Day length and temperature are two major environmental factors that shape plant growth and development and affect the flowering time and photoperiod sensitivity that determines the seasonal and regional adaptation of crops. Rice (Oryza sativa L.) is the staple food for approximately half of the world’s population. Improvements in its productivity are an important way to meet the need of the continuously increasing human population. The high yield potential in rice is usually associated with a long growth time, which is the basic requirement for developing enough vegetative source to support the large sink of the reproductive organ. As a short-day plant, rice flowers more rapidly when the day-length becomes shorter and heading more difficult when the day-length grows longer. Rice is planted worldwide, mostly in Asia area that have a wide latitude range with changing photoperiod conditions, from long-day (LD) conditions to short-day (SD) conditions. This wide expansion of rice throughout complicated geographic environments suggests that the rice genome experiences profound changes, with several key genomic regions being under natural selection for adaption.

The natural variation of genes controlling adaptive traits, such as flowering time, has been extensively studied. In cultivated rice, the variation of florigen Heading date 3a (Hd3a) expression has been found to be highly correlated with the diversification of flowering time1. Hd3a expression might be partly linked to the combination of allelic variation in the Heading Date 1 (HD1) protein and to the divergent expression levels of Early heading date 1 (Ehd1)1. In addition, natural nucleotide variations in several flowering time inhibitors, such as PRR37, Ghd7, and Ghd8/DTH8/LHD1, and flowering time accelerators (Ehd1, Ehd4, Hd3a, OsELF3/Hd17/Ef7, and DTH2) show a strong association with flowering time and contribute to the wide distribution of rice cultivation at various latitudes2,3,4,5,6,7,8,9,10,11,12. However, most of the important natural variations in rice that determine the function of genes are located in the coding region1,10,11,13,14,15. The role of the variation in the gene promoter region that modulates the function to cope with changing environmental conditions is elusive. Given that the flowering genes have the characteristics of accurate spatiotemporal expression and plasticity in response to environmental signals16,17, variations in both the promoter and coding region of a gene might combine in an optimal form to affect the adaption ability, which has currently been rarely reported in rice.

Nuclear factor Y (NF-Y) is a ubiquitous CCAAT-box binding transcription factor formed by NF-YA, NF-YB and NF-YC subunits18,19,20. NF-Y has been reported to regulate downstream gene expression to improve plant tolerance to abiotic and biotic stresses21. NF-Y has also been identified as a flowering time regulator in plants and plays a pivotal role in regulating diverse aspects of plant growth and development22. In particular, the subunits NF-YB have a function on flowering time, which has been well studied. NF-YB is called HAP3 (heme activator protein 3) or CBF-A (CCAAT-binding factor A) in yeast and animals22. Ghd8 was identified as a transcription factor belonging to NF-YB. It was cloned as a major quantitative trait locus (QTL) controlling heading date, plant height and grain productivity in rice. It acts as a flowering inhibitor in LD conditions and an activator in SD conditions8,23,24. Recently, it has been reported that GHD8 regulates flowering time through interacting with HD1 and other NF-Y/HAP family members such as OsHAP2 and OsHAP5. The GHD8-HD1-OsHAP5 complex directly binds the CORE2 element on the 5′UTR region of florigen Hd3a to affect the flowering time25,26. Although the mechanism of Ghd8 in controlling flowering time is well elucidated, the contribution of the cis-regulatory variations of Ghd8 to rice adaptation to the environments such as adverse temperature condition and unfavorable day-length in the changing climate is not clear.

To study whether natural variations of Ghd8 exhibit an important determinant function in rice adaptation, we conducted sequence analysis of the entire Ghd8 gene, including the promoter region and coding region, in a panel of 198 rice germplasm accessions. Several haplotypes in the promoter region and coding region of Ghd8 were observed in this population. Two dominant haplotypes of the Ghd8 promoter had a significant difference in the transcription level. Transgenic tests in rice demonstrated that the expression level of Ghd8 was significantly associated with flowering time and cold tolerance. Moreover, a single nucleotide polymorphism (SNP) in the promoter region of Ghd8 was found to be involved the transcriptional change. Our results demonstrate that the cis-regulatory variation of the Ghd8 promoter modulating its transcription level confers a selective advantage and plays a critical role in adapting rice to a wide range of environments.


Nucleotide variation of Ghd8

Previous studies have shown that several variants occur in the coding region of Ghd8 that alter the function of flowering time or heading date8,13,23. To understand the nucleotide variation of the promoter region and its relationship with the coding region, we sequenced a 3.8-kb region of Ghd8 in a rice collection of 198 accessions (Table S1). The sequence comparison revealed a total of 41 variants identified across the 3.8-kb Ghd8 region, showing extensive nucleotide variations in both the promoter and coding regions of Ghd8 in the rice collection.

Sixteen nucleotide variants were detected in the coding region of Ghd8, which classified the 198 accessions into 10 haplotypes (Fig. 1, Fig. S1). These coding haplotypes were divided into two clades (clade I and clade II) corresponding well to the subspecies indica and japonica. Seven haplotypes, GHD8-4, GHD8-5, GHD8-6, GHD8-7, GHD8-8, GHD8–9 and GHD8–10, were in clade I, and four haplotypes, including GHD8-1, GHD8-2 and GHD8-3, belonged to clade II (Fig. 1). In clade II, 90% of accessions shared the GHD8-1 haplotype23. In parallel, clade I was dominated by three haplotypes GHD8-5, GHD8-7 and GHD8-8, with respective frequencies of 24%, 15% and 18% in the 198 accessions (Table S1, Fig. S1). Notably, some haplotypes, such as GHD8-2, GHD8-3, GHD8-5, and GHD8-8, were reported to be nonfunctional, due to a frame-shift or stop codon occurring that caused a truncated or premature protein (Fig. 1), while GHD8-4, GHD8-6, GHD8-7 and GHD8–9 had a strong function13,23.

Figure 1
figure 1

Phylogeny of the coding and promoter regions, showing two main clusters of the nucleotide variants in the promoter region (left) and coding region of Ghd8 (right). The relatedness of the major haplotypes was inferred by using the UPGMA method. The bootstrap values >50% (500 replicates) are indicated next to the branches. Ten Ghd8 alleles based on the coding region variation are presented in the middle panel. The red and blue dots next to the coding region indicate the functional and loss-of-function alleles, respectively, based on the variation/mutation in the coding region and on previous reports. Open and closed arrowheads represent insertions and deletions, and the number denotes base pairs of the insertions or deletions compared to Ghd8-1. The black and gray vertical lines in the boxes indicate the nonsynonymous and silent SNPs, respectively. ‘STOP’ represents a premature stop codon.

There were 25 variants detected in the promoter region of Ghd8 in the 198 accessions. A phylogenetic analysis of the variants in the promoter region revealed that eight major haplotypes (minor frequency >0.02, or n ≥3) in the rice collection were identified and were also divided into the two clades as well (Fig. 1). Pro3 from clade I (indica type) and Pro13 from clade II (japonica type) were two dominant haplotypes (Table S1), existing in 50% and 24% of the rice collection, respectively.

A strong linkage disequilibrium (p < 0.0001, R2 > 0.9) was observed among the nucleotide variations in the promoter and coding regions of Ghd8. In particular, the strongest linkages existed between the dominant promoter haplotype Pro13 and the coding haplotype GHD8-1, and between Pro3 and GHD8-6 (Fig. 1, Table S2). These results suggest that the nucleotide variations in the promoter affecting the expression might be directly linked to the functional Ghd8 alleles.

Ghd8 transcriptionally regulates flowering time

To investigate whether the transcriptional changes of Ghd8 are highly related to the flowering time, we generated overexpression lines of Ghd8 (OE) with a ubiquitin promoter to drive the Ghd8 allele from Nipponbare (designated as Ghd8-1), which is functional23. The OE lines showed delayed flowering of 5–7 days compared to wild type (WT) Nipponbare under both LD (14.5 h light: 9.5 h dark) (Fig. 2a,c) and SD conditions (9 h light: 15 h dark) (Fig. 2b,d). Consistent with the delayed flowering, the expression of Ghd8 was 10−15 fold higher than that in the WT, and the downstream flowering gene Hd3a was significantly downregulated in the OE lines under both of the photoperiod conditions (Fig. 2e,f).

Figure 2
figure 2

Altered expression of Ghd8 affects rice heading date. (ad) Phenotype performances of wild type (WT) and Ghd8 overexpression line (OE) under LD conditions (14.5 h light: 9.5 h dark) (a) and SD conditions (9 h light: 15 h dark) (b). Arrows indicate the first emerging panicle. Heading date of WT and OE in LD (c) and SD (d) conditions. (e,f) The expression of Ghd8 (e) and Hd3a (f). (gj) Performances of WT and artificial microRNA line (AMI) of Ghd8 in LD (g) and SD (h) conditions. Heading date of WT and AMI plants under LD (i) and SD (j) conditions. (k,1) The expression of Ghd8 (k) and Hd3a (l) in WT and AMI plants. Asterisks ** and * represent significant difference compared to WT by t-test at p < 0.01 and p < 0.05 levels, respectively. Error bars indicate the mean ± SD, with three to five replicates.

In addition, higher expression levels of Ghd8 causing a delayed of flowering time was also found in a T-DNA mutant of Ghd8 in which a 1.4-kb insertion fragment was detected from −170 bp to −1609 bp upstream of ATG in the Ghd8-1 allele (Fig. S2). This insertion led to the overexpression of Ghd8-1 that was 10–13 fold higher in the mutant compared to that in the WT. The overexpressing Ghd8 caused a delayed (3 days) flowering time in the mutant lines (Fig. S2).

We also generated the amiRNA-Ghd8 lines (AMI) in which the expression of Ghd8 was repressed (Fig. 2g–l). The knock-down Ghd8 in the amiRNA lines caused a dual phenotype. Compared to Nipponbare, the flowering time of AMI was advanced by approximately six days under LD conditions but was delayed by three days under SD conditions (Fig. 2g–j). Consistent with the earlier flowering time in LD conditions, the expression of the downstream gene Hd3a was induced in AMI (Fig. 2k,l). However, the expression of Hd3a was reduced in AMI under SD conditions, which matched the delayed flowering time in the short-day conditions (Fig. 2l). These results indicate that the Ghd8-1 is functional allele and its expression levels are highly associated with flowering time in rice.

Genetic variations modulate Ghd8 expression

To test whether natural variations in the promoter contributed to the Ghd8 expression that affected flowering time, we generated two constructs (named as ZN and NN) that harbored the common Ghd8-1 driven by either a truncated promoter Pro3-1186 of Pro3 or a promoter Pro13-1469 of Pro13 (Fig. 3a). Higher expression levels of Ghd8 were displayed in the ZN (Pro3-1186) transgenic lines than in the NN (Pro13-1469) transgenic lines (Fig. 3b,d). Moreover, ZN showed a significantly delayed flowering time compared to that of NN under both LD and SD conditions (Fig. 3c), which matched well with the overexpressing Ghd8 delayed flowering time in the OE lines. In line with the flowering time delay and photosensitivity enhancement (Fig. 3e), reduced expression of downstream genes Ehd1, Hd3a and RFT1 was observed in ZN (Fig. S3). Since both ZN and NN had the same Ghd8-1 allele (Fig. 3a), the differences in flowering times between them should be a result of the functional variations in their promoters. These results indicate that the promoter region of Ghd8 is associated with its differential expression, which plays a role in the flowering time of rice.

Figure 3
figure 3

A negative cis-regulatory region of Ghd8 regulates its expression level. (a) Schematic diagram of the Ghd8-1 allele driven by two types of the Ghd8 promoter, where the sequence variations are indicated. NN and ZN indicate the construct with the promoter type Pro13-1469 from Nipponbare and Pro3-1186 from ZS97, respectively. (b,d) Expression profiles in the NN and ZN transgenic plants under LD and SD conditions. The white and black bars represent light and dark periods, respectively. The zeitgeber times (ZT) on the axis are corresponding to the sampling times. Leaf samples were collected every 4 h from 35-day-old plants. (c) Heading date of the transgenic lines and ZS97 under SD and LD conditions. (e) Photosensitivity of the transgenic lines and ZS97. Different letters indicate significant difference among the lines by the Tukey’s HSD test at p < 0.05. Error bars indicate the mean ± SD with three replicates.

SNP (T1279G) affects Ghd8 expression

To determine which variants in the promoter play a key role in the regulation of Ghd8, we developed two more truncated promoter constructs Pro3-1161 and Pro13-1161 from Pro3-1469 and Pro13-1469, respectively. The expression of the GUS (β-glucuronidase) reporter driven by the truncated promoter Pro3-1161 showed a higher expression level than that driven by Pro3-1469. Likewise, the GUS expression by the truncated promoter Pro13-1161 was much higher than that by Pro13-1469 (Fig. 4a, Fig. S4). A higher expression level of GUS was observed in Pro3-1161 than in Pro13-1161. In contrast, GUS expression in the transgenic line driven by Pro3-1469 was two times lower than that in the Pro13-1469 transgenic line (Fig. 4a, Fig. S4). In addition, the GUS-staining results showed there was no difference in tissue-specific expression between the Pro3::GUS and Pro13::GUS transgenic lines (Fig. S5). There was no significant difference in GUS expression between the transgenic lines with Pro3-1469 and the other truncated promoter, Pro3-1279G (Fig. 4a). These results suggest that the distal region (−1285 to −1161 bp) has a negative cis-regulatory element reducing the expression of Ghd8.

Figure 4
figure 4

Expression of GUS and LUC driven by truncated Ghd8 promoters. (a) Schematic diagram of the reporter gene GUS driven by the truncated promoters of Ghd8 that were used to generate the transgenic GUS lines. Pro13-1161 represents the truncated promoter of Ghd8 from Nipponbare (Pro13-1469). Pro3-1161 and Pro3-1279G indicate the truncated promoters from ZS97 (Pro3-1469). Pro3-1279T indicates a G replacement of T at the −1279 site of Pro3-1279G, where a predicted pyrimidine box is shown. The GUS expressions in 35-day-old seedling leaves of the corresponding transgenic lines, as indicted in the right panel. (b) Schematic diagram of constructs with the truncated promoters Pro3-1161, Pro3-1279G and Pro3-1279T for luciferase assays in rice protoplasts. (c) Relative firefly luciferase activity to Renilla luciferase activity of Pro3-1161, Pro3-1279G and Pro3-1279T. (d) GUS expression of Pro3-1279G and Pro3-1279T transgenic plants under the treatments of 0 mM (as CK) and 20 mM exogenous GA3. ** and * indicate significant differences among the indicated lines by ANOVA at p < 0.01 and p < 0.05 levels, respectively. Error bars indicate the mean ± SD with three replicates.

Remarkably, only the SNP (T1279G) at site −1279 was found within the 125-bp segment of the distal region between Pro3 and Pro13. Transient expression assays with the additional site-mutated construct (Pro3-1279T), in which 1279 G was replaced by 1279 T (Fig. 4b), showed that the expression level of GUS in Pro3-1279T was significantly increased relative to Pro3-1279G and partially recovered up to 50% of the expression level in Pro13-1469 (Fig. 4a). Likewise, Pro3-1279T revealed a significantly higher luciferase activity in the rice protoplasts than that in Pro3-1279G but was significantly lower than that in Pro3-1161, which lacked the 125-bp segment (Fig. 4c). Thus, the variant (e.g., 1279 G) might negatively affect Ghd8 expression regulation.

Ghd8 has a conserved function in Arabidopsis

GHD8 protein in rice shares a highly conserved domain with the early flowering time gene Hap3B in Arabidopsis. Our previous results showed that overexpression of Ghd8, driven by 35 S, significantly accelerated the flowering time in Arabidopsis Col-023. To validate whether the rice promoter normally drives Ghd8 expressed in Arabidopsis, we also introduced the rice constructs ZN and NN into the late-flowering mutant hap3b in Arabidopsis. The transgenic test results showed that both the ZN and NN transgenic lines could positively recover the mutant phenotype in bolting date, flowering time, and rosette and cauline leaf number to the level of wild type Col-0 (Fig. 5a–c). Furthermore, the ZN transgenic lines flowered much earlier than the NN lines. Similar to the expression pattern in rice, much higher Ghd8 expression levels were found in the ZN lines relative to that of NN in Arabidopsis (Fig. 5d). As expected, the downstream florigen gene FT and the flowering integrator SOC were upregulated more profoundly in ZN than in NN (Fig. 5d). These results indicated that the rice Ghd8-1 could rescue the late flowering time in the mutant hap3b, and the functional effect was regulated through the rice promoter. Therefore, Ghd8 has a conserved function to control flowering time in Arabidopsis and rice, and its expression is functionally associated with the flowering time.

Figure 5
figure 5

Ghd8 rescues the late flowering of hap3b mutant in Arabidopsis. (a) Performance of wild-type Col-0, late flowering mutant hap3b, and transgenic plants generated with the NN and ZN constructs (four pots of each column belong to the same genotype) under LD conditions (16 h light: 8 h dark). (b) Bolting time and flowering time for the indicated genotypes as shown in (a). (c) Number of rosette leaves and cauline leaves scored at the maturity stage. At least 10 plants for each indicated genotype were recorded, and the error bars represent the mean ± SE. (d) Expression of Ghd8, FT and SOC in the hap3b mutant, NN and ZN transgenic plants in Arabidopsis under LD conditions. Different letters above the bar indicate significant differences by Tukey’s HSD test at p < 0.05.

High expression of Ghd8 confers cold tolerance in rice

Low temperature is one of the major abiotic stresses limiting rice distribution in specific environments such as the high-latitude regions. To understand whether Ghd8 regulates cold tolerance in rice, we developed two near-isogenic lines (NILs) within the common ZS97 background carrying two contrasting alleles: a nonfunctional allele Ghd8-5 and a functional allele Ghd8-1. The NILs with different functional Ghd8 alleles showed a significant difference in seedling survival rate after 4 °C treatment. NILGhd8-1 displayed an improved cold tolerance, with approximately 30% higher survival rate compared to that of NILGhd8-5 (Fig. 6a,b). Notably, in line with Ghd8-1 affecting the flowering time, the average survival rate of the Ghd8-1 transgenic line was 27% higher than that of ZS97. The average survival rate of the ZN transgenic line with the increased expression level of Ghd8 was 40% higher than that of NN (Fig. 6c,d). Consistently, a high expression level of Ghd8 conferring a strong cold tolerance in rice was also observed in the Ghd8 overexpression line (Fig. S6).

Figure 6
figure 6

A higher expression level of Ghd8 confers better cold tolerance in rice. (a) Difference in seedling survival rate between near-isogenic line (NIL) carrying the Ghd8-1 alleles and Ghd8-5 under cold treatment. (b) The performances of NILGhd8-1 and NILGhd8-5. (c) Survival rate of the ZN and NN transgenic lines and WT (ZS97) under cold stress. (d) The seedling performances of the indicated lines. (e) Comparison of survival rates of the accessions carrying the 1279 T variant and those carrying the 1279 G variant. N represents the number of the assayed rice accessions. (f) Expression level of Ghd8 in the accessions carrying the T or G alleles as shown in (e). The error bars indicate the mean ± SE. Asterisks ** represent significant differences between the indicated lines at p < 0.01 by t-test; and different letters above the bar indicate significant differences by ANOVA at p < 0.05.

To understand how Ghd8 responds to low temperature, we carried out an expression analysis of Ghd8 in cold treatment using the ZN and NN transgenic lines in rice. Quantitative real-time PCR analyses showed that Ghd8 expression in the transgenic lines of ZN and NN were all rapidly induced by six to eight times at 1 h after cold treatment, and ZN showed a continuously higher expression level of Ghd8 than did NN up to to 6 h after cold treatment (Fig. 7a). In addition, several cold responsive genes, such as AOX1b, DREB1a and DREB1b27,28, were detected with a significant upregulation in both ZN and NN under the cold stress (Fig. 7b–d). Consistent with the high survival rate of ZN in the cold treatment, the cold responsive genes in ZN showed a significantly higher expression level than those in NN across all the surveyed time points (Fig. 7b–d). These results indicate that Ghd8 is induced by low temperatures and is associated with cold tolerance through positively upregulating cold-response genes in rice.

Figure 7
figure 7

The expressions of Ghd8 and several cold response genes were induced by cold treatment. (ad) Expression profiles of Ghd8 (a) and three cold response genes, AOX1b (b), DREB1a (c), and DREB1b (d), in the transgenic lines ZN and NN. C0, C1 and C12 represent 0, 1 and 12 h after cold treatment. The error bars indicate the mean ± SD with three replicates.

The association of Ghd8 with cold tolerance was also confirmed by analyzing data from the variant T1279G in natural rice cultivars. The rice accessions with the promoter haplotype carrying T1279G displayed a significant difference in the expression level of Ghd8, in which the 1279 T had a higher expression level than the 1279 G (Fig. 6f). These results support that the variant T1279G located in a negative regulatory element, is key to determining the expression level of Ghd8. Furthermore, the rice varieties with the promoter haplotype carrying 1279 T exhibited a significantly higher cold tolerance than those of the accessions carrying the 1279 G (Fig. 6e,f). Taken together, these results confirm that T1279G is a strong candidate for the causal regulatory variant of Ghd8, leading to differences in the cold tolerance of rice.

Phytohormone GA might mediate Ghd8 expression

To determine which conserved regulatory elements might be associated with the SNP T1279G, we analyzed the sequences within the 125-bp fragment surrounding the SNP using the software PLACE29. It predicted that the fragment contains one potential pyrimidine box with a CCTTTT/G motif that is involved in gibberellic acid (GA) responsiveness30. The SNP (T/G highlighted above) at position −1279 is located at the pyrimidine box (Fig. 4a). To test whether T1279G alters the GA responsiveness, we performed luciferase (LUC) transient assays in rice protoplasts. The assays showed that the relative activity of LUC by the construct Pro3-1279T was significantly higher than Pro3-1279G under the normal condition without GA treatment. Notably, an approximately twofold greater increase of LUC activity was observed in Pro3-1279T relative to Pro3-1279G when treated with 5 μM GA3, but it remained unchanged in Pro3-1279G (Fig. S7). The results imply that the GA response of the regulatory element was abolished when the CCTTTT motif was mutated. Furthermore, quantification of the GUS expression in leaf tissue driven by Pro3-1279T also exhibited a significantly stronger GUS expression level than that by Pro3-1279G, especially after 20 mM GA3 was applied to the GUS transgenic seedlings (Fig. 4d). These results suggest that the variant T1279G, located in the GA-responsive motif, might modulate the expression of Ghd8.

Genome region encompassing Ghd8 in japonica subspecies under selection

To gain insight into the divergent variants in the promoter that are linked to adaptive traits, we compared the nucleotide diversity between cultivated and wild rice and conducted neutrality tests for the sequences across a 100-kb genomic region flanking Ghd8. Diversity analyses revealed that the promoter region exhibited an obvious reduction of nucleotide diversity in both indica and japonica relative to that in the wild rice (Table S3), potentially indicating that this region might be targeted by selection. However, the diversity ratio in the promoter region of Ghd8 between japonica and wild rice is slightly higher than that between indica and wild rice. The neutrality test revealed that an approximately 25-kb genome region surrounding Ghd8 was subjected to selection, with a highly positive Tajima’s D value (average = 2.2) in the japonica group. By contrast, no significant Tajima’s D (average = −0.01) was detected in the same region of indica (Fig. S8). The significant and positive Tajima’s D value suggests that selection occurred in the Ghd8 region, thereby causing prevalent haplotypes such as Pro13 and Ghd8-1 in the japonica group growing in temperate areas.


Previous studies have shown that several variants occur in the coding region of the transcription factor Ghd8 that alter the function on flowering time or heading date8,13,23. Variations in the expression levels of several flowering time genes due to cis-regulatory polymorphisms have been reported in maize and Arabidopsis31,32. In our study, we demonstrate that naturally occurring variants in the promoter region of Ghd8 have effects on flowering time and cold tolerance in rice, which are two adaptive traits important for crop adaption in diverse environments.

One of the findings in the present study is that the functions of Ghd8 are largely dependent on its transcription, with the upregulation of Ghd8 not only delaying flowering time (Figs 2, 3) but also improving cold tolerance (Fig. 6). Ghd8 was identified as a major locus affecting flowering with the dual function: inhibit flowering under LD conditions and promote flowering under SD conditions23,26. In the current study, this dual function of Ghd8 on flowering was also found in the transgenic lines by regulating Ehd1 and Hd3a/RFT1 (Fig. 2, Fig. S3). Intriguingly, a high expression level of Ghd8 enhanced the inhibition of Ghd8 on flowering time by suppressing the expression of the downstream genes Ehd1-Hd3a/RFT1 under both LD and SD conditions (Fig. S3). We further identified the variant T1279G that occurs in the promoter region of Ghd8, regulating Ghd8 transcription, plays an important role in flowering time and cold tolerance in rice (Figs 4, 6). Several lines of evidence suggest that T1279G in the cis-regulatory region of Ghd8 is highly associated with its transcription and contributes to the variations in flowering time and cold tolerance in rice. First, the promoter variations of Ghd8 were clustered into two main haplotypes, which corresponded well to the subspecies indica and japonica. The haplotype japonica (e.g., Pro13) displayed a high expression of Ghd8 relative to the haplotype indica (e.g., Pro3). However, we observed that the transgenic lines with a truncated Pro3 promoter (e.g., Pro3-1161) displayed an unexpected higher expression level of Ghd8 than those with the Pro13 promoter in rice and Arabidopsis, suggesting there is a negative cis-regulatory element in Pro13 repressing the expression of Ghd8. Importantly, the elevated expression level of Ghd8 in rice transgenic lines delayed the flowering time and increased the tolerance to low temperatures (Figs 2, 3, 6). Second, GUS transgenic lines revealed that the promoter lacking the negative cis-regulatory element within the distal region (−1285 to −1161 bp) of Ghd8 induced the gene expression (Fig. 4, Fig. S4). This note was also supported by the transient assays using various truncated promoters in rice protoplasts. Third, site-mutated experiments provided a strong support for T1279G as a causal variant in the negative cis-regulatory element that regulates Ghd8 expression.

Notably, the variant T1279G is located on the potential GA-responsive motif that might contribute to a different response to GA treatment. The 1279 T mutant revealed significantly increased expression compared to the 1279 G variant after the application of GA (Fig. 4d, Fig. S7). It has been reported that the pyrimidine box CCTTTT and the other two motifs, the TAACAAA box and TATCCAC box, together are required for a GA response in transiently transformed barley aleurone layers30,33,34,35. Damage to either of them caused a reduction of promoter expression in GA-treated aleurone layers, but did not completely abolish GA induction. Consistent with Ghd8 being induced by GA, we found that a TATCCAC box (at −644 bp) and a TAACAAA box (at −572 bp) exist in the Ghd8 promoter. Together with the CCTTTT motif at −1279 bp, they likely form a GA-responsive complex, which could play a role in regulating the Ghd8 expression in response to GA. Furthermore, the japonica cultivars carrying the 1279 T variant had a higher expression level of Ghd8 compared to the indica cultivars that harbored the 1279 G variant (Fig. 6f). Therefore, we suggest that low temperatures might activate the GA-related genes that recognize the pyrimidine motif at the Ghd8 promoter to induce the expression of Ghd8, although this requires further experimental verification.

The other finding in our investigation is that regulation of Ghd8 is necessary for rice adaptability. Our results indicate that the divergent variants in the promoter region and coding region of Ghd8 produce functional diversity in flowering time and cold tolerance, which driving rice adaption to specific environments. Haplotype analyses in rice accessions have revealed that Ghd8 has two major haplotypes showing some indica-japonica specificity. One major haplotype, GHD8-1, with function at a high frequency (70.1%) exists in the japonica rice grown in temperate areas or northern latitudes (Fig. S9). The other haplotypes, such as GHD8-5 and GHD8-8 with nonfunctional, occur in 21% and 16% of indica cultivars, respectively, that are widely distributed throughout the subtropical regions. Different geographical distributions of these haplotypes potentially indicate that they might have some ability to adapt to particular environments. A contrasting pattern of distribution was reported for the flowering gene Ghd7; the strong function alleles of Ghd7 tend to be found in the southern parts of Asia, and weaker or nonfunctional alleles appear more frequently in the northern region13. It is believed that the northward expansion of rice into the long-day environments is largely facilitated by early flowering, reduced photoperiod sensitivity and tolerance to low temperature. In terms of cold tolerance, it is an ecologically significant trait for the adaptation of plants to cold climates. However, molecular evidence supporting such a claim is very limited. We found that all of the japonica cultivars from high latitudes or northern regions carry the promoter haplotype with the 1279 T variant accompanied by the Ghd8-1 allele, which has a function on flowering time23. It is particularly interesting that the japonica rice with the 1279 T variant is strongly associated with a high expression of Ghd8-1, increasing cold tolerance. In parallel, we observed that the Ghd8-5 and Ghd8-8 alleles experienced a loss of function and appear more frequently in indica cultivars, which might have occurred later during rice domestication or breeding selection, since they were not detected in the wild progenitors (Fig. S1). Regarding the alleles of Ghd8 with a strong function on flowering time13,23, a lower expression of the functional alleles is preferable in rice if grown at lower latitudes, otherwise a high expression would result in a seriously delayed flowering time that might be harmful to productivity in southern regions. Thus, weak alleles with a moderate effect on photo-sensitivity and a strong effect on cold tolerance would facilitate expanding rice growing to the northern regions. Consistent with this proposal, most indica accessions grown in southern regions carry the 1279 G variant of Ghd8 and confer a lower expression level compared to the japonica rice. The distribution of rice is closely correlated with latitudes might result from the joint functional effects of regulatory and coding variants in the indica-japonica breeding history. Our study provides a novel view of the functional variations in the promoter region of Ghd8, offering a powerful means to fine-tune gene expression for rice adaptability in various environments.

It has been reported that cis-regulatory element modifications are tend to be the targets that undergo selection during plant domestication36,37,38,39. We propose that Ghd8 is a likely target of selection during rice domestication based on three lines of evidence. First, the promoter region exhibits an obvious reduction of nucleotide diversity in cultivated rice relative to wild rice (Table S3). Second, genetic divergence of the Ghd8 promoter is found in the cultivated rice, and a high level of diversity is maintained at the Ghd8 locus in both subspecies. Third, a significant positive Tajima’s D value is detected in the 25-kb genome fragment harboring Ghd8 in the japonica cultivars and indicates a strong haplotype structure appearing in the subspecies japonica (Fig. S8). This reflects that some certain GHD8 protein types are always linked to a specific promoter with the divergent T1279G. As mentioned above, this SNP variant is most likely to be responsible for the expression change. Our results provide a new example supporting the theory that cis-regulatory variations underwent selection for local adaptation during rice domestication.

In conclusion, we reveal that cis-regulatory variation in the Ghd8 promoter, along with the coding region function, is critical for rice to adapt to different regions. The T1279G variant, located on a negative cis-regulatory region of Ghd8, causes a high expression level of Ghd8 in japonica rice and leads to a change in cold tolerance, thus contributing significantly to the ecological adaptation of rice varieties at high latitudes or in northern regions. These findings provide insights into the cis-regulatory variant that functionally affects gene expression associated with adaptive traits in rice.

Materials and Methods

Plant materials and growth conditions

A rice mini-core collection comprised of 198 rice accessions was used in the nucleotide diversity analysis40,41. For function analysis of Ghd8 regulating flowering time, the T-DNA mutant PFG_3A-18183 in cv. Hwayoung background was obtained from POSTECH RISD ( The homozygous T-DNA insertion mutants were confirmed by using border primers on the rice genome and the vector T-DNA (Table S4). The homozygous mutant lines, along with the negative plants were grown in natural LD conditions to score flowering time. We also obtained the Arabidopsis mutant SALK_025666 hap3b from the Nottingham Arabidopsis Stock Centre for function analysis of Ghd8.

The overexpression line of Ghd8 was generated by using the plasmid pCAMBIA-1301, with the Ghd8 coding region from the japonica cultivar Nipponbare driven by a plant Ubiquitin promoter. Artificial microRNA (amiRNA) lines (named amiRNA-Ghd8) were generated, as previously described42. Using the WMD3 tool (, we designed 21-mer small RNA carrying the mismatch sites at Ghd8 to create amiRNA-Ghd8 that specifically silenced the gene expression. The mutant and transgenic lines were grown in growth chambers under LD conditions (14.5 h light: 9.5 h dark). Three transgenic lines for each construct were analyzed for the phenotype of interest. The flowering time and leaf number was scored in Arabidopsis mutant, as previously described43.

Gene expression analyses

Rice leaves were harvested from the 35-day-old seedlings for expression analysis. Rice plants were grown under both natural LD conditions (14.5 h light: 9.5 h dark) and SD conditions (9 h light: 15 h dark). Total RNA was extracted from leaf tissues using TRIzol and was treated with DNase I following the TRIzol reagent protocol (Invitrogen). cDNA (20 µL) was synthesized from 1 µg of RNA using SuperScriptII Reverse Transcriptase (Invitrogen). One microliter of cDNA was used for real-time PCR with SYBR Green PCR master mix (Applied Biosystems). The data were collected using an Applied Biosystems 7500 Real-time PCR System. All expression levels were normalized to that of the ubiquitin gene using the relative quantification method44.

Transgenic analysis for functional variations of Ghd8

To compare the transcription levels caused by variation in the Ghd8 promoter region, two constructs with a binary vector pCAMBIA1301 were generated and transformed into the recipient cv. Zhenshan97B (ZS97). The first construct, named as NN, contained a 3.8-kb fragment from cv. Nipponbare, including 1469-bp of promoter region (Pro13) upstream of the ATG initiation codon of Ghd8, 893-bp of coding region, and approximately 1400-bp of the 3′ UTR23. The second construct, designated as ZN, was created by using a promoter fragment (from −1186 bp to ATG) from ZS97 driving the identical coding region of Ghd8 from Nipponbare. The ZN and NN constructs were also introduced into the Arabidopsis hap3b mutant using the floral dipping method45.

To study whether the Ghd8 promoter variants altered the expression of Ghd8, two main types of promoter: Pro3 (named Pro3-1469) and Pro13 (named Pro13-1469), were respectively isolated from ZS97 and Nipponbare by PCR and inserted into the DX2181 vector that carries a GUS (β-glucuronidase) reporter gene46. These two constructs were transformed into the japonica variety Hejiang19 (HJ19) by the agrobacterium-mediated transformation47. In addition, four truncated promoters from the above Pro3 and Pro13 were amplified by PCR and inserted into the DX2181 vector as well. The truncated Ghd8 promoter constructs were transformed into cv. HJ19 or Zhonghua 11 (ZH11) by the agrobacterium-mediated method.

Luciferase transient expression assay in rice protoplasts

To construct the reporter Pro3-1161::LUC, Pro3-1279G::LUC and Pro3-1279T::LUC, the segments of Ghd8 promoter (Pro3-1161, Pro3-1279G, Pro3-1279T) were amplified by PCR and cloned into the pGreen II 0800-LUC vector using the Hind III and BamH I restriction sites48. Rice protoplast isolation and transfection were based on the method described in a published protocol49. For cotransfection assays, 5 µg of construct were used for each polyethylene glycol transfection. After 12 h of transient transformation, protoplasts were incubated for 2 h with an addition of the gibberellic acid (GA3) at a final concentration of 5 μM at 28 °C under dark conditions for luciferase (LUC) activity measurement. The Renilla luciferase gene (rLUC), driven by the 35 S promoter of Cauliflower mosaic virus (CaMV), was used as an internal control. Firefly and Renilla luciferase activity were quantified using a Dual-Luciferase Reporter Assay System (Promega; No. E1910), according to the manufacturer’s instructions, and chemiluminescence was measured using a TECAN Infinite M200 microplate. Three replicates were performed.

DNA sequence analysis

The promoter region of Ghd8 was amplified by using the ExTaq and GC buffer I (Takara) from genomic DNA of the 198 rice accessions. The purified PCR products were used as templates and sequenced using the Big Dye Terminator v3.1 Cycle sequencing kit (Applied Biosystems). All sequences were assembled using Sequencher v4.5 software and aligned against the Nipponbare sequence of Ghd8 (LOC_Os08g07740.1) downloaded from the website A phylogenetic tree was constructed for various Ghd8 alleles using MEGA software ( The number beside the branches obtained by a bootstrapping approach represents a measure of support for the node, where 100 represents maximal support. Diversity analysis and Tajima’s D analysis were performed with a 2-kb window size based on a previous method50 for the 100-kb region surrounding Ghd8 in 156 japonica and 295 indica varieties, which were downloaded from the website All primers used for PCR and quantitative real-time PCR are listed in Table S4.

Cold tolerance experiment

To test cold stress tolerance, rice accessions, near-isogenic lines (NIL) of Ghd8, and the NN and ZN transgenic lines were used. Thirty-six germinated seeds for each sample were sown in 6 × 6 × 10 (cm) pots and placed in a 30 °C growth chamber for 7 days. They were then transferred into a 4 °C chamber for 18 h for the transgenic lines, or 24 h for the natural rice accessions or NILs. After the cold treatment, the seedlings were returned to a 30 °C growth chamber for recovery. The survival rate was determined after 6 days of recovery growth by counting the percentage of the total seedlings that were alive, as previously described51. The cold tolerance experiment was replicated three times.