heading date caused by two FNPs in DTH2 resulted in relatively large differences in rice grain yield and reproductive fitness. The present study provides an empirical case showing how plants adapt to specific or local environments by keeping the balance between resource accumulation and stress avoidance. Our results also echo the report that minor-effect QTLs contributed to the adaptation of maize to northern environments (47). Thus, we anticipate that more minor-effect QTLs contributing to crop adaptation and diversification will be uncovered with increased knowledge of plant genomes and more genome-wide association studies.

Methods

Fine Mapping of DTH2.

The 66 chromosomal-segment substitution lines (CSSLs) were derived from a cross between Aso and IR24 (50). We selected CSSL23 to cross once and backcross four times to Aso, and then the self-pollinated BC₄F₁ plants were used to construct a CSSL23/Aso BC₄F₂ population with 374 individuals. An NIL harboring a 9.6-cM segment of IR24 containing the DTH2 locus in the Aso background was isolated from this population. As the NIL is derived by four backcrosses, it is therefore 96.875% isogenic with Aso. The IR24 fragment is flanked by the SSR markers RM318 and RM240 on chromosome 2 (SI Appendix, Fig. S1). To fine-map the DTH2 locus, we crossed Aso with NIL and self-pollinated the F₁ to generate a population of 4,312 F_2:3 families containing 86,240 plants. From the 4,312 F_2:3 families, we identified 1,099 late-heading homozygous families (SI Appendix, Fig. S3) for mapping of DTH2 by using the primers listed in SI Appendix, Table S3.

Sequencing of DTH2 and 20 Genes Flanking DTH2.

To determine whether DTH2 is subject to selection, we amplified and sequenced the 2,275-bp DTH2 coding region and fragments of 20 genes surrounding DTH2 for a core collection of 127 cultivated and wild rice accessions (SI Appendix, Table S1). Sequencing primers of DTH2 and the information for all 20 genes are listed in SI Appendix, Tables S4 and S5, respectively.

Network Analysis and Test for Selective Sweep.

A phylogenetic network was constructed using the median-joining model implemented in network version 4.6 (34). The coalescent simulation was performed to test for a selective sweep in the region surrounding DTH2 (SI Appendix, SI Methods).