Importance of the Interaction between Heading Date Genes Hd1 and Ghd7 for Controlling Yield Traits in Rice

Appropriate flowering time is crucial for successful grain production, which relies on not only the action of individual heading date genes, but also the gene-by-gene interactions. In this study, influences of interaction between Hd1 and Ghd7 on flowering time and yield traits were analyzed using near isogenic lines derived from a cross between indica rice cultivars ZS97 and MY46. In the non-functional ghd7ZS97 background, the functional Hd1ZS97 allele promoted flowering under both the natural short-day (NSD) conditions and natural long-day (NLD) conditions. In the functional Ghd7MY46 background, Hd1ZS97 remained to promote flowering under NSD conditions, but repressed flowering under NLD conditions. For Ghd7, the functional Ghd7MY46 allele repressed flowering under both conditions, which was enhanced in the functional Hd1ZS97 background under NLD conditions. With delayed flowering, spikelet number and grain weight increased under both conditions, but spikelet fertility and panicle number fluctuated. Rice lines carrying non-functional hd1MY46 and functional Ghd7MY46 alleles had the highest grain yield under both conditions. These results indicate that longer growth duration for a larger use of available temperature and light does not always result in higher grain production. An optimum heading date gene combination needs to be carefully selected for maximizing grain yield in rice.


Introduction
Flowering time is a pivotal factor in the adaption of cereals to various ecogeographic environments and agricultural practices, which is controlled by an intricate genetic network.Florigens are at the core of the network, which are encoded by Hd3a and RFT1 in rice [1,2].The expression of Hd3a and RFT1 are regulated by two important pathways mediating by Hd1 and Ehd1, respectively [3].Hd1 has dual functions, which enhances florigen genes expressions under short-day (SD) conditions but inhibits florigen genes expressions under long-day (LD) conditions.The function conversion of Hd1 is related to PhyB, Se5, Ghd7 and Ghd8 [4][5][6][7][8].Function loss of any of these genes attenuates the conversion and maintains Hd1 as an activator under any day-length conditions.Ehd1 activates florigen genes expressions to promote flowering under both the SD and LD conditions [9].Ehd1 likely acts as a signal integrator, and its expression is regulated by many genes [3].Recent studies revealed that Hd1 represses expression of Ehd1 through interaction with Ghd7 or DTH8 [6][7][8].
Flowering time is closely related to the grain yield for crop, owing to its key role in maintaining an appropriate balance between full use of resources and avoidance of environmental stresses.Many heading date (HD) genes were reported to affect yield traits, and their natural variations have been used in rice breeding, such as Ghd7 [10], DTH8/Ghd8 [11,12], Hd1 [13,14], OsPRR37/Ghd7.1/DTH7/Hd2[15][16][17], RFT1 [18] and OsMADS51 [19,20].Abiotic stresses during flowering, such as high temperature, low temperature, and drought, can pose a serious threat to spikelet fertility and consequently induce yield loss.The relationship between HD gene and abiotic stress has been given attention in recent years.The Ehd1-Hd3a/RFT1 pathway responses stress signals mediated by Ghd7 [21], OsABF [22] or OsMADS51 [20].They integrate low temperature, high temperature, and drought signals, respectively, into HD pathway, which induce or repress floral transition to avoid flowering in the stress environments.Moreover, Ghd7 and other four HD genes, including Ghd2 [23], OsHAL3 [24], OsWOX13 [25] and OsJMJ703 [26], were found to be involved in drought or salt tolerance during vegetative phase.
When the pleiotropic effects of individual HD genes on yield traits have become recognized, the role of gene-by-gene interaction remains to be explored.In the present study, influences of Hd1 and Ghd7 on HD and yield traits were analyzed using near isogenic lines (NILs) and NIL-F 2 populations derived from a cross between indica rice cultivars Zhenshan 97 (ZS97) and Milyang 46 (MY46).Our results showed that Hd1 and Ghd7 could independently promote and repress flowering, respectively, whereas the flowering-repressor function of Hd1 under natural long-day (NLD) conditions required functional Ghd7.With delayed flowering, spikelet number and grain weight increased under both natural short-day (NSD) and NLD conditions, but the spikelet fertility and panicle number fluctuated.Rice lines with genotype of hd1Ghd7 produced the highest grain yield under both conditions.
Highly significant effects (p < 0.0001) of Hd1 and Ghd7 on HD were detected in all the three trials (Table 1).In the two trials under NSD conditions (16LS and 17LS), the functional Hd1 ZS97 and Ghd7 MY46 alleles promoted and delayed flowering, respectively, no matter whether its counterpart was functional or non-functional (Figure 2a,b).In 16LS and 17LS, the proportion of phenotypic variance explained (R 2 ) were estimated to be 80.74% and 75.69% for Hd1, and 5.79% and 6.50% for Ghd7, respectively.The interaction between Hd1 and Ghd7 was non-significant in the 17LS trial and significant in the 16LS trial with a small R 2 of 1.30%.Overall, Hd1 and Ghd7 largely act additively in regulating HD under NSD conditions.In the 17HZ trial under NLD conditions, the effects of Hd1, Ghd7 and their interaction were all highly significant (p < 0.0001).The R 2 were estimated to be 3.03% for Hd1, 56.54% for Ghd7, and 16.43% for the interaction between the two genes (Table 1).Compared with NILs having the hd1 MY46 ghd7 ZS97 genotype, those having the Hd1 ZS97 ghd7 ZS97 genotypes flowered earlier by 3.51 d; compared with NILs having the hd1 MY46 Ghd7 ZS97 genotype, those having the Hd1 ZS97 Ghd7 MY46 genotype flowered later by 8.75 d (Figure 2c; Table 2).These indicated that Hd1 regulates flowering dependent on Ghd7 under NLD conditions, and its flowering-repressor activity requires the functional allele of Ghd7.For Ghd7, it delays flowering regardless of genotype of Hd1 but its effect is enhanced by Hd1.HD was longer by 5.24 d in lines of hd1 MY46 Ghd7 MY46 than hd1 MY46 ghd7 ZS97 , whereas it was longer by 17.49 d in lines of Hd1 ZS97 Ghd7 MY46 than of Hd1 ZS97 ghd7 ZS97 (Table 2).

Expressions of Genes Involved in the Photoperiod Pathway
The transcript levels of Hd1, Ghd7, Ehd1, Hd3a and RFT1 at 2 h after sunrise were examined in seven-week-old rice lines in the R1-NIL population grown in the 17LS and 17HZ trials (Figure 3).In the 17LS trial under NSD conditions (Figure 3a), expression of Hd1 and Ghd7 was not affected by each other.The Ehd1 expression was also not affected by either Hd1 or Ghd7.For florigen genes, the expression of Hd3a was 7.87 times larger in lines of Hd1 ZS97 ghd7 ZS97 than hd1 MY46 ghd7 ZS97 , and 12.46 times larger in lines of Hd1 ZS97 Ghd7 MY46 than hd1 MY46 Ghd7 MY46 .These results indicate that Hd1 promotes Hd3a expression regardless of Ghd7 function, which was in accordance with that Hd1 promotes flowering regardless of Ghd7 function under NSD conditions.In addition, Hd1 was also found to promote RFT1 in the Ghd7 background.At the same time, slightly repression of Hd3a by Ghd7 was detected in the hd1 background.These were consistent with the small effect of Ghd7 under NSD conditions.In the 17HZ trial conducted under NLD conditions (Figure 3b), expression of Hd1 was not affected by Ghd7, but Hd1 up-regulated Ghd7 expression.The Ghd7 expression was 2.12 times larger in lines of Hd1 ZS97 Ghd7 MY46 than hd1 ZS97 Ghd7 MY46 .The expression of Ehd1 in lines of Hd1 ZS97 ghd7 ZS97 was 1.24 times as large as that in lines of hd1 MY46 ghd7 ZS97 , but the expression in lines of Hd1 ZS97 Ghd7 MY46 was only 0.42 times as large as that in lines of hd1 MY46 Ghd7 MY46 .These suggest that Hd1 significantly represses Ehd1 expression in the Ghd7 background.For florigen genes, the expressions of Hd3a and RFT1 in lines of Hd1 ZS97 ghd7 ZS97 were 4.86 and 1.55 times as large as that in lines of hd1 MY46 ghd7 ZS97 , indicating Hd1 promotes expressions of florigen genes in the ghd7 background.However, Hd1 was converted to severely repress the florigen gene expressions in the Ghd7 background.The expressions of Hd3a and RFT1 in lines of Hd1 ZS97 Ghd7 MY46 were only 0.07 and 0.32 times as large as those in lines of hd1 MY46 Ghd7 MY46 .In the meantime, significant repression of the Ehd1, Hd3a and RFT1 expressions by Ghd7 were detected in both the Hd1 and hd1 background, and the effect were larger in the Hd1 background.The expressions of the three genes in lines of hd1 MY46 Ghd7 MY46 were 0.77, 0.24 and 0.68 times as large as those in lines of hd1 MY46 ghd7 ZS97 ; and the expressions in lines of Hd1 ZS97 Ghd7 MY46 were 0.26.0.004 and 0.14 times as large as those in lines of Hd1 ZS97 ghd7 ZS97 .These agreed with that flowering-repressor function of Ghd7 could be enhanced by Hd1.

Influence of Hd1 and Ghd7 on Yield Traits and Its Relationship with HD
Grain yield per plant (GY), and five yield components traits including number of panicles per plant (NP), number of spikelets per panicle (NSP), number of grains per panicle (NGP), spikelet fertility (SF), 1000-grain weight (TGW), were measured in the R1-NIL population grown in the 17LS and 17HZ trials.
In the 17LS trial under NSD conditions, Hd1 showed significant effects (p < 0.01) on all the six yield traits except NP; and Ghd7 showed significant influences (p < 0.01) on all the six yield traits except SF and GY (Table 1).Interaction between the two genes were all non-significant at p < 0.01.Relationships between HD and the yield traits were further investigated (Table 2).The lines of Hd1 ZS97 ghd7 ZS97 had the shortest HD, followed by Hd1 ZS97 Ghd7 MY46 , hd1 MY46 ghd7 ZS97 and hd1 MY46 Ghd7 MY46 .Significant differences (p < 0.05) were detected for all the five yield determinants among the four genotypic groups.Three of the traits, NSP, NGP, and TGW, were positively correlated with HD, having correlation coefficients (r) of 0.823, 0.828, and 0.614, respectively (Table S1).Values of these three traits increased with delayed heading.On the other hand, NP and SF were not significantly correlated with HD.For GY, the values increased with delayed flowering among the three genotypic groups having the shortest to third shortest HD, and then remained stable when the HD became longer.Consequently, the two genotypic groups having the longest and second longest HD, hd1 MY46 Ghd7 MY46 and hd1 MY46 ghd7 ZS97 , had little difference on GY.
In the 17HZ trial under NLD conditions, Hd1 showed significant effects only on NP; and Ghd7 showed significant influences on NP, NSP, and TGW (p < 0.0001).Significant interaction between the two genes was detected on TGW (p < 0.001).The interaction acted for increasing the values of the recombinant types, which was in accordance with the epistasis on HD.The HD and six yield traits were also compared among the four homozygous genotype groups (Table 2).The lines of Hd1 ZS97 ghd7 ZS97 had the shortest HD, followed by hd1 MY46 ghd7 ZS97 , hd1 MY46 Ghd7 MY46 and Hd1 ZS97 Ghd7 MY46 .Significant differences (p < 0.05) among the four genotypic groups were detected on four yield determinants, including NP, NGP, NSP, and TGW.Variations of TGW and NSP were positively correlated with HD, having r values of 0.708 and 0.355, respectively (Table S1).The two traits tended to increase with delayed heading.Similar tendency was observed for NGP though it was not significantly correlated with HD.Conversely, NP was negatively correlated with HD (p < 0.05), having r value of −0.670.SF also appeared to decrease with delayed heading though no significant difference was observed.Consequently, the largest value of GY in the four genotypic groups was observed for hd1 MY46 Ghd7 MY46 which had the second longest HD.

Validation of the Influences of Hd1 and Ghd7 on HD and Yield Traits under NLD Conditions
The relationship between Hd1 and Ghd7 was further analyzed using the R2-F 2 population, which was segregated at Hd1 and Ghd7 loci but homozygous at all the remaining 12 cloned flowering QTL mentioned above.The 775 plants of this population were grown in Hangzhou in 2017 under NLD conditions.Significant effects were identified for both genes.The additive effect, dominance effect and R 2 were estimated to be 1.89 d, -0.89 d and 6.4% for Hd1, and 6.04 d, 1.91 d and 59.3% for Ghd7, respectively.The plants were classified into nine genotypic groups based on the Hd1 and Ghd7 alleles, and the HD values were compared (Figure 2d).Hd1 promoted flowering in the ghd7 background, but delayed heading when the genotype of Ghd7 was functional or heterozygous.Ghd7 delayed flowering regardless of the genotype of Hd1 but its effect was enhanced by the functional Hd1 allele.
GY and five yield components traits were also measured in the R2-NIL population.Hd1 showed significant effects on SF (p < 0.0001) and GY (p < 0.01), and Ghd7 exhibited highly significant effects on NSP, NGP, SF and TGW (p < 0.0001) (Table 1).Highly significant epistatic effects of the two genes were detected on all the traits except NGP (p < 0.0001).For NSP and TGW, the interactions acted for increasing the values of the recombinant types, which were consistent with the epistasis on HD.For NP, SF, and GY, the opposite direction was found.The relationships between HD and the yield traits were further analyzed (Table 2).Lines of Hd1 ZS97 ghd7 ZS97 had the shortest HD, followed by hd1 MY46 ghd7 ZS97 , hd1 MY46 Ghd7 MY46 , and Hd1 ZS97 Ghd7 MY46 .Significant differences were detected for all the yield traits among the four genotypic groups.NSP, NGP and TGW were positively correlated with HD (p < 0.05), having r values of 0.806, 0.507 and 0.672, respectively (Table S1).Values of these traits increased with delayed heading.On the other hand, SF and NP were negatively correlated with HD (p < 0.05), having r values of −0.855 and −0.349, respectively.SF decreased with delayed heading; compared with lines having the shortest HD, SF in lines having the third longest, the second longest, and the longest HD decreased by 1.9%, 4.9% and 10.2%, respectively.Similar tendency was observed for NP.Consequently, lines in the hd1 MY46 Ghd7 MY46 genotypic group having the second longest HD produced the highest GY.

Discussion
The bi-functional action of Hd1 has been well recognized, promoting flowering under SD conditions and inhibiting flowering under LD conditions [27].Recent studies revealed that flowering repressing function of Hd1 is dependent on Ghd7 [6,7].In the present study, this relationship between Hd1 and Ghd7 was confirmed.Under NSD conditions, Hd1 always up-regulated expressions of the two florigen genes (Figure 3) and promoted flowering regardless of Ghd7 genotype (Figure 2).Under NLD conditions, Hd1 still promoted flowering (Figure 2) by up-regulating florigen genes in the ghd7 background (Figure 3).In the Ghd7 background, however, Hd1 was found to up-regulate Ghd7, and down-regulate Ehd1 and florigen genes, consequently leading to late flowering.For Ghd7, its flowering-repressor action was observed under both NSD and NLD conditions regardless of Hd1 function.Taken together, our results suggest that Hd1 and Ghd7 could promote and repress flowering independently, whereas flowering-repressor function of Hd1 under LD conditions requires the functional Ghd7.
Among the four homozygous genotypic combinations of Hd1 and Ghd7, the Hd1ghd7 group exhibited the shortest HD under NLD conditions.Compared to Hd1ghd7, heading was delayed by 3.4-4.3d and 7.5-8.7 d in the hd1ghd7 and hd1Ghd7 groups, respectively.Strikingly, HD in the Hd1Ghd7 group was delayed by 16.1-20.9d, owing to the genetic interaction between Hd1 and Ghd7 under NLD condition.This is likely the reason the Hd1Ghd7 genotype was hardly carried by early season indica cultivars grown in middle-lower regions of the Yangtze River and South China regions [28] and japonica cultivars in northeast China [29], where early flowering is essential to ensure sufficient grown period for late season indica cultivars or secure a harvest before cold weather approaches.
It is generally accepted that long growth duration is associated with high-yielding production in rice [29,30], if varieties are harvested before cold weather approaches.A larger number of HD genes were found to have pleiotropic effects on yield traits, and their late-flowering alleles were frequently used to enhance grain yield mainly by increasing spikelet number and partially by increasing grain weight [10][11][12][13][14][15][16][17][18][19][20].As expected, NSP and TGW gradually increased with delayed flowering under both the NSD and NLD conditions in this study.However, SF and NP tended to decrease under NLD conditions when the HD has become relatively long.As a consequence of trade-off among different yield components, rice lines having the hd1Ghd7 genotype which had the second longest HD produced the highest grain yield, rather than the lines having the Hd1Ghd7 genotype which had the longest HD.These results indicate that longer growth duration for a more use of available temperature and light does not always result in higher grain production.
Spikelet sterility is a key determinant of grain yield and frequently used as an indicator for stress tolerance.Two alternative explanations could be given to the decrease of spikelet sterility with delayed flowering.Firstly, alteration of time of flowering causes some loss of seasonal adaptability of rice.Secondly, Ghd7 and Hd1 participate in the stress tolerance of rice.Ghd7 has been found to respond to multiple abiotic stress, such as high temperature, low temperature, and drought.Moreover, overexpression of Ghd7 increases drought sensitivity, whereas knock-down of Ghd7 enhances drought tolerance [21].Our study showed that Ghd7 expression was dramatically up-regulated in the Hd1 background.This may be a reason that caused low SF in lines of Hd1Ghd7.Moreover, alteration of SF by Hd1 was also observed in the ghd7 background (Table 2), suggesting Hd1 could be involved in stress response independently.
Panicle number is generally recognized as an unstable trait among yield traits.Few genes were reported to have pleiotropic effects on flowering time and panicle number [21,31,32].Ghd7 is found to regulate panicle number in a density-dependent manner.It decreases and increases panicle number at normal field condition and low-density conditions, respectively, though it always suppresses flowering time [21].In the NIL populations used in our study, negative correlation between NP and HD was detected in both trials conducted under NLD conditions at normal planting density (Table 2, Table S1).The lines of Hd1Ghd7 with the longest HD always produced the least NP (Table 2), indicating that combination of Hd1 and Ghd7 could cause decrease of panicle number under NLD conditions.
Although late-flowering alleles of flowering genes generally increase spikelet number, their influences on panicle number and spikelet sterility are not necessarily positive.Thus, an optimum HD genes combination needs to be carefully selected for maximizing grain yield in rice.In the present study, lines carried hd1 and Ghd7 alleles from MY46 produced the highest grain yield in both trials conducted in Hangzhou (Table 2) where is in the middle-lower region of the Yangtze River.Among the 14 middle-season indica rice cultivars tested by Wei et al [28], MY46 is one the 10 cultivars having the combination of non-functional hd1 and functional Ghd7.These indicate that this combination could have undergone intensive artificial selection and play a significant role in the adaption of middle-season rice.

Plant Material
Three rice populations segregating at both the Hd1 and Ghd7 loci were used in this study.The developing process was illustrated in Figure 1 and described below.One F 9 plant of ZS97/MY46 was crossed with MY46 for two generations.Two BC 2 F 1 plants which were heterozygous at both the Hd1 and Ghd7 loci were identified and selfed.In one of the two BC 2 F 2 populations produced, a plant which was heterozygous for both the genes was identified and selfed.The resultant BC 2 F 3 population was assayed with functional or closely linked DNA markers for the two genes.A total of 49 plants which were homozygous at Hd1 and/or Ghd7 loci were identified and selfed.One NIL population namely R1-NIL, comprising all the four homozygous genotypic combinations of Hd1 and Ghd7, was constructed.
Another BC 2 F 2 population was advanced to the BC 2 F 4 generation.A BC 2 F 4 plant which was heterozygous for both the genes was identified.In the resultant BC 2 F 5 population, plants which were heterozygous for both the genes were selected and selfed.A NIL-F 2 population in the BC 2 F 6 generation, namely R2-F 2 population, was constructed.A total of 114 plants which were homozygous at Hd1 and/or Ghd7 loci were selected and selfed.One NIL population namely R2-NIL, which consisted of all the four homozygous genotypic groups, was constructed.

Field Experiments and Phenotyping
The rice populations were tested in the experimental stations of the China National Rice Research Institute located at either Hangzhou or Lingshui.During the period of floral transition in the rice materials tested, day length in Hangzhou and Lingshui were corresponding to NLD and NSD conditions, respectively [14].In all the trials, the planting density was 16.7 cm × 26.7 cm.Field management followed the normal agricultural practice.For NIL sets, the experiments followed a randomized complete block design with two replications.In each replication, one line was grown in a single row of ten plants.HD was recorded for each plant.At maturity, five middle plants in each row were harvested in bulk and measured for six yield traits, including NP, NSP, NGP, SF (%), TGW (g) and GY (g).Of which TGW was evaluated using fully filled grain followed the procedure reported by Zhang et al. [33].

DNA Marker Genotyping and Quantitative Real-time PCR Analysis
For population development and QTL mapping, total DNA was extracted using 2 cm-long leaf sample following the method of Zheng et al. [34].PCR amplification was performed according to Chen et al. [35].The products were visualized on 6% non-denaturing polyacrylamide gels using silver staining or on 2% agarose gels using Gelred staining.Three DNA markers were used, including functional marker Si9337 for Hd1, functional marker Se9153 and closely linked marker RM5436 for Ghd7 [10,17].
For expression analysis, penultimate leaves of rice lines in the R1-NIL population were harvested at 7:00 am in 17HZ and 9:00 am in 17LS, 2 h after sunrise.Total RNA was extracted using RNeasy Plus Mini Kit (QIAGEN, Hilden, German).First-strand cDNA was synthesized using ReverTra AceR Kit (Toyobo, Osaka, Japan).Quantitative real-time PCR was performed on Applied Biosystems 7500 using SYBR qPCR Mix Kit (Toyobo, Osaka, Japan) according to the manufacturer's instructions.Actin1 was used as the endogenous control.The data were analyzed according to the 2 -∆Ct method.Three biological replicates and three technical replicates were used.The primers were selected from previous studies [10,20,36].

Data Analysis
For the NIL-F 2 population, QTL analysis was performed with single marker analysis in Windows QTL Cartgrapher 2.5 [37].For the NIL populations, two-way ANOVA was conducted to test the main and epistatic effects.Duncan's multiple range test was used to examine the phenotypic differences among genotypic groups.The analysis was performed using the SAS procedure GLM [38].

Figure 1 .
Figure 1.Development of the rice populations used in this study.

Figure 3 .
Figure 3. Transcript levels of five heading date genes in the R1-NIL population.(a) Under the NSD conditions in Lingshui.(b) Under the NLD conditions in Hangzhou.Data are presented in mean ± s. e. m (n = 3).Bars with different letters are significantly different at p < 0.01 based on Duncan's multiple range tests.

Table 1 .
The effects of Hd1 and Ghd7 on heading date and six yield traits.

Table 2 .
Heading date and six yield traits of the four homozygous genotypes of Hd1 and Ghd7.