Is the yield change due to warming affected by photoperiod sensitivity? Effects of the soybean E4 locus

Abstract

Modeling studies predict that global warming will shorten growth duration and reduce seed yield of early‐maturing soybean (Glycine max) cultivars, but that late‐maturing cultivars could mitigate this reduction. This is widely discussed but has not been validated experimentally. Time of soybean maturation is determined by several photoperiod‐sensitive loci. Here, we focused on the E4 locus, and tested the hypothesis that this locus would mitigate the growth period shortening and yield reduction due to warming. We sowed cv. Enrei with the dominant E4 allele and a near‐isogenic line with recessive e4 allele (NIL‐e4) on two dates (normal vs. late, with a shorter photoperiod) and grown under three temperature regimes (near‐ambient, and 2.0 and 4.6°C above ambient) in sunlit greenhouses in a cool region of Japan. The period from sowing to flowering (R1) decreased with increasing temperature, regardless of genotype and sowing date. However, increased temperature prolonged the period from R1 to the beginning of pod filling (R3) in Enrei but not in NIL‐e4 for either sowing date. This indicates that increasing temperature shortened the period before R1, exposing Enrei with E4 to a longer photoperiod and therefore slowing its development. For both sowing dates, pod number and seed yield increased with warming in Enrei. Since the days from R1 to R3 and the cumulative radiation for this period were positively correlated with pod number in Enrei, the greater yield response was explained mostly by the prolongation of this period caused by warming. However, the yield increase resulted partially from the current mean growing season temperature, which was near or below the optimum for yield. We conclude that the E4 locus can increase seed yield under future warming in cool regions of Japan. This result demonstrates the potential importance of modifying photoperiod sensitivity to increase soybean yield under future warming.

INTRODUCTION

The global mean surface air temperature is expected to rise significantly by 2050 (IPCC, ). Warming is likely to impact on the phenology, growth, and yield of crops around the world. Effective strategies to adapt agricultural production to global warming require a deep understanding of crop responses to increasing temperature. Soybean (Glycine max [L.] Merr.) is a major source of plant protein and oil, and is a major contributor to the world's food supply. The impacts of high temperature on soybean production have been investigated using several approaches. A meta‐analysis of the impacts of high temperature on soybean yields obtained from four different analytical methods (global or local crop models, statistical regression, and warming experiments) demonstrated a nonsignificant decrease in the global average soybean yield of 3.1% per °C increase with large uncertainties (Zhao et al., ). Even among experimental studies, high temperatures have been reported to affect negatively or positively development, growth, and yield of soybean. For instance, Thomas, Boote, Pan, and Allen () investigated the developmental changes of soybean at a wide range of temperature (28/18, 32/22, 36/26, and 40/30°C) and found that the period from emergence to beginning of flowering (R1) decreased with increasing temperatures from 28/18°C to 36/26°C but increased at higher temperatures. In contrast, the period from beginning of seed filling (R5) to beginning of maturity (R7) was shortest at 28/18°C. A review by Hatfield et al. () showed that the optimum temperature for development of soybean shifted from 30°C in the vegetative phase to 26°C in the reproductive phase. Gibson and Mullen () found that an increase in day temperature from 30 to 35°C between R1 and R5 reduced seed number, that between R5 and R8 decreased seed size, and that for the entire reproductive period (R1‐R8) decreased seed yield. Tacarindua, Shiraiwa, Homma, Kumagai, and Sameshima (, ) reported that an increase in temperature during the growing season from 26 to 30°C reduced dry mass production, harvest index, seed number, pod number, and single‐seed size and thereby seed yield. On the other hand, Sionit, Strain, and Flint () showed that an increase in daytime temperature from 18°C to 26°C during the entire growing period increased seed number and yield. Pereira‐Flores et al. () also showed in their open‐top chamber study that a temperature rise of 2.7°C above the ambient air (25.0°C) increased seed yield. These varying effects can be due to differences in the timing, intensity of exposure to elevated temperature, the photoperiod during the growing season, and the cultivar's sensitivity to temperature and photoperiod.

Process‐based crop models are effective tools for assessing the impact of climate change on crop production. Modeling studies have predicted the impacts of high temperature on soybean production in many regions. For example, a statistical model based on long‐term data (1994–2013) across the USA predicted that soybean yields would decrease by around 2.4% for every 1°C increase in the growing‐season temperature, but the response differed greatly among individual states, ranging from −22% in the warmer states to  + 9% in the cooler states (Mourtzinis et al., ). Most crop models parameterize phenology by defining developmental stages that are sensitive to temperature, photoperiod, or both. Phenology affects dry matter accumulation and partitioning, ultimately affecting seed yield (e.g., Boote, Jones, & Hoogenboom, ; Setiyono et al., ; Soltani & Sinclair, ). Soybean is a short‐day plant, and longer photoperiods slow its phenological development. Therefore, the maturity group (MG) of a cultivar and its development rate are strongly determined by a cultivar's photoperiod sensitivity. In the USA, late‐maturing soybean cultivars with a later MG showed greater sensitivity to longer photoperiod, whereas early‐maturing cultivars with an earlier MG are nearly insensitive to photoperiod (e.g., Setiyono et al., ).

The CROPGRO‐Soybean model predicts the highest soybean seed yield at a mean temperature of 22 to 24°C (Boote et al., ). By using this model, Boote () predicted that a projected 2°C increase would cause shorter vegetative growth, smaller leaf area, lower cumulative solar radiation intercepted, lower canopy photosynthesis, and seed yield in early‐MG cultivars that are insensitive to photoperiod in the mid‐western USA, where the current growing season temperature averages 22.5°C. He also predicted that the magnitude of the yield reduction caused by increasing temperature would be smaller in late‐MG cultivars with greater photoperiod sensitivity, suggesting that cultivars with greater photoperiod sensitivity would benefit from future global warming. Hatfield et al. (), using the same model, predicted that a 0.8°C temperature increase would cause a 2.4% decrease in soybean yield in the southern USA (where the current growing season temperature averages 26.7°C), but that the same temperature increase would increase yield in the upper mid‐western USA (where the mean air temperature averages 22.5°C) by approximately 1.7%. Thus, both the magnitude and the direction of soybean's response to warming will depend on the current mean growing season temperature, the photoperiod during the growing season, and the cultivar's photoperiod sensitivity. Since the photoperiod at a given location in a given time is constant, whereas the growing season temperature increases, optimization of a cultivar's photoperiod sensitivity and sowing date could potentially maximize soybean productivity under future warming.

Recently, genetic researchers have identified 11 major loci (E1 to E10 and J) that control flowering time and maturity, and these loci have been characterized at the phenotypic and genetic levels. Except for E6 and J, their dominant alleles increase the time to flowering and maturity in response to long photoperiod (Li et al., and references therein). This suggests that the dominant alleles of most of soybean's photoperiod‐sensitive loci could alleviate the decreases in the growth period and yield associated with increasing temperature. However, to the best of our knowledge, this hypothesis has not been tested. In our previous study, we grew two determinate soybean genotypes with different photoperiod sensitivities (late MG IV, Enrei; early MG II, Yukihomare) during the normal growing season under three temperature regimes in a temperature gradient chamber (TGC) in a cool region of Japan. In this region, the current mean growing‐season temperature ranges from 19.4 to 22.6°C, which is near or below the optimal temperature for yield (Kumagai & Sameshima, ). The responses of seed yield to temperature differed between cultivars with different MGs: Enrei showed increased yield with increasing growth temperature, whereas Yukihomare showed no change. The greater response of Enrei was explained mostly by an increase in the number of pods, which resulted from a longer period from R1 to the beginning of pod filling (R3) caused by the warmer temperature. This slower transition from R1 to R3 in Enrei presumably results from the longer photoperiod and the cultivar's sensitivity to photoperiod when it is sown at conventional dates (the beginning of June). However, we have not yet identified the genetic factors responsible for these differences in phenology and yield. Recent studies showed that various allelic combinations at the E1, E3, and E4 loci affect flowering time, as well as the preflowering and postflowering photoperiod responses, and contribute greatly to soybean's wide adaptability (Jiang et al., ; Tsubokura et al., ; Wolfgang & An, ; Xu et al., ). Xu et al. () suggested that two Phytochrome A genes, E3 and E4, regulate the response of pod filling to photoperiod. Enrei and Yukihomare are different in the E1, E3, and E4 loci; Enrei's genotype is E1/e3/E4 but that of Yukihomare is e1/E3/e4 (Tsubokura et al., ). In this study, however, we focused on the dominant E4 allele and considered that this allele, in conjunction with a longer summer photoperiod, appears to be linked with the slower transition from R1 to R3 in Enrei. To confirm this, we investigated the effects of season‐long warmer temperatures on phenology and seed yield of Enrei and a near‐isogenic line (NIL, E1/e3/e4) with the Enrei genetic background, sown at two different dates (early June and early July) in two consecutive years in TGCs. We showed that the dominant E4 allele prolonged the period from R1 to R3 in higher temperatures and thereby increased the pod number and seed yield compared with the recessive e4 allele under June sowing, despite the total growth duration being shortened by warming. We conclude that the phenological trait of E4 is useful for yield enhancement under future warming in relatively high latitude regions like northern Japan.

MATERIALS AND METHODS

Plant materials and growth conditions

The experiments were conducted at the NARO Tohoku Agricultural Research Center in Morioka, Japan (39°44′N, 141°7′E), from June to October in 2014 and 2015. We grew Enrei, which has the dominant E4 allele (a late‐maturing cultivar in MG IV), and an early‐maturing NIL with the Enrei background but with recessive e4 allele (NIL‐e4). NIL‐e4 was developed as the 6th backcross generation (BC₆) between Enrei and Harosoy‐e4. The allele at the E4 locus was identified as e4‐SORE‐1 in Harosoy‐e4 (Tsubokura et al., ). Both cultivars had a determinate growth habit.

Before sowing, each pot was supplied with 15 g of compound fertilizer, which included 0.45 g of N, 1.5 g of P₂O₅‐equivalent, and 1.5 g of K₂O‐equivalent, and with 20 g of fused magnesium phosphate and 20 g of dolomitic lime, according to standard regional agronomic practices. The soil was a low‐humic Andisol. The pot size was 10 L (300‐mm diameter × 255‐mm height). Each pot contained 8 kg of soil. Seeds were inoculated with Bradyrhizobium japonicum (Mamezou, Tokachi‐Nokyouren, Japan). We sowed three seeds per pot on each day of 4 June and 1 July 2014 and 4 June and 3 July 2015.

After seedling emergence, plants were thinned to one per pot and grown until harvest in two TGCs, each of which was a naturally sunlit greenhouse (6‐m wide, 30‐m long, and 3‐m tall) with an air inlet at one end and exhaust fans at the other end. Okada, Hamasaki, and Sameshima () provide a detailed description. The air in the TGC flowed continuously from the inlet to the exhaust fans. A temperature gradient inside the TGC was continuously maintained along the longitudinal axis by cooling the air with an air conditioner at the inlet end, warming the air by solar radiation or a supplemental heat input (a heater and air ducts) at the outlet end, or both. To create a point where the air temperature was equivalent to the ambient temperature outside the greenhouse, we maintained the air temperature at the inlet end at 2 to 3°C below the outside temperature by cooling the air. We established three temperature regimes along the temperature gradient: low (T1), medium (T2), and high (T3) temperature. The air temperature in T1 was equivalent to that outside the greenhouse. We aimed to create a temperature difference of 4 to 5°C between T1 and T3 by regulating the fans and heater. We used the two TGCs as replicates in both years. We arranged five pots per cultivar in each temperature regime, without mutual shading, and replicated this design in the second TGC. The size of each treatment area was 3 m × 4 m, and pots were rotated at 7‐day intervals to minimize the effects of environmental differences.

Pt‐100 resistance thermometers with an aspirated double‐tube radiation shield were used to monitor air temperature in each regime and periodically moved their position upward to match the plant height. Air temperature was measured every 5 s, and the means were recorded every 30 min for 24 hr on a datalogger (CR1000; Campbell Scientific Inc.). Daily mean solar radiation during the experiments was recorded at the research center's weather station, 1 km NNE of the TGC. Photoperiod (astronomical day length) was calculated according to the method of Horie and Nakagawa (). Plants were irrigated with tap water once or twice a day to maintain soil water near field capacity. Pesticides were applied when necessary.

Phenology and yield

The phenology of four or five plants of each cultivar in each treatment was surveyed once a day; we recorded the date of the beginning of flowering (R1, the date when 50% of the plants had at least one flower at a node on the main stem), the date of the beginning of pod setting (R3, the date when 50% of the plants had at least one 5‐mm‐long pod at one of the four uppermost nodes on the main stem with a fully developed leaf), and the date of the beginning of maturity (R7, the date when 50% of the plants had at least one mature pod on the main stem) according to the growth stage definitions of Fehr and Caviness ().

At harvest maturity, we cut the plants at ground level and separated the aboveground parts of four or five plants per regime in each TGC into leaves, stems, pod shells, and seeds. Simultaneously, we recorded the numbers of nodes on the main stem and branches, of fertile pods, and of seeds per plant. The dry weight of stems, pod shells, and seeds were measured after oven‐drying at 80°C for 5 days to provide the total biomass. Seed yield per plant was adjusted to 15% moisture content, and the mean single‐seed weight was determined by dividing the seed yield per plant by the number of seeds.

Statistical analysis

We conducted analysis of variance (ANOVA) for a split‐split‐plot design, treating temperature as the main factor, sowing date as the split factor, and genotype as the split‐split factor, using the general linear model procedure. We used the two TGCs as replicates in each year but treated this design as if there were four replicates (2 years × two TGCs), because identifying the yield difference between the years was not our focus (i.e., our analysis focused on differences among treatments within a year). As the temperatures were sufficiently similar during the growing season, this approach seems reasonable. We compared the responses of developmental rate (DVR, the reciprocal of days) from R1 to R3 to the two factors (temperature and photoperiod) for the two genotypes by means of analysis of covariance (ANCOVA). First, we used separate‐slope covariance analysis to test for slope differences. Whenever the genotype × factors (temperature or photoperiod) interaction term was not significant, we continued the analysis using a common‐slope model. All statistical procedures were performed in SPSS for Windows v. 22.0 or 23.0 software (IBM, Japan).

RESULTS

Meteorological conditions

In T1, the mean 4‐month temperature from June to September was approximately 19.6°C in both the years 2014 and 2015, which was 1°C lower than the 30‐year mean outside temperature of 20.6°C (Table ). The corresponding mean 4‐month air temperatures were 21.6 in T2 and 24.1°C in T3. Thus, the difference between T1 and T2 was 2.0°C, and that between T1 and T3 was 4.6°C. For each year, the standard error (SE) of the mean 4‐month temperature was small (<0.06°C) in each TGC. As shown in Figure S1, differences in daily maximum and minimum temperature between the temperature treatments were consistently maintained during the growing seasons. The mean 4‐month solar radiation in 2014 was close to that in a normal year but was higher than normal in 2015.

Mean monthly air temperature, solar radiation, and photoperiod during the growth season under three different temperature regimes (T1, low temperature; T2, medium temperature; T3, high temperature) in 2014 and 2015

Parameter	Year	Treatment	June	July	August	September	4‐month mean
Air temperature (°C)	30‐year mean	18.3	21.8	23.4	18.7	20.6
2014	T1	19.0	21.0	21.1	17.2	19.6 ± 0.06
T2	21.1	23.1	23.1	19.1	21.6 ± 0.02 (2.0)
T3	23.8	25.7	25.6	21.5	24.2 ± 0.04 (4.6)
2015	T1	17.8	21.6	21.1	17.5	19.5 ± 0.06
T2	19.9	23.6	23.0	19.4	21.5 ± 0.04 (2.0)
T3	22.5	26.1	25.5	21.9	24.0 ± 0.01 (4.5)
Solar radiation (MJ/m2 per day)	30‐year mean	15.7	14.5	14.9	11.3	14.1
2014		17.1	15.1	15.1	11.8	14.8
2015		19.6	17.8	15.6	13.3	16.6
Photoperiod (hr)			14.8	14.5	13.5	12.2	–

¹Measured outside of the greenhouse.

²Values are the mean ± SE of the two greenhouses.

³Values in parentheses indicate the temperature difference (an increase) compared with the T1 treatment.

⁴Photoperiod (astronomical day length) was calculated according to the method of Horie and Nakagawa ().

Phenology and cumulative solar radiation

Warming reduced the total growth duration (from sowing to R7) and the total cumulative solar radiations over the seasons (p < .001, Figure , Table , Table S1, Figure S2). This was largely due to the reduction in the vegetative growth phase (from sowing to R1, Figure ). The reductions in these durations were similar between two genotypes and sowing dates. Enrei had a longer growth duration than NIL‐e4 (p < .001, Table ), but the difference was less pronounced in July sowing (sowing × genotype interaction, p < .001). On the other hand, the effects of increasing growth temperature on the periods and the cumulative solar radiations of postflowering (from R1 to R3 and from R3 to R7) were inconsistent among the four sowing date and genotype combinations; there were a strong and significant temperature × genotype interaction for the number of days and the cumulative solar radiations from R1 to R3 (p < .001) and significant temperature × sowing date × genotype interactions for the numbers of days and the cumulative solar radiations from R1 to R3 (p < .001) and from R3 to R7 (p < .05). Compared to T1, the period from R1 to R3 was increased by 4 to 6 days in T3 for Enrei, versus 1 to 2 days for NIL‐e4. In addition, the greatest increase of the cumulative solar radiation from R1 to R3 (151 MJ/m²) was found for the June + Enrei combination. The magnitude of the decrease in the period and the cumulative solar radiations from R3 to R7 due to increasing temperature differed among the four combinations: This period and this radiation was 3 to 5 days shorter and 24 to 62 MJ/m² lower, respectively, in T3 than in T1 in the combination of June + NIL‐e4, July + Enrei, and July + NIL‐e4, but no effect of temperature was found for the June + Enrei combination.

Effects of temperature (T1, low; T2, medium; T3, high), sowing date (June or July), and alleles (E4 or e4) on the phenology (days from sowing to R1, from R1 to R3, and from R3 to R7) in Enrei (E4) and a NIL in the Enrei background (NIL‐e4). Phenological stages: R1, beginning of flowering; R3, beginning of pod filling; R7, beginning of maturity

ANOVA results of the effects of temperature, sowing date, genotype (E4/e4 alleles), and their interactions on the phenology of Enrei (E4) and a NIL in the Enrei background (e4). Phenological stages: R1, beginning of flowering; R3, beginning of pod filling; R7, beginning of maturity

Sources of variation	Duration (phenological stages)
Sowing–R7	Sowing–R1	R1–R3	R3–R7
Temperature (T)	***	***	***	***
Sowing date (S)	***	***	***	***
Genotype (G)	***	***	***	***
T × S	ns	***	**	**
S × G	***	***	***	***
T × G	ns	ns	***	ns
T × S × G	ns	ns	***	*

Abbreviation: ns, not significant.

Statistical significance:

^***p < .001;

^**p < .01;

^*p < .05.

We further analyzed the effects of temperature and photoperiod on the developmental rate from R1 to R3 (DVR, which is the reciprocal of the number of days from R1 to R3) (Figure ). When we combined the data for the two sowing dates, we found that DVR decreased with increasing temperature in both genotypes (Figure a, p < .001). On the other hand, the pooled data showed a significant negative correlation between photoperiod and DVR only for Enrei (Figure b, p < .001). ANCOVA showed that the slopes of the relationship between the mean temperature and DVR differed significantly between the two genotypes (p < .05), whereas the slopes of the relationship between photoperiod and DVR did not. These results suggest that the prolonged period from R1 to R3 in Enrei caused by higher growth temperature was regulated by the E4 locus.

Relationships between (a) the mean growing‐season temperature and the development rate (DVR, which equals the reciprocal of the duration of a period in days) from R1 to R3 (from the beginning of flowering to the beginning of pod filling), and (b) between the mean photoperiod and DVR from R1 to R3. Values are for Enrei (E4) and a NIL in the Enrei background (e4). Each data point is the mean ± SE of four replicates. In these relationships, each genotype was regressed for all sowing dates combined. Statistical significance: ***p < .001 for each sowing date and genotype combination (n = 24). There were no significant differences between the slopes indicating the relationships of temperature with DVR in each genotype (ANCOVA; p ＝ .130). However, the slopes indicating the relationship of photoperiod with DVR for each genotype differed significantly (ANCOVA; p < .05)

Yield and yield components

Seed yield and pod number increased with increasing temperature (p < .001, Table ) and decreased with late sowing (p < .001), and Enrei outyielded NIL‐e4 in all treatments (p < .001), but the magnitude of the increase caused by increasing temperature differed between the genotypes and among the sowing date + genotype combinations, as evidenced by the strong sowing date × genotype, temperature × genotype, and temperature × sowing type × genotype interactions for seed yield and pod number (p < .001, Table ). In Enrei, seed yield and pod number both increased with increasing temperature in the June sowing (by 26% and 25%, respectively, for T2, and by 41% and 44% for T3, compared with T1). In the July sowing, the corresponding increases were 18% and 22% for T2, and 32% and 54% for T3, compared with T1. On the other hand, seed yield and pod number in NIL‐e4 increased with increasing temperature in the July sowing (by 8% and 6% for T2, and by 31% and 50% for T3, compared with T1), but either did not increase greatly or decreased slightly in the June sowing.

Effects of temperature (T1, low; T2, medium; T3, high), sowing date (June or July), and E4/e4 alleles on seed yield, pod number, pods per node, and total node number of Enrei (E4) and a NIL in the Enrei background (e4)

Sowing date (S)	Genotype (G)	Seed yield (g/plant)	Pod number	Pods per node	Total node number
T1	T2	T3	T1	T2	T3	T1	T2	T3	T1	T2	T3
June	Enrei (E4)	74.0	92.9 (26)	104.7 (41)	107.2	134.1 (25)	154.4 (44)	1.31	1.55 (19)	1.94 (48)	82.9	86.7 (5)	80.4 (−3)
NIL‐e4	63.4	60.1 (−5)	57.7 (−9)	85.1	86.9 (2)	92.1 (8)	1.72	1.82 (6)	2.07 (20)	50.0	48.6 (−3)	44.7 (−11)
July	Enrei (E4)	35.6	42.1 (18)	47.1 (32)	50.7	61.8 (22)	78.0 (54)	1.37	1.39 (2)	1.90 (39)	38.3	45.2 (18)	41.9 (9)
NIL‐e4	32.1	34.6 (8)	42.2 (31)	46.7	49.5 (6)	70.0 (50)	1.40	1.56 (11)	2.06 (47)	33.9	32.2 (−5)	34.2 (1)
ANOVA	Source of variation
Temperature (T)		***			***			***			*
Sowing date (S)		***			***			***			***
Genotype (G)		***			***			***			***
T × S		ns			ns			*			*
S × G		***			***			*			***
T × G		***			***			ns			*
T × S × G		***			***			*			ns

Note:

Values are the means (n = 4, based on 2 houses × 2 years).

Abbreviation: ns, not significant.

¹²Values in parentheses indicate the % change compared with the T1 value: positive, increase; negative, decrease.

¹³Statistical significance:

^***p < .001;

^**p < .01;

^*p < .05.

Among the yield components examined, growth temperature had positive effects on the number of pods per node (p < .001, Table ) and total node number (p < .05). Late sowing decreased greatly total node number (p < .001) but increased the number of pods per node (p < .001). The increase in the number of pods per node with increasing temperature was lower in the June + NIL‐e4 combination (by 20% for T3, compared with T1) than in the other combinations (by 39% to 48% for T3 compared with T1), resulting in a significant temperature × sowing date × genotype interaction (p < .05, Table ). For total node number, there was no significant temperature × sowing date × genotype interaction. The main temperature effect was a decrease in the number of seeds per pod (p < .001, Table S2) and a decrease in the single‐seed mass (p < .01), but there were no significant sowing date, genotype, or interaction effects.

Relationships among seed yield, yield attributes, and phenological durations and cumulative solar radiation

Seed yield increased with pod number for both sowing dates for Enrei combinations and for the July + NIL‐e4 combination (p < .001, Figure S5a), but not for the June + NIL‐e4 combination. However, no relationship was found between pod number per plant and total node number for any combination of sowing date and genotype (Figure S5b). Pod number per plant increased with the number of pods per node for all combinations of sowing date and genotype (p < .05, Figure S5c).

We also analyzed the effects of temperature during the reproductive stage on the yield components (Figure ). The number of pods per node was significantly and positively linearly related to the mean temperature from R1 to R7 for each combination of sowing date and genotype (Figure , p < .001). Although ANCOVA detected no differences among the slopes of the four sowing date + genotype combinations, the slope of the June + NIL‐e4 combination was somewhat smaller than those of the other combinations. It is also worth noting that the pod number per plant increased with increase in the number of days from R1 to R3 in the June + Enrei, July + Enrei, and July + NIL‐e4 combinations (Figure ; p < .01) but not in the June + NIL‐e4 combination. Furthermore, pod number per plant increased with increase of the cumulative solar radiation from R1 to R3 in the June + Enrei only (Figure S3). These indicate that longer periods and higher cumulative radiation from R1 to R3 in Enrei contributed to the greater yield increase caused by increasing growth temperature.

Relationships between the mean temperature from the beginning of flowering (R1) to the beginning of maturity (R7) and the number of pods per node for the four combinations of sowing date (June or July) and genotype (E4/e4). Values are for Enrei (E4) and a NIL in the Enrei background (e4). Phenological stages: R1, beginning of flowering; R7, beginning of maturity. Each data point is the mean ± SE of four replicates. Statistical significance for each sowing date and genotype combination (n = 12): ***p < .001, **p < .01, *p < .05. There were no significant differences between the slopes for each sowing date + genotype combination (ANCOVA; p = .3)

Relationships between the number of days from R1 (beginning of flowering) to R3 (beginning of pod filling) and pod number per plant. Values are for Enrei (E4) and a NIL in the Enrei background (e4). Each data point is the mean ± SE of four replicates. Statistical significance for each sowing date + genotype combination (n = 12): ***p < .001, **p < .01. There were no significant differences between the slopes for each sowing date + genotype combination (ANCOVA; p = .184)

DISCUSSION

Some modeling studies showed that increasing the growth temperature would shorten the total growth duration and thereby reduce seed yield in early‐MG cultivars with less photoperiod sensitivity, but that this reduction could be alleviated in late‐MG cultivars with greater photoperiod sensitivity (e.g., Boote, ). However, since modeling studies generally have large uncertainties in predictions of how phenology and yield will respond to projected future warming (e.g., Migliavacca et al., ), it is necessary to validate their predictions experimentally. Here, we tested the hypothesis that alleles at photoperiod‐sensitivity loci of soybean might mitigate the shortening of the growth period and yield reduction caused by warming. Specifically, we investigated whether the soybean E4 locus could affect the changes of phenology and yield caused by increasing temperature for plants sown on two widely separated sowing dates (June and July, thus different photoperiods) in a cool climate in northern Japan.

As the modeling studies predicted, the higher temperatures decreased the total growth duration (number of days from sowing to R7), regardless of genotype and sowing date (Table and Figure ). On the other hand, we observed significant effects of temperature, genotype, and their interactions on seed yield and yield components (pod number per plant or per node and node number). We also observed that the magnitude of the increases in seed yield and pod number in the photoperiod‐sensitive Enrei caused by increasing temperature differed among the sowing date + genotype combinations: the increased growth temperature increased the seed yield and pod number in Enrei under both the June and July sowing dates and in NIL‐e4 under the July sowing, but not in NIL‐e4 under the June sowing. The increase for June + Enrei was consistent with our previous study (Kumagai & Sameshima, ). Although these results contradict model predictions of the yield reduction due to warming, the increases of seed yield and pod number we observed can be explained by the longer reproductive duration (days from R1 to R3) at the higher growth temperature as discussed later.

The durations of the reproductive phases are critical for the determination of soybean seed yield. For instance, soybean pod and seed numbers increased when the reproductive phase was lengthened by artificial daylength extension (Kantolic & Slafer, ). Rowntree et al. () also demonstrated that the seed yield enhancement caused by early planting was associated with a decrease in the duration of the vegetative phase and an extension of the reproductive phase. Our previous study also showed that increased seed yield at high temperatures in Enrei under the June sowing involved increased numbers of open flowers and pods, which resulted from extension of the flowering period (the days from R1 to last flowering) and the R1 to R3 period (Kumagai & Sameshima, ). The present study showed that the R1 to R3 and R3 to R7 periods responded differently to increasing temperature: the R1 to R3 period increased in Enrei but not in NIL‐e4, regardless of the sowing date, whereas the R3 to R7 period decreased in all treatments except for June + Enrei. Our analysis showed that pod number increased with the duration of the R1 to R3 period in all combinations except June + NIL‐e4 (Figure ). When the data were combined with our previous study (Kumagai & Sameshima, ), there was a positive linear relationship between pod number per plant and the duration of the R1 to R3 period in Enrei sown at June (Figure S6). These findings suggested that for Enrei, at least, extension of the period from R1 to R3 contributed to the increased pod number, which in turn increased the seed yield. It was widely accepted that the critical period for determination of pods and seeds in soybean is R1 to the beginning of seed filling (R5) period (e.g., Van Roekel, Purcell, & Salmerón, ). Nakano et al. () showed that the duration from R1 to R5 also increased by 8 to 10 days with increasing growth temperature in Enrei at the same TGC experiment. Although we did not investigate the last flowering stage and R5 in this study, it is likely that the greater increase of seed yield and pod number per plant at high temperatures in June + Enrei can be attributed to a longer flowering duration and R1 to R5 period.

There is ample evidence that a longer photoperiod slows the postflowering development (from R1 to R7) in a wide range of soybean cultivars (Kantolic & Slafer, ; Kumudini, Pallikonda, & Steele, ; Summerfield, Asumadu, Ellis, & Qi, ). Recently, Nico, Miralles, and Kantolic () revealed that the R1 to R3 phase was more sensitive to longer photoperiod than the R3 to R7 phase of soybean. In crop growth models, the DVR of the R1 to R7 period was a linear‐plateau function of photoperiod under optimal temperatures (Boote et al., ; Soltani & Sinclair, ). It is generally accepted that the optimum temperature for development is lower in the reproductive phase than in the vegetative phase (Boote et al., ; Setiyono et al., ), but the optimum temperature for developmental rates during the R1‐R3 period is not well documented. This study performed linear regression for the relationship between the DVR from R1 to R3 and the mean temperature and photoperiod for each genotype (Figure ). It was interesting to note that although DVR of two genotypes decreased linearly with increasing temperature in similar way, DVR responded linearly to increasing photoperiod in Enrei only. Since our control was based on the ambient temperature and the experiment was run under natural photoperiod, we could not clearly separate the effects of temperature and photoperiod from the temperature treatments and sowing dates. The relationship between DVR and photoperiod for NIL‐e4 within each sowing date is seemingly negative (Figure b) but is confounded by the effect of temperature and/or a temperature by photoperiod interaction. However, comparison of DVR between sowing dates in the respective temperature treatment clearly indicates that a difference in photoperiod between sowing dates does not have any effect on DVR for NIL‐e4. Xu et al. () reported that the magnitude of the extension of the reproductive period (which they evaluated as the number of days from R1 to R8) caused by artificially extended photoperiod was greater in the e1/e3/E4 genotypes than in the e1/e3/e4 genotypes in northern Japan. One simple and possible explanation for this differential response of the length of R1 to R3 to increasing temperature could be genotype‐specific photoperiod sensitivity. In our study, higher temperature prolonged the R1 to R3 period and increased pod number and seed yield in Enrei sown in July (a shorter photoperiod than for the June‐sown plants). Thus, we conclude that the photoperiod sensitivity at the E4 locus could be an effective option for increasing soybean yield under future warming with a wide range of sowing dates in the cool climate of northern Japan.

The increased seed yield of soybean due to longer postflowering photoperiod has been previously associated with increase of cumulative solar radiation because longer photoperiod extended the duration of postflowering phase (Nico, Miralles, & Kantolic, , ). This has been considered an indirect photoperiodic effect, mediated by the capture of resources that affect crop growth. Likewise, this study showed that temperature, sowing dates (photoperiod) and E4 locus affected vegetative and reproductive developments (Figure ), and thereby cumulative solar radiations during different stages (Figure S2). The greater extension of days from R1 to R3 caused by higher temperature resulted in the greater increase of cumulative solar radiation, which was positively related to pod number in June + Enrei (Figure S3) and thereby seed yield. The cumulative solar radiation intercepted by soybean is determined by the radiation capture efficiency. Our previous study showed that leaf area per plant at the full leaf expansion stage, which is related to the radiation capture efficiency, increased with increasing growth temperature in Enrei sown at beginning of June (Kumagai & Sameshima, ). Therefore, it is likely that the indirect effect associated with increase of cumulative solar radiation caused by the prolongation of days from R1 to R3 in presence of dominant E4 alleles under higher temperature and longer photoperiod is as the result of the greater improvements of seed yield and pod number per plant in June + Enrei.

The effects of temperature on soybean yield are complex, as they are determined by the growth and partition as well as phenological development. Furthermore, all these responses had different ranges of optimum temperature. We considered that the different responses of yield to the sowing date + genotype combinations resulted partially from the current mean growing season temperature at our experimental site, which was near or below the optimum for soybean yield and yield components. Soybean yield is determined by the number of pods per plant, the number of seeds per pod, and the single‐seed mass. Previous controlled‐environment studies with temperatures similar to those in our study (mean values ranging from 19.6 to 25.4°C from sowing to R7 and from 18.4 to 26.0°C from R1 to R7) have shown that seed yield and yield components are both affected by increasing temperature. For instance, Sionit et al. () showed that an increase in temperature from 22/16°C (day/night) to 26/19°C during the entire growing period increased the number of pods and seed yield. The number of seeds per pod was the yield component least affected by temperature (Baker, Allen, Boote, Jones, & Jones, ). Single‐seed mass was insensitive to increases in season‐long temperatures from 18/12 to 26/20°C (Sionit et al., ), but it decreased as temperature increased above 26/20°C (Baker et al., ). Our regression analysis showed that seed yield was strongly explained by the number of pods per plant (Figure S5a). The number of seeds per pod and the single‐seed mass decreased slightly with increasing temperatures in all treatments (Table S1). Therefore, the increases in seed yield compared with T1 in the T2 and T3 treatments for all combinations except June + NIL‐e4 can be attributed to the increased number of pods per plant rather than the increased number of seeds per pod or single‐seed mass.

In general, the number of pods per plant is determined by the product of the total node number and the number of pods per node. In a study with day/night temperature treatments of 18/14, 22/18, 26/22, and 30/26°C, Thomas and Raper () found that the number of pods per node increased with increasing temperature, except at the highest temperature. Van Shaik and Probst () showed that increasing mean temperature from 15.6 to 32.2°C increased the number of pods per node for the soybean cultivar Midwest but did not increase for the cultivar Clark. We found that the number of pods per plant could be explained by the number of pods per node, but not by the total node number, for all treatment combinations (Figure S5b,c). Moreover, the June + NIL‐e4 treatment showed the smallest increase in pods per node in response to temperature increases (Table ). We also observed linear increases in the number of pods per node with increasing mean temperature from R1 to R7 in all treatment combinations (Figure ). The slope of this relationship was somewhat lower in the June + NIL‐e4 combination, although there was no significant difference from the other slopes. We considered the response of pods per node to higher temperature was directly determined by the number of aborted and abscised flowers, as well as by the pod set ratio. Borthwick and Parker () reported that flower initiation was inhibited above 32°C. During flowering and pod setting, increasing both day and night temperatures from 20/8°C to 30/20°C improved the pod set ratio in tropical and temperate soybean cultivars (Lawn & Hume, ). In contrast, temperatures at 38/28°C severely reduced the pod set ratio as a result of lower pollen viability and pollen germination rates (Djanaguiraman, Prasad, & Schapaugh, ). These observations suggest that increasing temperatures up to 30/20°C favored greater soybean pod set. Genetic variation in the responses of the traits that contribute to pod setting, namely pollen germination and pollen tube growth, to extremely high temperatures (38/30°C) has been documented in determinate and indeterminate soybean cultivars (Salem, Kakani, Koti, & Reddy, ). However, there has been no report on the direct effects of the E4 allele on pod set. Therefore, it is unlikely that there is a difference in the temperature response of the number of pods per node per se between the genotypes. In our study, the June + NIL‐e4 combination had a higher range of mean temperature from R1 to R7 (21.0 to 26.0°C) than the other combinations (18.4 to 24.6°C) (Figure ) and could reach the optimal temperature range for soybean and slightly increase pod set in the T3 regime. This is one possible explanation for why the June + NIL‐e4 combination had the smallest increase in the number of pods per node. Pod number per node also increased with the duration of the R1 to R3 period in all combinations except June + NIL‐e4 (Figure S4). Nico, Mantese, Miralles, and Kantolic () reported that artificial photoperiod extension increased pods per node on the main stems, by increasing pods and flowers on lateral racemes and extending the flowering period. Therefore, another possible explanation for the difference in the temperature response of pod number at the node level was the indirect effect such as the phenological change caused by temperature, sowing date (photoperiod), and E4/e4 alleles in this study.

Although our case study focused on northern Japan, our results will help researchers in other countries in their research on the impact of global warming on soybean productivity by encouraging them to take advantage of genetic information on photoperiod sensitivity. The cultivated area of soybean includes a wide diversity of environments between the equator and 50°N or even higher. The American MG classification system was released in 1970 and was based on the theoretical response to photoperiod or latitude. MG zones represent the defined areas where a cultivar is best adapted, and range from 000 for the very‐early‐maturing cultivars to X for the latest‐maturing cultivars (Zhang et al., and references therein). Recently, Mourtzinis and Conley () used yield data from trials conducted between 2005 and 2015 at 312 locations across the USA to show that the current MG zones were defined by a downward deflection of MG lines in contrast to the 1970s. They have shown that that increased growing season temperatures caused by global warming could have allowed earlier planting of late‐MG cultivars in northern regions, and suggested that the effect of temperature, driven by climate change, is also important for determining the optimal MG for a given location. This suggestion leads to the hypothesis that moving the cultivation zones for soybean cultivars poleward could help farmers increase soybean yield under future global warming.

Since many researchers in Japan, China, Brazil, India, and Europe have adopted the American MG system and classified their local cultivars into American MGs (Liu et al., and references therein), the poleward migration should be validated in many regions in the future. A simulation study using a crop phenological model would be appropriate for this purpose. Some recent studies have deepened the genetic and molecular understanding of the variation of MGs in diverse soybean panel data (Jiang et al., ; Wolfgang & An, ). However, it remains to be seen whether the empirical relationships established for estimating temperature and photoperiod responses will hold for cultivars with a wider range of MG grown in multiple environments and whether the relationships can be linked to the loci (alleles) that control flowering and maturity in soybean (Cober, Curtis, Stewart, & Morrison, ; Stewart, Cober, & Bernard, ). Crop models that can simulate the influence of genetic information on temperature and photoperiod responses would be useful for identifying the optimal MGs for breeding programs as well as for optimal scheduling of sowing as global warming proceeds.

Although our study highlights the importance of the E4 locus for increasing soybean yield under future global warming in cool climates, some additional points should be considered. First, the loci for quantitative traits are generally influenced by the plant's genetic background. The positive effect of E4 on the warming responses of phenology and yield should be tested in additional genotypic backgrounds in future studies. Second, several loci that control flowering time and maturity have been identified in soybean; the E1, E3, and E4 loci have especially strong effects on postflowering photoperiod responses (Xu et al., ). This is interesting because the genotypes at these loci differ among cultivars from northern Japan (Tsubokura et al., ). In fact, Enrei's genotype is different from Yukihomare not only in the E4 locus but also in the E1 and E3 loci. It is possible that the other loci such as E1 and E3 are involved in the different yield responses to temperature between the two cultivars. Therefore, further investigation of the effects of these allelic combinations on temperature and photoperiodic responses in cool regions of northern Japan will be needed. Third, we used determinate soybean cultivars which are predominantly used in Japan, but indeterminate cultivars may have different responses to environment. An earlier work suggested that indeterminate soybean cultivars are potentially more resilient to environmental stresses after flowering than determinate cultivars, as indeterminate cultivars can recover from occasional growth restrictions if environmental conditions allow (Fehr, Caviness, & Vorst, ). Another work using indeterminate and determinate cultivars showed that the reproductive development differed in response to planting dates (Kumagai, ; Wilcox & Frankenberger, ). However, to our knowledge, no studies have compared the responses to temperature and photoperiod between determinate and indeterminate types under the temperature‐controlled condition. Further studies are required to clarify the roles of Dt1 and Dt2 loci, which control growth habit (e.g., Bernard, ), combined with the major maturity loci, in response to high temperature and photoperiod of soybean. Finally, this study investigated the response of soybean to high temperatures in the enclosed chambers, but ideally the response of crops to environment should be investigated in the open field condition. Further research on the roles of maturity loci in the temperature response of field‐grown soybean should be conducted using methodologies such as infrared heating array systems (e.g., Siebers et al., ).

ACKNOWLEDGMENTS

This work was supported in part by the Ministry of Agriculture, Forestry and Fisheries, Japan, through a research project entitled “Development of Technologies for Mitigation and Adaptation to Climate Change in Agriculture, Forestry and Fisheries.” We are greatly indebted to K. Segawa, A. Umihata, E. Kumagai, F. Saitou, and H. Tamura for their technical assistance.

CONFLICTS OF INTEREST

The authors declare no conflict of interest.