Functional Characterization of the Steroid Reductase Genes GmDET2a and GmDET2b from Glycine max

Brassinosteroids are important phytohormones for plant growth and development. In soybean (Glycine max), BR receptors have been identified, but the genes encoding BR biosynthesis-related enzymes remain poorly understood. Here, we found that the soybean genome encodes eight steroid reductases (GmDET2a to GmDET2h). Phylogenetic analysis grouped 105 steroid reductases from moss, fern and higher plants into five subgroups and indicated that the steroid reductase family has experienced purifying selection. GmDET2a and GmDET2b, homologs of the Arabidopsis thaliana steroid 5α-reductase AtDET2, are proteins of 263 amino acids. Ectopic expression of GmDET2a and GmDET2b rescued the defects of the Atdet2-1 mutant in both darkness and light. Compared to the mutant, the hypocotyl length and plant height of the transgenic lines GmDET2a and GmDET2b increased significantly, in both darkness and light, and the transcript levels of the BR biosynthesis-related genes CPD, DWF4, BR6ox-1 and BR6ox-2 were downregulated in GmDET2aOX-23 and GmDET2bOX-16 lines compared to that in Atdet2-1. Quantitative real-time PCR revealed that GmDET2a and GmDET2b are ubiquitously expressed in all tested soybean organs, including roots, leaves and hypocotyls. Moreover, epibrassinosteroid negatively regulated GmDET2a and GmDET2b expression. Sulfate deficiency downregulated GmDET2a in leaves and GmDET2b in leaves and roots; by contrast, phosphate deficiency upregulated GmDET2b in roots and leaves. Taken together, our results revealed that GmDET2a and GmDET2b function as steroid reductases.

In animals, steroid hormones are perceived through nuclear receptors. By contrast, in plants, BR is sensed by the BR receptors such as brassinosteroid insensitive 1 (BRI1) in the plasma membranes [12,13]. Recently, the linear BR signaling pathway in Arabidopsis has been revealed. In the absence of BRs, BRI1 is bound by the membrane-located kinase BKI1 [14]. Upon reception of BR, BRI1 disassociates BKI1 [14] and physically interacts with the membrane kinase BAK1, which is considered as a co-receptor [15]. BRI1 and BAK1 transphosphorylate each other [14]. Brassinosteroid insensitive 2

Identification of Steroid Reductases in Soybean
To identify steroid reductases in the soybean proteome, we took two approaches. First, we searched the soybean proteome for the PF02544 domain (http://pfam.xfam.org/), the signature domain of steroid reductase [40], and found a total of six putative steroid reductase. Second, the amino acid sequence of the known steroid reductase AtDET2 was used to search the soybean proteome with BLAST-P (http://blast.ncbi.nlm.nih.gov), and the two proteins encoded by Glyma.07g144400 and Glyma.11g010000 showed the highest identity and similarity to AtDET2. Collectively, the soybean genome encodes eight steroid reductase (Table 1). Although the PF02544 domain was not found in Glyma.07g144400 and Glyma11g010000, the BLAST search against the soybean proteome with Arabidopsis DET2 showed that Glyma.07g144400 and Glyma.11g010000 appear to have a similar function to that of AtDET2, and thus, they have been named GmDET2a and GmDET2b, respectively. Putative cell localization of soybean steroid reductases was predicted with PSORT. Based on the released genome sequences and cDNA sequences of soybean, the number of introns and exons were determined through GSDS as described in the Materials and Methods. EST, expressed sequence tag; Plas, plasma membrane; Nucl, nucleus; Cyto, cytoplasm; AA, amino acid; Int, intron; Ext, extron.
Based on previous RNA-Seq study on soybean [41], we found that GmDET2a and GmDET2b are ubiquitously expressed in young leaves, flowers, pods, seeds, roots and nodules, with relatively higher transcript abundance in roots and nodules ( Figure 1). The transcript levels of GmDET2d and GmDET2h are higher in roots than in other organs, while those of GmDET2b and GmDET2d are higher in nodules. The transcripts of GmDET2c and GmDET2h are higher in leaves, and the expression levels of GmDET2c and GmDET2h are higher in flowers. Interestingly the transcripts of all eight steroid reductase genes can be detected in seeds, with transcripts of GmDET2c in 10 days after fertilization (DAF) seeds being significantly higher. The abundance of GmDET2c and GmDET2h is somewhat higher in one-cm pod, pod shell 10 DAF and pod shell 14 DAF ( Figure 1B).
Legend:  Figure 1. Gene structure of Arabidopsis and soybean steroid reductase and transcript levels of the soybean steroid reductase family. The gene structure of seven and eight steroid reductase from Arabidopsis and soybean are shown (A). CSDS2.0 was used to analyze the CDS (coding DNA sequence), UTR (untranslated region) and intron of these genes. The heat map (B) demonstrates the expression levels of GmDET2a to GmDET2h. The normalized expression data from RNA-Seq were downloaded from SoyBase [41]. DAF, day after fertilization.

Phylogenetic Analysis of Steroid Reductases
We aligned the amino acid sequences of eight soybean DET2s with 12 homologs from the reference species Arabidopsis thaliana and Oryza sativa and reconstructed the phylogenetic tree of DET2s with MEGA6 [42]. As expected, GmDET2a and GmDET2b grouped together with AtDET2 and Os01g63260. GmDET2c and GmDET2d clustered with At5g16010, GmDET2e with Os04g48750, GmDET2f and GmDET2g with At3G55360 and GmDET2h with At1g73650. In short, DET2s from soybean, Arabidopsis and rice can be grouped into five sub-clades ( Figure 2).
To extensively decipher the evolutionary history of steroid reductase in the kingdom Plantae, we further collected steroid reductases from multiple plants, including Physcomitrella patens, Selaginella moellendorffii, Oryza sativa, Brachypodium distachyon, Sorghum bicolor, etc. After alignment with the CLUSTALW program, sequences were analyzed with MrBayes 3.2 [43]. As shown in Figure 3, steroid reductases of Plantae are grouped into five subgroups. Of note, except for group V , all subgroups have members from Physcomitrella patens and Selaginella moellendorffii. This indicates that higher plants have retained ancestral steroid reductase genes and that steroid reductase plays an important role in growth and development.  [44]. The tree with the highest log likelihood (−2073.6839) is shown. The percentage of trees in which the associated taxa clustered together is shown next to the branches. Initial tree(s) for the heuristic search were obtained automatically by applying the neighbor-join and BioNJalgorithms to a matrix of pairwise distances estimated using a JTTmodel and then selecting the topology with the superior log likelihood value. A discrete gamma distribution was used to model evolutionary rate differences among sites (five categories (+G, parameter = 5.1876)). The tree is drawn to scale, with branch lengths measured in the number of substitutions per site. The analysis involved 20 amino acid sequences. All positions containing gaps and missing data were eliminated. There was a total of 74 positions in the final dataset. Evolutionary analyses were conducted in MEGA6 [42]. Steroid reductases from soybean, Arabidopsis and rice can be grouped into five subgroups (I, II, III, IV and V).

Transcript Levels of GmDET2a and GmDET2b in the Seedling Stage
Taking into consideration the important roles of brassinosteroid in plant growth and development, that GmDET2 appear to be involved in BR biosynthesis, and that GmDET2a and GmDET2b are the closest homologs of AtDET2, we thus focused on them in subsequent experiments. We first employed quantitative real-time PCR (qRT-PCR) to detect the transcripts of GmDET2a and GmDET2b in the seedlings of soybean. As shown in Figure 5, qRT-PCR results showed that GmDET2a and GmDET2b were both expressed in apical buds, leaves, hypocotyls, primary root (PR) tips and lateral roots. The transcripts of GmDET2a in apical buds, PR tips and lateral roots are somewhat higher than those in leaves and hypocotyls ( Figure 5). A similar pattern was observed in the case of GmDET2b ( Figure 5). . Transcript levels of GmDET2a and GmDET2b in soybean seedlings. RNA was extracted from seven-day-old soybean YC03-3. Quantitative real-time PCR was exploited to detect the transcripts of GmDET2a (in blue) and GmDET2b (in green) in apical buds, leaves, hypocotyls, primary root (PR) tips and lateral roots. Results are the means ± SD from four replicates. One-way ANOVA was used to compare the differences between organs, with uppercase letters or lowercase letters indicating the difference at the 0.05 level.

GmDET2a and GmDET2b Were Negatively Regulated by Epibrassinosteroid
As previously reported, AtDET2 is negatively regulated by exogenous BR [45,46]. To explore this relationship in soybean, we looked at the expression pattern of GmDET2a and GmDET2b in soybean seedlings treated with epibrassinosteroid (EBL) for 6 h. As shown in Figure 6A, in general, GmDET2a was negatively regulated by 6 h EBL treatment. The abundance of GmDET2a was significantly reduced in apical buds (p < 0.01), leaves (p < 0.01), primary root (PR) tips (p < 0.01) and lateral roots (p < 0.05), but not negatively regulated by EBL in hypocotyls. GmDET2b was negatively regulated by EBL in only primary root tips (p < 0.01) and lateral roots (p < 0.05), but not in apical buds, leaves and hypocotyls ( Figure 6B). GmDET2b (B) in apical buds, leaves, hypocotyls, primary root (PR) tips and lateral roots. Results are the means ± SD from 4 replicates. Student's t-test was used to compare the differences between EBL treatment and control, respectively (*, 0.01 < p < 0.05; **, 0.001 < p < 0.01).

Responses of GmDET2a and GmDET2b to Nutrient Deficiencies
As our group is interested in plant nutrition, we also explored the responses of GmDET2a ( Figure 7A) and GmDET2b ( Figure 7B) to nutrient deficiencies. Soybean seedlings were starved of phosphate (-P), nitrogen (-N), potassium (-K) or sulfate (-S) for 14 days. Analysis by qRT-PCR showed that the transcripts of GmDET2a in leaves were decreased by -S, and those of GmDET2b in leaves and roots were reduced by -S (p < 0.05; Figure 7B), while being upregulated by -P in both leaves and roots (p < 0.05). The transcripts of GmDET2a and GmDET2b were not regulated by -N or -K in either leaves or roots.

Ectopic Expression of GmDET2a and GmDET2b in Atdet2-1 Rescued the Growth and Development Defects
A loss of function mutation in DET2 affects the growth and development of Arabidopsis in both darkness and light [5,35]. To determine the functions of GmDET2a and GmDET2b, we constructed over-expression plasmids with the pCHF3 vector and transformed the loss-of-function mutant, Atdet2-1 [5]. We selected transgenic lines and identified single-copy lines based on the segregation ratio of 3:1. The transcript levels of GmDET2a and GmDET2b were explored in transgenic lines GmDET2aOX-20, GmDET2aOX-23, GmDET2bOX-16 and GmDET2bOX-23. As shown in Figure S1, the transgenic lines of GmDET2a and GmDET2b were overexpressing lines.
Consistent with previous results [5], the leaves of Atdet2-1 mutant are curled with a shorter petiole, as compared to wild-type (WT, Columbia (Col-0)) ( Figure 8). As expected, when overexpressing GmDET2a or GmDET2b in Atdet2-1, the defective leaf growth and development were rescued. Images were taken after 15 days ( Figure 8A) and 30 days of germination ( Figure 8B). Consistent with previous study, the Atdet2-1 mutant is dwarf with short siliques ( Figure 9A). We measured the height of transgenic Arabidopsis after 50 days of germination. The height of transgenic lines is higher in contrast to Atdet2-1 (p < 0.05, Figure 9B). This indicates that like AtDET2, GmDET2a or GmDET2B complement AtDET2, a steroid reductase.
BRs have been documented to play crucial roles in skotomorphogenesis [12], and the Atdet2-1 mutant has obviously different phenotypes from wild-type in the dark [5]. Hence, we compared the differences of Atdet2-1 and transgenic lines in the dark. In line with previous reports, different from Col-0, the Atdet2-1 mutant has a shorter hypocotyl (0.86 cm VS 1.84 cm) and primary roots and open cotyledons ( Figure 10A). Overexpression of GmDET2a or GmDET2b in Atdet2-1 rescued the hypocotyl growth and primary root growth in the dark. The cotyledons in transgenic lines are less open in contrast to Atdet2-1. Data in Figure 10B verified that ectopic expression of GmDET2a or GmDET2b increases the hypocotyl growth in darkness. Compared to Atdet2-1, the hypocotyl length in the GmET2aOX-20, GmDET2aOX-23, GmDET2bOX-16 and GmDET2b-23 lines was increased by 110%, 87%, 146% and 121%, respectively (p < 0.05). Taken together, ectopic expression of GmDET2a and GmDET2b rescued the growth and development defects of Atdet2-1 in both the dark and light conditions.

Over-Expression of GmDET2a or GmDET2b Rescued the Responses of BR Biosynthesis-Related Genes to Exogenous BR
As reported, BR biosynthesis-related genes are feedback regulated by BR [47]. Therefore, we sought to determine whether the expression of GmDET2a or GmDET2b in the Atdet2-1 mutant background could rescue the responses of BR biosynthesis-related genes CPD, DWF4, BR6ox-1 and AtBR6ox-2.
As demonstrated in Figure 11, compared to WT, the abundances of BR6ox-1, BR6ox-2, CPD and DWF4 were relative higher in the det2-1 mutant. When overexpressing GmDET2a in Atdet2-1, BR6ox-1, BR6ox-2, CPD and DWF4 transcripts were significantly reduced (p < 0.05); with similar results observed when overexpresing GmDET2b in Atdet2-1. These results verified that soybean DET2a and DET2b complement the function of AtDET2 at the molecular level.  Figure 11. Effect of ectopic over-expression of GmDET2a and GmDET2b on the transcripts of BR biosynthesis-related genes in the Atdet2-1 mutant. Arabidopsis seedlings were grown as described in the Materials and Methods. Quantitative RT-PCR was used to quantify the relative expression levels of BR6ox-1 (A), BR6ox-2 (B), CPD (C) and DWF4 (D) in det2-1 and transgenic lines. Results are the means ± SD from four independent experiments with three technical repeats. One-way ANOVA was used to compare the difference among lines, with different letters indicating significant differences at the 0.05 level.

Discussion
Brassinosteroids are classic phytohormones, playing crucial roles in plant growth, development, abiotic and biotic stress responses [12]. Although the BR biosynthesis pathway in Arabidopsis has been extensively studied, the understanding of BR biosynthesis in soybean is poor. In this study, we identified eight soybean steroid reductase genes (GmDET2a to GmDET2h) and functionally characterized GmDET2a and GmDET2b at the physiological and molecular levels.
BR biosynthesis has been documented, thus far, in 61 plant species, including 53 angiosperm, 6 gymnosperms, 1 pteridophyte (Equisetum arvense) and one bryophyte (Marchantia polymorpha) [48]. Single-celled green freshwater algae (Chlorella vulgaris and Hydrodictyon reticulatum) and the marine brown alga Cystoseira myrica biosynthesize BRs, as well [49], indicating that BRs are highly conserved phytohormone. Based on studies in Arabidopsis and rice, DET2, encoding steroid reductase, a crucial enzyme in BR biosynthesis, catalyzes the rate-limiting step reaction [47]. Homologues of DET2 in tomato and cotton have been functionally identified, as well [50,51] In this study, we found that the soybean genome encodes eight steroid reductase (Table 1). This indicates that brassinosteroid biosynthesis in soybean is more complex than in other species. GmDET2a and GmDET2b are intron-less genes, and consistent with AtDET2, GmDET2a and GmDET2b do not have the PF02544 steroid reductase signature (Table 1). Transgenic experiments verified that GmDET2a and GmDET2b function as steroid reductases in Arabidopsis, but more studies are needed to reveal the functions of the other six soybean DET genes. We also used PF02544 to search against the Phytozome database and retrieved 144 proteins containing steroid reductase/dehydrogenase domains from 19 species (Additional File 1). Evaluated by Prottest-2.4, JTT + G + F was selected as the best model to rebuild the evolutionary tree, utilizing MrBayes 3.2 software. As shown in Figure 3, steroid reductases were grouped into five subgroups. Except for Subgroup V, the other four subgroups contain steroid reductases from both Physcomitrella patens and Selaginella moellendorffii, implying that at least four ancestral steroid reductase genes exist in Plantae. Of note, fungus Ustilago maydis steroid 5α-reductase complements Arabidopsis DET2 function [52], and overexpressing a human steroid 5α-reductase rescues the growth and development of Atdet2-1 [35]. This evidence further verifies the existence of a common ancestor for steroid reductase between fungal, plant and mammalian proteins.
We found that in most branches, the Ka/Ks of steroid reductases was less than 1.0 (Figure 4), indicating that steroid reductases have been subjected to purifying selection. Ka/Ks values are also consistently less than one in BR receptor genes [32]. Similarly, as high as 76% analyzed jasmonate ZIM-domain (JAZ) genes from 13 monocot and dicot species have also been subject to purifying selection [53], and the Ka/Ks of the squamosa promoter binding protein (SBP)-box genes in plants were less than 0.5 [54]. Our results support the notion that the Ka/Ks of crucial genes in many plant species are less than 1.0 [55].
RNA-Seq data indicate that transcripts of eight soybean DET genes can be detected in roots, leaves, flowers, seeds and nodules ( Figure 1B), implying that BR regulates the development and growth of these organs. Notably, the transcript abundances of GmDET2a and GmDET2b are higher in nodules and roots, indicating their important role in nodulation and rooting. The interaction between BR and auxin has been shown to stimulate lateral rooting [56], and BR signaling has been proven to boost root hair development [57].
Similarly, in tomato, the transcripts of LeDET2 can be detected in leaves, stems, roots, seeds and callus, with the highest abundance observed in leaves [50]. In this study, qRT-PCR results revealed that GmDET2a and GmDET2b have similar expression patterns ( Figure 5) with higher expression in primary root tips, apical buds and lateral roots, implying that BR biosynthesis is active in these tissues. Differing from auxin, BR does not experience polar transport; thus, it is understandable that DET2 shows ubiquitous expression in primary roots, lateral roots, seeds and nodules. The future challenge will be to reveal the differing roles of soybean DET2s in different organs.
We found that application of EBL for 6 h significantly decreased the transcripts of GmDET2a and GmDET2b in apical buds, leaves, PR tips and lateral roots ( Figure 6A,B). DET2, CPD, DWF4, BR6ox-1 and BR6ox-2 are key enzymes for BR biosynthesis [47] and are negatively regulated by exogenous BR or BR signal [47,58]. In line with this, we found that the transcripts of CPD, DWF4, BR6ox-1 and BR6ox-2 were downregulated in WT in contrast to Atdet2-1, but overexpressing GmDET2a or GmDET2b negatively regulated the levels of these four genes in Atdet2-1 (Figure 11). This is possibly due to the fact that ectopic expression of GmDET2a or GmDET2b in Atdet2-1 increases endogenous BR amounts. Molecular evidence also supports the idea that GmDET2a or GmDET2b encodes steroid reductase. Transcripts of DWF4 in WT are low, but those of DWF1, CPD and BRI1 are higher, indicating that transcription of DWF4 is downregulated by BR [47]. The fact that Arabidopsis CPD, DWF4 promoters contain binding elements of BZR1 or BES1 prompted us to conceive of the idea that the same cis-elements also exist in the promoter region of GmDET2a and GmDET2b. The homologue of BES1 or BZR1 can be blasted in soybean proteome.
Phosphate starvation induces the expression of GmDET2b, and S starvation downregulates the transcripts of GmDET2a and GmDET2b (Figure 7). This hints at potential crosstalk between nutrients and BR biosynthesis. Whether it is phosphate-or sulfate-related transcription factors binding the promoters of GmDET2a or GmDET2b would be worthy of future study.
Atdet2-1 mutants exhibit a dwarf phenotype, and in darkness, the hypocotyl length of det2-1 is shorter than WT [5]. By ectopically expressing GmDET2a or GmDET2b in Atdet2-1, we show that GmDET2a and GmDET2b function as steroid reductase, as their overexpression rescued the defects of the mutant (curled leaf, reduced plant height in light and reduced hypocotyl length in dark) (Figures 8-10). In conclusion, we functionally identified GmDET2a and GmDET2b as steroid reductases involved in BR biosynthesis.

Soybean and Arabidopsis Growth
Glycine max genotype YC03-3 was used in this study. Soybean seeds were sterilized with 10% NaClO and germinated in sand until the cotyledon was open. The seedlings were then transferred to Hoagland solutions (pH = 5.9) for three days, until the first true leaf developed. The seedlings were then cultured in 10 µM EBL solution for 0 h and 6h, and the leaves were sprayed with 10 µM EBL (+EBL) or 0 µM EBL (-EBL). The apical buds, true leaves, hypocotyls, primary root tips and lateral roots were then removed and stored in liquid nitrogen. To study nutrient stress, germinating soybean seedlings were first cultured in complete Hoagland solution (pH = 5.9) for seven days, as described [59], and the seedlings with the first fully-developed trifoliate leaves were transferred into phosphate (Pi)-sufficient (Control, 250 µM KH 2 PO 4 ) or phosphate deficient (-P) (0 µM KH 2 PO 4 ) nutrient solution, for 14 days. For nitrogen starvation (-N), NH 4 NO 3 was omitted, and KCl was substituted for KNO 3  Arabidopsis seeds were sterilized with 75% ethanol for 2 min followed by 100% ethanol for 2 min. Seeds were dried in a hood and sown on half strength MS media containing 1% sucrose and 0.8% agar (pH 5.8). Square Petri dishes were maintained in a growth chamber with a 16-h light (100 µE m −2 s −1 )/8-h dark cycle and with a temperature cycle of 23 • C light/21 • C dark. For the darkness treatment, the square Petri dishes were wrapped with aluminum foil.

Extraction of Genomic DNA, RNA and Reserve Transcription of mRNA
Arabidopsis and soybean RNA were extracted with the TRIzol method. RNA were reverse transcribed through MLV-transcriptase.

Determination of Soybean GmDET2 Family Structure
Based on the full cDNA and genomic DNA sequences of GmDET2a to DET2h (www.phytozome.org), we exploited CSDS2.0 to plot the gene structure of GmDET2s.

Phylogenetic Analysis
Alignment of steroid reductases from Glycine max, Arabidopsis and Oryza sativa was done with ClustalW2.1. The evolutionary history was inferred by using the maximum likelihood method based on the Jones et al. w/freq.model [44]. The tree with the highest log likelihood (−2073.6839) is shown. The percentage of trees in which the associated taxa clustered together is shown next to the branches. Initial trees for the heuristic search were obtained automatically by applying neighbor-join and BioNJalgorithms to a matrix of pairwise distances estimated using a JTT model and then selecting the topology with the superior log likelihood value. A discrete gamma distribution was used to model evolutionary rate differences among sites (5 categories (+G, parameter = 5.1876)). The tree is drawn to scale, with branch lengths measured in the number of substitutions per site. The analysis involved 20 amino acid sequences. All positions containing gaps and missing data were eliminated. There was a total of 74 positions in the final dataset. The bootstrap values were set to 1000. Evolutionary analyses were conducted in MEGA6 [42].

Analysis of Expression Patterns of Soybean DET2 Family at the Seedling Stage
Based on the cDNA sequence and genomics sequences of GmDET2a and GmDET2b (www. phytozome.org), we designed specific primer pairs with PerlPrimer to detect their expression levels. Seven-day-old soybean cultivar YC03-3 seedlings were transferred to half strength Hoagland nutrient solutions to grow a further 7 days treated with different nutrient solutions. The roots, stems and leaves were sampled to extract total RNA. Quantitative real-time PCR was used to test the expression levels of GmDET2a and GmDET2b, and soybean EF1a was used to normalize the data. Forty qRT-PCR cycles were run (95 • C 1 min, 95 • C 15 s, 60 • C 15 s, 72 • C 30 s, 72 • C 45 s), and Rotorgen software was used to calculate the PCR results. Data were from three independent biological experiments.

Overexpression of GmDET2a and GmDET2b in Arabidopsis det2
As the coding region sequences of GmDET2a and GmDET2b are nearly 100% identical, we did two rounds of PCR to clone the open reading frame of GmDET2a and GmDET2b. First, we designed PCR primers in the 5 UTR and 3 UTR of GmDET2a and GmDET2b and then used the common forward and reverse primers GmDET2a/b.oxF and GmDET2a/b.oxR that contain the KpnI and SalI restriction sites, respectively, to amplify the coding fragment. The coding fragments of GmDET2a and GmDET2b and pCHF3 plasmid were digested with KpnI and SalI, respectively, and the digested pCHF3, GmDET2a and GmDET2b coding fragments were ligated using DNA ligase. The Atdet2-1 mutant was transformed with pCHF3-GmDET2a or pCHF3-GmDET2b using the floral dip method and the GV3101 agrobacterium strain [60]. Transformed lines were selected in media containing 50 µg/mL kanamycin. The single copy T-DNA transformed lines were determined based the 3:1 segregation ratio. The homozygous F3 were used in further experiments.

Phenotypic Analysis of Transgenic Arabidopsis
Seeds of wild-type Col-0, Atdet2-1 and overexpression lines in mutant background, as indicated, were stratified in 4 • C for 2 days to break dormancy and then sown into soils, and after being cultured as indicated, seedling height was measured with a ruler.

Determination of the Expression Levels of BR Biosynthesis-Related Genes in Col-0, det2-1 and Transgenic Lines
To test the expression levels of BR biosynthesis-related genes CPD, DWF4, BR6ox-1 and BR6ox-2, 3-week-old Arabidopsis seedlings of Col-0, Atdet2-1 and overexpression lines in mutant background were transplanted to the soil for one month under standard growth conditions, and then, the whole plants were used to extract total RNA with the TRIzol method. cDNAs were obtained with reverse transcriptase-mediated reactions. Transcript abundance of CPD, DWF4, BR6ox-1 and BR6ox-2 was determined using qRT-PCR, and AtEF1a was used as a reference gene to normalize the qRT-PCR results. Specific primer pairs are listed in Supplementary Table S2.

Data Analysis
All data were analyzed with Excel2003. Student's t-test and one-way ANOVA were employed to compare the differences. R 3.0.1 package [61], ggplot2 and gplots 2.12.1 [62] were used to draw figures and the heat map, respectively.

Conclusions
The soybean genome has eight genes that encode steroid reductase (GmDET2a to GmDET2h), and GmDET2a and GmDET2b were negatively regulated by exogenous BR. BR reductases in plants were subjected to purifying selection during evolution. Ectopic overexpression of GmDET2a and GmDET2b in Atdet2-1 rescues the BR biosynthesis deficiency-related growth and development defects of the det2-1 under dark and light conditions. Supplementary Materials: The following are available online at www.mdpi.com/1422-0067/19/3/726/s1.

Acknowledgments:
This study was partially supported by NSFC (No. 31572184). We thank Joanne Chory for providing det2-1 seeds and Jennifer Mach for her comments and help in English writing.
Author Contributions: Weige Huo participated in the study design, carried out the experiments and data analysis and drafted the manuscript. Bodi Li, Jiebing Kuang, Pingan He and Zhihao Xu carried out the experiments and analyzed the data. Jinxiang Wang conceived and designed experiments, analyzed the data and authored the manuscript. All authors have read and approved the final manuscript.

Conflicts of Interest:
The authors declare no conflict of interest.