agronomy Evolutionary Variation of Accumulative Day Length and Accumulative Active Temperature Required for Growth Periods in Global Soybeans

: Soybean ( Glycine max (L.) Merr.) is a typical short-day and thermophilic crop. This study aimed to reveal the required accumulative day length (ADL) and accumulative active temperature (AAT) for DSF (days of sowing to ﬂowering) and DFM (days of ﬂowering to maturity) in global soybeans. A sample consisted of 354 varieties from 27 countries in ﬁve geographic regions, which were tested in Nanjing, China in two spring-sowing and two summer-sowing seasons. The ADL and AAT were calculated from the climatological data provided by the Public Service of Nanjing Meteorological Bureau. The results showed that the average DSF and DFM of global soybeans were 41.0 d and 83.3 d, which required ADL DSF of 606.6 d · h and AAT DSF of 1185.9 d · ◦ C, ADL DFM of 1126.7 d · h and AAT DFM of 2145.1 d · ◦ C, respectively, all with a wide variation among/within geographic and MG(maturity-group)-set subpopulations. From the multiple regression of DSF and DFM on required ADL and AAT, the ADL, AAT and ADL × AAT contributed 38.5%, 44.79% and 17.10% to DSF variation and 86.98%, 11.42% and 0.54% to DFM variation, respectively, and their relative importance to DSF and DFM varied among the geographic and MG subpopulations. The geographic subpopulations matched only partially with the genomic marker clusters, indicating multiple genetic sources of each subpopulation and that genetic exchange happened among subpopulations.


Introduction
Soybean (Glycine max (L.) Merr.) is a typical short-day and thermophilic crop, and its growth and development are affected by photo-thermal conditions [1][2][3]. Garner [4] discovered the phenomenon of photoperiodism and proposed that photoperiod is an important environmental factor that affects the growth period of soybean. It was shown that temperature is also one of the key factors affecting soybean flowering and maturity [5]. Thus, the two factors, photoperiod and temperature, affect flowering and the development of soybeans, especially, affect the adaptability of soybean varieties to geographic regions, which limits the extension of elite varieties to broad areas.

Measurement of Growth-Period Traits
According to Fehr and Caviness [28], the emergence date (Ve), first flowering date (R1) and maturity date (R8) were recorded for all the tested materials. From the data, DSF and DFM were calculated for further analysis.
Daily day length and temperature data were obtained from the Public Service of Nanjing Meteorological Bureau. The daily maximum and minimum temperature were used to calculate the daily average temperature, from which the daily active temperature was calculated as the accumulated daily average temperature over all the days except those less than 10 • C (in fact, all daily average temperature values were more than 10 • C in the present study). At the same time, the daily sunrise time and sunset time were obtained, from which the daily day length was calculated from their difference (Supplementary Figure S1).
The accumulative day length and accumulative active temperature for DSF and DFM were obtained from the summation of daily day length and daily active temperature during the respective periods, which were designated as ADL DSF and AAT DSF, and ADL DFM and AAT DFM , respectively.

Statistical Analysis
The DSF, DFM, ADL DSF , ADL DFM , AAT DSF and AAT DFM for the 4 environments' 8 blocks were calculated using SAS/STAT 9.4 software package (SAS Institute Inc., Cary, NC, USA). PROC UNIVARIATE was used to perform descriptive statistical analysis on the 6 traits, with their significance among subpopulations tested. PROC GLM was used to perform an analysis of variance (ANOVA) and multiple regression analysis. The linear model of ANOVA is: where µ is the population mean, g k is the effect of the kth genotype, t i is the effect of the ith environment, r j(i) is the jth block effect in the ith environment, (gt) ik represents the interaction effect between ith genotype and jth environment and ε ijk is the residual. The heritability was estimated as [29]: where σ 2 g is the genotypic variance, σ 2 ge is the variance of genotype-by-environment interaction, σ 2 ε is the error variance, n is the number of environments and r is the number of replications. The genetic coefficient of variation is calculated as GCV (%) = σ g /µ × 100%. The multiple regression model was: where y i , x 1i and x 2i are BLUP values of ith genotype for a growth period trait, a is the intercept, b 1 , b 2 and b 12 are corresponding regression coefficients and ε i is the random residual following N(0, σ 2 ). Here, x 1i and x 2i represent ADL and AAT values, whereas x 1i x 2i represents the interaction or correlation between x 1 and x 2 , assuming the three terms are independent of each other. The phenotypic variation of DSF or DFM explained by required ADL, AAT and their product was estimated as: where R 2 is the coefficient of determination of the multiple regression model, S 1 is the sum of squares of ADL, S 2 is the sum of squares of AAT and S 12 is the sum of squares of the product of ADL and AAT.

SNP Genotyping and Clustering Analysis
Restriction-site-associated DNA sequencing (RAD-seq) was conducted at BGI Tech, Shenzhen, China for genotyping the materials, which has been reported in Liu et al. [9,30]. A total of 97,706 SNPs were finally obtained for the 354 varieties and then divided into SNPLDB (SNP linkage disequilibrium block) genomic markers using the software RTM-GWAS [31,32] under linkage disequilibrium D > 0.7 criterion. The SNPLDB marker is characterized by multiple haplotypes, which is especially appropriate for the germplasm population. Finally, 15,954 SNPLDB markers, each with 2~13 haplotypes, were identified.
A genetic similarity coefficient matrix was constructed with SNPLDB data using the corresponding program in RTM-GWAS (https://github.com/njau-sri/RTM-GWAS, accessed on 16 June 2021 [33]). Based on the genetic similarity coefficient matrix, a neighborjoining cluster analysis of the global soybeans was conducted using MEGA 7.0 software (Mega Limited, Auckland, New Zealand) [34].

Wide Variation of ADL and AAT Required for DSF and DTM in Global Soybeans
Under field environment conditions, it is presumed that the experienced ADL and AAT are essential for the appropriate growth period. The DSF and DFM with their required ADL and AAT are shown in Figure 1, which showed a linear positive correlation for DSF on ADL DSF and AAT DSF , and for DFM on ADL DFM and AAT DFM . This indicated that more ADL and AAT are required for longer DSF and DFM. The analysis of variance for DSF, ADL DSF , AAT DSF , DFM, ADL DFM and AAT DFM showed that significant variations existed among genotypes, environments and genotype-by-environment interactions (GEIs), respectively. In particular, variations among environments and GEIs were mainly due to the different sowing seasons (spring-sowing and summer-sowing) involved (Supplementary Table S3).
AAT are essential for the appropriate growth period. The DSF and DFM with their required ADL and AAT are shown in Figure 1, which showed a linear positive correlation for DSF on ADLDSF and AATDSF, and for DFM on ADLDFM and AATDFM. This indicated that more ADL and AAT are required for longer DSF and DFM. The analysis of variance for DSF, ADLDSF, AATDSF, DFM, ADLDFM and AATDFM showed that significant variations existed among genotypes, environments and genotype-by-environment interactions (GEIs), respectively. In particular, variations among environments and GEIs were mainly due to the different sowing seasons (spring-sowing and summer-sowing) involved (Supplementary Table S3). Scatter plots of ADLDFM (blue) and AATDFM (red) for DFM of geographic subpopulations and MG subpopulations in WSGP. "Total" represents the total 354 varieties in WSGP. "O" represents the varieties from the center of origin in both Huang-Huai River Valleys (HCHN) and Chang-jiang River Valleys and its south (SCHN). "A" represents the varieties from Northeast China (NCHN), Far-East Russia (RUFE) and southern Sweden (SSWE); "B" represents the varieties from the Korean peninsula (KORP) and Japan islands (JPAN); "C" represents the varieties from Southeast Asia (SEAS), South Asia (SASI) and Africa (AFRI); "D" represents the varieties from northern North America (NNAM), southern North America (SNAM) and Central and South America (CSAM). "E" represents the early MG-set varieties (MG 000~0); "P" represents the primary MG-set varieties (MG I~VII); "L" represents the late MG-set varieties (MG VIII~X). The same is true for the latter figure.
The present study concentrated on the main effect of ADL and AAT on DSF and DFM, leaving the environment and GEI for future studies. The average DSF of the global soybeans was 41.0 d (32.1% of the entire growth period, i.e., days from sowing to maturity) with a wide range of 20.6-95. 5   Scatter plots of ADL DFM (blue) and AAT DFM (red) for DFM of geographic subpopulations and MG subpopulations in WSGP. "Total" represents the total 354 varieties in WSGP. "O" represents the varieties from the center of origin in both Huang-Huai River Valleys (HCHN) and Chang-jiang River Valleys and its south (SCHN). "A" represents the varieties from Northeast China (NCHN), Far-East Russia (RUFE) and southern Sweden (SSWE); "B" represents the varieties from the Korean peninsula (KORP) and Japan islands (JPAN); "C" represents the varieties from Southeast Asia (SEAS), South Asia (SASI) and Africa (AFRI); "D" represents the varieties from northern North America (NNAM), southern North America (SNAM) and Central and South America (CSAM). "E" represents the early MG-set varieties (MG 000~0); "P" represents the primary MG-set varieties (MG I~VII); "L" represents the late MG-set varieties (MG VIII~X). The same is true for the latter figure.
The present study concentrated on the main effect of ADL and AAT on DSF and DFM, leaving the environment and GEI for future studies. The average DSF of the global soybeans was 41.0 d (32.1% of the entire growth period, i.e., days from sowing to maturity) with a wide range of 20.6-95.5 d and high heritability (h 2 95.7%). The global soybeans required an average ADL DSF of 606.6 d·h (day·hour) for flowering (with h 2 95.8%), which was about 34.1% of the ADL for the entire growth period. ADL DSF varied greatly among global soybeans with a wide range of 311.7-1382.3 d·h. The global soybeans required an average AAT DSF of 1185.9 d· • C for flowering (with h 2 99.2%), which was about 34.8% of the AAT for entire growth period. AAT DSF varied also greatly with a range of 622.6-2726.2 d· • C ( Table 1).
The average DFM of the global soybeans was 83.  (Table 1). Thus, the reproductive growth period needs about two times the ADL and AAT of the vegetative growth period on average. However, depending on geographical origin and maturity group, the ADL and AAT of DSF and DFM varied greatly among global soybeans, as global soybeans adapted to their respective local day length and temperature conditions. The maximums of DSF, ADL DSF and AAT DSF were 4.64, 4.43 and 4.38 times of their minimums, whereas the maximums of DFM, ADL DFM and AAT DFM were 2.52, 2.28 and 2.20 times their minimums, respectively. This suggested that the vegetative growth stage (DSF) is more sensitive to environmental resources, whereas the reproductive stage (DFM) is more stable for its required environmental resources.

Evolutionary Changes from the Center of Origin to Various Geographic Regions in ADL and AAT Required for Growth-Period Traits in Global Soybeans
From the center of origin O (HCHN and NCHN), soybean was disseminated northward to A (NCHN, RUFE and SSWE), and the average DSF decreased from 41.9 d to 26 In summary, DSF, ADL DSF and AAT DSF decreased significantly in the northward dissemination of soybean from the center of origin but increased significantly in the southward dissemination. DFM, ADL DFM and AAT DFM decreased significantly also in the northward dissemination, but no significant change was observed for ADL DFM and AAT DFM in the southward dissemination. In the eastward and westward dissemination, the reductions in or changes of ADL and AAT for both DSF and DFM were not as large as the former two dissemination regions. Among the four dissemination routes, D was the most recent one but with the widest variation in DSF and DFM along with their variation of required ADL and AAT, indicating a very wide extension and a fruitful breeding effort, especially extending and adapting to South America. The relationship between DSF/DFM and ADL/DFM can also be observed roughly in Figure 1, in which the linear gradients for DSF was generally higher than those for DFM and both varied among subpopulations for DSF and DFM, respectively.
In this study, the percentage of ADL DSF (PDL DSF ) and the percentage of ADL DFM (PDL DFM ) to the total ADL required for the entire growth period were used to evaluate the relative amounts of ADL for DSF and DFM, respectively. The relative amounts of AAT for DSF and DFM were defined in the same way and referred as PAT DSF and PAT DFM , respectively. The PDL DSF in O, B and D (34.2%, 32.9% and 32.2%, respectively) was at about the same level but was significantly lower in A (29.9%) and significantly higher in C (47.7%) than that in O, B and D. The PAT DSF was similar to PDL DSF and was also at about the same level in O, B and D (34.5%, 33.2% and 32.8%, respectively) but lower in A (30.3%) and higher in C (49.8%). The PDL DFM was also at about the same level for O, B and D (65.8%, 67.1% and 67.8%, respectively) but was significantly lower in A (70.1%) and significantly higher in C (52.3%) than in O, B and D. The PAT DFM was also similar to PDL DFM , with O, B and D at the same level (65.5%, 66.8% and 67.2%, respectively) and lower in A (69.7%) and higher in C (50.2%) ( Table 1). Therefore, the patterns of required ADL and AAT were different among the geographic regions as well as between DSF and DFM.

Evolutionary Changes from the Primary MG Set to Early and Late MG Sets in ADL and AAT Required for Growth-Period Traits in Global Soybeans
As MGs are mainly determined by geographic regions in addition to local sowing seasons, evolutionary changes in ADL and AAT for MG sets are related to those for geographic changes. The primary maturity groups of soybeans in the center of origin were MG I~VII, but after its dissemination northward and southward and adaption to  to 1099.7 d·h (ranging in 895.5~1487.4 d·h), which was not significant either, and AAT DFM reduced significantly to 1920.6 d· • C (ranging in 1437.0~2700.0 d· • C) ( Table 2).
The DSF, ADL DSF and AAT DSF all decreased significantly from the primary MG set to the emerged-early MG set, and increased significantly in the emerged-late MG set. The DFM, ADL DFM and AAT DFM decreased significantly from the primary MG set to the emerged-early MG set and decreased significantly from the primary MG set to the emerged-late MG set. That indicated that the emerged-late MG set reduced ADL and AAT for DFM in comparison to the primary and early MG sets.
The PDL DSF and PAT DSF of the primary MG set were 32.8% and 33.2%, whereas PDL DFM and PAT DFM were 67.2% and 66.8%, respectively. The PDL DSF and PAT DSF , as well as PDL DFM and PAT DFM , of the emerged-early MG set were similar to those of the primary MG set (30.5% and 31.4%, as well as 69.5% and 68.5%, respectively). However, the PDL DSF and PAT DSF of the emerged late MG set increased to 50.7% and 53.6%; accordingly, PDL DFM and PAT DFM decreased to 49.3% and 46.4%, respectively. Therefore, in the late MG set, the required day length and temperature resources for DSF increased, but the required day length and temperature resources for DFM decreased ( Table 2).

Relative Importance of ADL, AAT and ADL×AAT in Determining Growth Periods of Geographic and MG Subpopulations
Although the role of day length in DSF or transferring from vegetative growth to flowering development has been reported, there has been very little research on the role of day length in DFM or the reproductive period. Here, multiple regression analysis was used to evaluate the relative importance of ADL, AAT and ADL × AAT to DSF and DFM. In multiple regression of DSF or DFM on ADL and AAT, the contribution of the independent variables (ADL and AAT) to the dependent variable (DSF or DFM) can be estimated to include three parts, i.e., contribution from ADL(x 1 ), from AAT(x 2 ) and the correlation/interaction between ADL and AAT (ADL × AAT or x 1 x 2 ). The regression model is as Equation (3), where x 1 x 2 represents the correlation or interaction between x 1 and x 2 , but assuming x 1 , x 2 and x 1 x 2 are independent from each other [35].
In the whole population, the ADL DSF , AAT DSF and (ADL × AAT) DSF contributed 38.05%, 44.79% and 17.10% variation to DSF, which indicated that, in addition to ADL DSF , AAT DSF was somewhat more important to DSF variation, and their interaction (ADL × AAT) DSF accounts for a certain part to DSF variation ( Table 3).
The relative contributions of ADL, AAT and (ADL × AAT) to DSF and DFM varied among the geographic subpopulations. In O, A, B, C and D, DSF was composed of 29.33%, 67.51%, 84.79%, 0.54% and 44.66% contribution from ADL DSF , and 15.38%, 23.65%, 0.23%, 97.56% and 19.92% contribution from AAT DSF , and 55.25%, 8.55%, 14.89%, 1.66% and 35.37% contribution from (ADL × AAT) DSF , respectively. Therefore, in O and D, DSF was characterized with higher contribution from (ADL × AAT) DSF interaction or joint ADL-AAT contribution (55.23% and 35.37%, respectively). In A and B, DSF was characterized with higher contribution from ADL DSF (67.51% and 84.79%, respectively). In C, DSF was characterized with higher contribution from AAT DSF . Thus for DSF, ADL DSF is most important for A and B subpopulations, and AAT DSF is most important for C subpopulations, whereas (ADL × AAT) DSF is more important for O and D subpopulations, or the importance of accumulative day length ADL DSF varied among geographic subpopulations. In other words, during the evolutionary process, from O to the subregions, the major contributor to DSF changed from (ADL × AAT) DSF to ADL DSF in A, B and D, and to AAT DSF in the C subregion (Table 3, Supplementary Tables S4 and S5).
The regression of DFM on ADL DFM and AAT DFM in each geographic group showed that in O, A, B, C and D, DFM was composed of 88.69%, 1.57%, 71.87%, 93.24% and 82.30% contribution from ADL DFM , 9.57%, 34.44%, 27.43%, 4.25% and 16.26% contribution from AAT DFM , and 0.52%, 62.30%, 0.07%, 1.07% and 0.97% contribution from (ADL × AAT) DFM , respectively. Therefore, in O, B, C and D, DFM was characterized with higher contribution from ADL DSF , whereas only in A was DFM characterized with higher contribution from (ADL × AAT) DFM and AAT DFM (62.30% and 34.44%). Thus, for DFM or turning from flowering to maturity, ADL DFM is most important for all subregions except A, whereas (ADL × AAT) DFM and AAT DFM are the most important for the A subpopulation. Therefore, during the evolutionary process, from O to the subregions, the major contributor to DFM remained as ADL DFM in O, B, C and D, except A, in which it changed to (ADL × AAT) DFM (Table 3, Supplementary Tables S4 and S5).  For MG sets, the regression of DSF on ADL DSF and AAT DSF showed that AAT DSF contributed the most to DSF in the primary MG set (54.30%) and the emerged-late MG set (84.74%), whereas ADL DSF contributed the most to DSF in the emerged-early MG set (71.35%). The regression of DFM on ADLDSF and AAT DSF showed that ADL DFM contributed the most to DFM in the primary MG set (78.64%) and the emerged-late MG set (93.98%), whereas (ADL × AAT) DFM and ADL DFM contributed majorly to the emergedearly MG set (47.25% and 39.08%, respectively). Thus, the AAT was important to DSF in primary and emerged-late MG sets, but the ADL was most important to DSF in the emerged-early MG set. The ADL was most important to DFM in primary and emergedlate MG sets, but (ADL × AAT) DFM and ADL DFM were more important to DFM in the emerged-early MG set (Table 3, Supplementary Tables S4 and S5).

Genetic Clustering of the Global Soybeans and ADL and AAT Variation among and within the Clusters
According to He et al. [33], the 97,706 SNPs were grouped into 15,954 SNPLDB markers each with 2~13 haplotypes/alleles. Neighbor-Joining tree analysis based on SNPLDB markers showed that the 354 varieties may be grouped into eight clusters (Figure 2, Table 4). Among these clusters, clusters i~iv are minor ones consisting of only 1 to 10 varieties as neighboring clusters of geographic regions O, A, B and C, mainly in the primary MG set. Clusters v~viii are the major ones, consisting of 104, 17, 52 and 156 varieties, respectively. In cluster v, the varieties were mainly from the center of origin O (52) and southward route region C (37), containing mainly primary MG set varieties (79) and emerged-late MG set varieties (25), indicating O and C having a close genetic relationship. In clusters vi and vii, the varieties are mainly from northward and eastward route regions A and B, including a major part from the primary MG set and some from the emerged-early MG set. Cluster viii was the largest one, consisting of 156 varieties, mainly from northward and westward route regions, A (33) and D (112), indicating that A and D have a close genetic relationship, including a major part of varieties belonging to the primary MG set; both emerged early (17) and the late MG set (5) was involved. Here, the varieties in region A have most of their genetic sources from the center of origin O, but A has separated from O and is becoming close to B and D. Therefore, in current global soybeans, there are two major cluster groups: one is cluster v, containing mainly O and C varieties, and the other is clusters vi, vii and viii, containing mainly A, B and D varieties. All the emerged-early and -late MG sets are included in clusters v~viii. The genetic clustering was not parallel to geographic differentiation, and the ADL and AAT required for DSF and DFM on average did not vary obviously among the eight clusters, with no significant difference in ADL DFM and AAT DFM , respectively. to geographic differentiation, and the ADL and AAT required for DSF and DFM on average did not vary obviously among the eight clusters, with no significant difference in ADLDFM and AATDFM, respectively. and Chang-jiang River Valleys and its south (SCHN). "A" represents the varieties from Northeast China (NCHN), Far-East Russia (RUFE) and southern Sweden (SSWE); "B" represents the varieties from the Korean peninsula (KORP) and Japan islands (JPAN); "C" represents the varieties from Southeast Asia (SEAS), South Asia (SASI) and Africa (AFRI); "D" represents the varieties from northern North America (NNAM), southern North America (SNAM) and Central and South America (CSAM). "E" represents the early MG set varieties (MG 000~0); "P" represents the primary MG set varieties (MG I~VII); "L" represents the late MG set varieties (MG VIII~X).  Total  O  A  B  C  D  E  P  L  i  1  1  1  ii  4  2  4  10  10  iii  7  1  8  8  iv  1  5  6  6  v  52  3  8  37  4  79  25  104  vi  8  8  1  17  17  vii  1  21  26  3  1  14 38 52 Chang-jiang River Valleys and its south (SCHN). "A" represents the varieties from Northeast China (NCHN), Far-East Russia (RUFE) and southern Sweden (SSWE); "B" represents the varieties from the Korean peninsula (KORP) and Japan islands (JPAN); "C" represents the varieties from Southeast Asia (SEAS), South Asia (SASI) and Africa (AFRI); "D" represents the varieties from northern North America (NNAM), southern North America (SNAM) and Central and South America (CSAM). "E" represents the early MG set varieties (MG 000~0); "P" represents the primary MG set varieties (MG I~VII); "L" represents the late MG set varieties (MG VIII~X).

Geographic Subpopulation MG-Set Subpopulation
Here, the two major clusters were inspected for their ADL and AAT required ( Supplementary Table S6). The soybeans in cluster v showed that ADL DSF was 794.9 d·h (ranging 399.0~1382.0 d·h), and AAT DSF was 1552.1 d· • C (ranging 805.0~2726.0 d· • C). The ADL DFM was 1152.8 d·h (ranging 834.0~1479.4 d·h) and AAT DFM was 2160.4 d· • C (ranging 1437.0~2733.0 d· • C). The soybeans in cluster viii showed that ADL DSF was 536.0 d·h (ranging 312.0~1099.0 d·h), and AAT DSF was 1049.8 d· • C (ranging 628.0~2203.0 d· • C), whereas ADL DFM was 1138.1 d·h (ranging 731.0~1478.4 d·h) and AAT DFM was 2175.0 d· • C (ranging 1419.0~2755.0 d· • C). The required ADL and AAT in Cluster v and viii had their ranges extended more than that of their component regions. In summary, the inconsistency between geographic subpopulations and genomic clusters indicated the existence of multiple genetic sources in each subpopulation and genetic exchange happened already among the subpopulations. The results showed both wide ranges of DSF and DFM in global soybeans with an average DSF of 41.0 d and DFM of 83.3 d, respectively. The average ADLs required for DSF and DFM were 606.6 d·h and 1126.7 d·h, and the average AATs were 1185.9 d· • C and 2145.1 d· • C, respectively. About 34.1% and 65.9% of the total ADL(d·h) were required for vegetative and reproductive growth, whereas 34.8% and 65.2% were required for AAT, respectively. In northward dissemination from the center of origin, both DSF and DFM with their required ADL and AAT were reduced; this was consistent with previous studies on the vegetative period and reproductive period of northern spring soybean, shortened with the delay of phenological period [36,37], but in southward dissemination, DSF with its ADL and AAT increased, whereas DFM with its ADL and AAT did not change or changed slightly less. However, no obvious changes were found in eastward and westward dissemination. In comparison with the primary MG set, the late MG set increased their requirements of ADL for DSF but decreased their requirement of ADL and AAT for DFM.
To extend soybeans to further high or low latitude, breeding for earliness and long juvenile period are required, respectively. In global soybeans, the entire growth period (DSF + DFM) of the earliest variety N27294 was 70. ADLs of 2869.7 d·h and 5481.4 d· • C might also finish a life cycle. The latter required almost 3 times that of the former. Thus, it implies that using ADL and AAT might achieve further progress than using DSF and DFM in breeding for earlier emergence and/or longer juvenile period soybeans.
In addition to the merit of using ADL and AAT in studying the variety's requirements for and functions of eco-factors in the geographic dissemination process and breeding for phenological traits of soybean, it might also benefit the understanding and utilization of the properties of phenological traits in cultivation. Tan et al. [38] reported an investigation of the long-term soybean phenology data and climate-related data collected at 51 stations across China from 1992 to 2018, which showed that the growth-period traits varied along with the climate changes. Their results indicated that the average temperature (0.34 ± 0.09 • C/decade) increased, but cumulative sunshine hours decreased (−33.98 ± 1.05 h/decade); the vegetative growth period shortened (−0.52 ± 0.24 d/decade), but the reproductive growth period slightly extended (0.05 ± 0.26 days/decade); and the trends in soybean key growth periods diverged among regions. It seems that this study has noticed the influence of climate changes (average temperature and cumulative sunshine hours) on growth periods, which are similar to our ADL and AAT, but it focused on the influence of environmental changes on growthperiod traits, whereas our study emphasized on the required ADL and AAT for different regional soybean varieties. Since day length and temperature are the main environmental factors affecting soybean growth period varied among geographic regions [1][2][3][4][5], soybeans from different eco-regions have different responses to climate changes. Combining ours and Tan et al.'s results, it is suggested that growth periods might be predicted for future soybean production, utilizing the required ADL and AAT information of global varieties and predicted climate changes (predicted ADL and AAT) in different regional sites.

The Relative Importance of ADL, AAT and ADL×AAT in Determining Growth Period Traits and the Understanding of ADL × AAT Function
The present study also featured the detection of ADL × AAT effect based on separating DSF and DFM into their components of ADL and AAT, for which multiple regression was used to estimate the ADL, AAT and ADL × AAT effects and their relative contribution to explore their relative importance in the adaptation of soybeans to various geo-regions. The results showed that, in global soybeans, the contribution patterns of ADL, AAT and ADL × AAT were different for DSF and DFM variation. For DSF variation, AAT contributed the most (44.79%), whereas ADL and ADL × AAT contributed 38.50% and 17.10% variation, respectively. For DFM variation, ADL contributed the most (86.98%) with AAT and ADL × AAT contributed 11.42% and 0.54% variation, respectively. During the evolutionary dissemination, from O to the geo-regions, the major contributor for DSF variation changed from ADL × AAT to ADL in A, B and D and AAT in the C sub-region, whereas for DFM, its major contributor remained as ADL in O, B, C and D, but was changed to ADL × AAT in A. For DSF variation in different MGs, AAT was important to DSF variation in primary and late MG sets, but ADL was most important to DSF variation in the early MG set. For DFM variation in different MGs, ADL was most important to DFM variation in primary and late MG sets, but ADL × AAT and ADL were more important to DFM variation in the early MG set.
The contribution of ADL and AAT to DSF or DFM are easy to understand independently, but the contribution of ADL × AAT to DSF or DFM needs to be explained. It means ADL and AAT joint contribution or their correlated/interacted contribution, a positive ADL × AAT indicates ADL and AAT contributing to DSF or DFM in the same direction while a negative ADL × AAT indicates ADL and AAT contributing to DSF or DFM in an opposite direction. For example, in the global population, ADL, AAT and ADL × AAT contributed 38.05%, 44.79% and 17.10% to DSF variation, respectively, in which 17.10% DSF variation was due to ADL × AAT. Here, the regression coefficient of (ADL × AAT) DSF was negative, which means that, in this contribution, the ADL and AAT have a negative correlation or interaction, contributing to DSF or DFM in an opposite direction. In Supplementary  Table S4, for DSF, all the regression coefficients of (ADL × AAT) DSF , except that of the early MG set (E), were positive, and all the regression coefficients of (ADL × AAT) DFM , except those of the A sub-region and E MG set, were negative. In addition, all the regression coefficients of ADL and AAT were positive for DSF, and all the coefficients of ADL for DFM were positive, but all the coefficients of AAT DFM were negative, except the A sub-region and E MG set. That means the AAT and ADL × AAT performed different patterns for DSF and DFM. The above understanding is based on a statistical concept, and further study on their biological mechanism is needed.

Responses of Growth Periods to ADL and AAT in Genotypic Clusters
Global soybeans were genetically clustered to evaluate whether the genetic differentiation and geographic differentiation were consistent and to characterize their response of growth periods to ADL and AAT. The global soybeans were grouped into four major and four minor clusters: in the former (clusters v~viii), O and C were mainly located in cluster v, A and B mainly in cluster vi and vii and A and D mainly in cluster viii, whereas in the latter (clusters i~iv), there were scattered varieties from various geo-regions. The ADL and AAT required for DSF and DFM varied slightly among the eight clusters, with no significant difference for ADL DFM and AAT DFM . The required ADL and AAT in clusters v and viii had their ranges extended more than those of their component geo-regions.
It was assumed that the genetic clusters should coincide with the geographic subpopulations if each geographic subpopulation had unique genetic sources. In the present study, each of the geographic subpopulations involved several different genetic clusters, and each cluster involved several subpopulations; therefore, each genetic cluster might have its ADL and AAT response extended to compose of several subpopulations. Thus, a geographic subpopulation may come from multiple genetic sources, or extensive germplasm exchange among the subpopulations happened in history.

Conclusions
DSF and DFM are important phenological and evolutionary traits. The present study separated the two traits into the degree and duration of two basic environment factors of day length and temperature, respectively, i.e., ADL and AAT. DSF is more sensitive to ADL and AAT than DFM, and their sensitivity also varies in different geographical subpopulations and MG sets. For DSF, ADL DSF and AAT DSF decreased significantly from geographical region O to northward region A but increased significantly to southward region C, whereas for DFM, ADL DFM and AAT DFM decreased significantly in region A, but there was no increase in region C in comparison to region O. Their changes in region B and D were not as large as in A and C. The westward region were the newest ones but with the widest variation in DSF and DFM as well as their required ADL and AAT. This study also explored the ADL and AAT independent and joint/interaction effect on DSF and DFM. In the global population, ADL and AAT are more important to DSF variation, and their interaction (ADL × AAT) accounts for a certain part of DSF variation, but ADL is the dominant contributor to DFM. The relative contributions of ADL, AAT and ADL × AAT to DSF and DFM varied greatly among the geographic subpopulations. Another feature of this study is that genetic cluster is not fully consistent with geographical differentiation in the global population, indicating that multiple genetic sources existed in each geographical subpopulation. In addition, this study found that not only DSF but also DFM was related to ADL, which has been minimally reported previously.
Supplementary Materials: The following supporting information can be downloaded at: https://www. mdpi.com/article/10.3390/agronomy12040962/s1, Figure S1: Variation of accumulative day length (ADL) and accumulative active temperature (AAT) among months in experiments site in 2015, 2016 and 2017. Table S1: Distribution of material subgroups and maturity group of World Soybean Germplasm Population (WSGP). Table S2: Source information of the tested varieties in the World Soybean Germplasm Population (WSGP). Table S3: Analysis of variance of growth period traits, accumulative day length and accumulative active temperature in the World Soybean Germplasm Population (WSGP). Table S4: Regression analysis of variance of growth period traits, accumulative day length and accumulative active temperature. Table S5: Estimation of parameters in regression of growth period traits on accumulative day length and accumulative active temperature. Table S6: Variation of growth period traits, accumulative day length and accumulative active temperature in genetic clusters of the World Soybean Germplasm Population (WSGP).
Author Contributions: J.G. designed the study. C.W. performed the experiments. C.W. and J.H. analyzed the data. X.H. and Y.P. participated in the data analysis. X.L., C.Z., W.Z., W.W. and G.X. participated in the field experiments. C.W., J.H. and J.G. drafted the manuscript. All authors approved the manuscript. All authors have read and agreed to the published version of the manuscript.