Soybean cultivar Tanlong1 carrying the dominant functional E1 allele was used for genetic transformation. T₀, T₁ and T₂ gene editing plants were planted in a growth chamber under long-day conditions (LD, 16 h/8 h, light/dark) and short-day conditions (SD, 12 h/12 h, light/dark) at 70% relative humidity and 200-300 μmol m⁻² S⁻¹ light fluency. Phenotypes of the T₂ plants was recorded during the cultivation season (May to October) under a naturally LD conditions (LD, 16 h/8 h, light/dark) and artificially controlled short-day conditions (SD, 12 h/12 h, light/dark) in Harbin. The plant height and flowering time of the R1 stage (the first flower to appear) were recorded according to Fehr’s system (Fehr, 1971). Thirty plants of each genotype were measured and the data collected were subjected to statistical analysis.

CRISPR/Cas9 expression vector was constructed as described previously (Zhai et al., 2022). The CRISPR/Cas9 expression vector was transformed into Agrobacterium strain EHA105 by electroporation. Soybean transformation was performed using the cotyledonary node transformation system described previously (Flores et al., 2008). To identify the e1 mutants, we extracted total genomic DNA from leaf samples of putative mutants using EasyPure^® Plant Genomic DNA Kit (TransGen, Beijing, China) according to the manufacturer’s instructions. Subsequently, PCR analysis was performed using E1 sequence-specific primer sets ( Supplementary Table 1 ). PCR products were detected using 1.5% agarose gel electrophoresis and then sequenced. The sequences of T₀, T₁, and T₂ generation plants were analyzed using DSDecodeM to characterize the mutations induced by CRISPR/Cas9. Successfully edited lines were identified via sequence peaks and alignment to the reference sequences. The heterozygous mutants exhibited overlapping peaks near the target site, and the homozygous mutations showed single peaks at the target. The homozygous mutants were then identified by sequence alignment with the WT sequence. To screen and obtain E1 targeted mutants without transgenic elements, we performed PCR analysis using Bar sequence-specific primer sets ( Supplementary Table 1 ) to determine sgRNA/Cas9 on T-DNA elements. Potential off-target sites were predicted with the online tool: CRISPR-P (http://crispr.hzau.edu.cn/CRISPR2/) (Lei et al., 2014). Two potential off-target sites with the highest sequence identities to E1 targets were analyzed. PCR was performed to amply the genome fragment containing the potential off-target sites, using primer pairs listed in Supplementary Table 1 . PCR products were detected by 1.5% agarose gel electrophoresis and then sequenced.

Short-day promotion rate (SDHR) was calculated to determine the effect of photoperiod on flowering time, maturity and plant height as follows (Yan et al., 2009):

where V_LD represents the phenotype value under LD, and V_SD represents the phenotype value under SD.

Wild-type soybean cultivar ‘Tianlong1’ and e1 mutant were used for RNA-seq assay. Stem tips were collected 4 h after dawn from 28-day-old seedlings grown under LD conditions. Total RNA was extracted using TRIzol reagent (Invitrogen, Carlsbad, CA, USA) according to the manufacturer’s instructions. Sequencing cDNA libraries were generated according to the protocol of VAHTS ^® Universal V6 RNA-seq Library Prep Kit for Illumina (Vazyme, Nanjing, China). cDNA libraries were then constructed for sequencing on the Illumina NovaSeq 6000 S4 Seq system with 150-bp paired-end read lengths. Tophat2 was used to map clean reads to the soybean reference genome, Glycine max Wm82.a2.v1, from Phytozome (https://phytozome-next.jgi.doe.gov/). Differentially expressed genes (DEGs) were detected using DEGseq under the following parameters: fold change 2.00 and adjusted P-value (Q-value) 0.05. DEGs are listed in Supplemental Data Set S1 . Sequence data of RNA-seq from this study were deposited in the National Genomics Data Center (SRA) database under accession number PRJCA010529.

Total RNA was extracted from the leaves of 28-DAE-old plants of wild-type soybean TianLong1 and E1 mutant grown under LD conditions (16 h/8 h, light/dark). Three independent replicates of the total RNA samples were prepared for each analysis.

Total RNA was extracted using TRIzol (Invitrogen, Carlsbad, CA, USA) method according to the manufacturer’s instructions. Subsequently, the RNA was treated with RNase-free recombinant DNase I (Takara, Dalian, China). The integrity of the RNA was determined using NanoDrop™ ND-2000c Spectrophotometer (Thermo Scientific, Wilmington, DE, USA). Equal amounts of isolated RNA were then reverse transcribed to cDNA using the SuperScript™ III Reverse Transcriptase kit. The quality of the cDNA samples was assessed by PCR using GmTubulin A (TUA5) specific primers.

Each cDNA sample was subjected to qRT-PCR analysis using the SYBR Green Master Mix (TransStart Top Green qPCR SuperMix, Beijing, China). The qRT-PCR analysis was performed on LightCycler^® 96 real-time PCR detection system (Roche, Switzerland) according to the manufacturer’s protocol. The measured Ct values were converted to relative copy-numbers using the 2^-ΔΔCt method. TUA5 gene was used as an internal control to normalize all gene expression data. The qRT-PCR was performed using three fully independent biological replicates and each sample was run in triplicate. Raw data were standardized as described previously (Willems et al., 2008). Primers used for the qRT-PCR and semi-quantitative RT-PCR analyses are listed in Supplementary Table 1 .

In our previous study, to investigate the regulation of GmMDE genes by E1, we used CRISPR/Cas9-mediated mutation to inactivate E1 in the soybean cultivar Tianlong1, which carries a functional E1 allele (Zhai et al., 2022). In this study, we obtained four types of homozygous mutations. The e1-1 mutant line harbored a 244-bp deletion in the E1 coding region, whereas the e1-2 mutant line harbored a 247-bp deletion in the E1 coding region. The e1-3 mutant line harbored a 243-bp deletion in the E1 coding region was obtained, whereas the e1-4 mutant line harbored a 209-bp deletion in the E1 coding region. ( Supplemental Figure 1 ). Notably, the e1-1 and e1-2 were frameshift mutations, causing premature termination of translation. The e1-1 and e1-2 produced a truncated protein encoding 98 and 97 amino acids, respectively, resulting in the deletion of all B3 domains and retaining part of the nuclear localization signal. The e1-3 and e1-4 produced a truncated protein encoding 153 and 165 amino acids, respectively, which retained part of the B3 domains and nuclear localization signal ( Supplementary Figure 2 ). Potential off-target sites were predicted and the top 2 genomic regions of homology were selected as most likely off-target sites. Subsequently, each of these regions was amplified by PCR using genomic DNA from the mutant lines as template. The PCR products were further analyzed by Sanger Sequencing. Sequencing analysis did not reveal any potential off-target variants in the T₁ mutants ( Supplementary Figure 3; Supplementary Table 2 ). Thus CRISPR/Cas9 expression vector had specific edits at two targets. To obtain trans-clean mutants without T-DNA elements, we performed PCR to examine check whether there were traces of T-DNA in the mutants using Bar gene specific primers ( Supplementary Table 1 ). Among the four T₁ mutant lines, two were free of T-DNA in T₁ generation derived from e1-1, e1-2, e1-3 and e1-4 ( Supplementary Table 3 ). All the T₁ mutant lines was flowering earlier than wild-type soybean cultivar Tianlong1 under LD conditions ( Supplementary Figure 4 ).

To analyze the effect of E1 mutation on soybean photoperiod sensitivity, we planted T₂ generation seeds of the homozygous mutant e1-1 and e1-2 without T-DNA elements under LD and SD conditions, respectively. The wild-type soybean cultivar Tianlong1 is extremely sensitive to photoperiod. The flowering time, maturation and plant height of Tianlong1 plants grown under LD conditions significantly differed from those under SD conditions ( Figures 1A-D ). Furthermore, the photoperiod sensitivity of the two E1 mutants decreased greatly. Although the flowering time, maturation and plant height of two E1 mutants under LD were different from those under SD conditions, differences in maturation and plant height between plants grown under LD and SD conditions decreased. The reproductive period R8 of the two E1 mutants under SD conditions was about 25 days earlier than that under LD conditions, while the reproductive period R8 of Tianlong1 under SD conditions was more than 53 days earlier than that under LD conditions. The plant height of Tianlong1 under SD conditions was 60 cm shorter than that under LD conditions, while the plant height of the two E1 mutants under SD conditions was about 25cm shorter than that under LD conditions. Collectively, these results indicate that loss of E1 function reduces the photoperiod sensitivity of soybeans. Short-day hastening rate of flowering time, maturity, and plant height of the two E1 mutants were significantly lower than that of wild-type Tianlong1 ( Figure 1E ).

Since the two E1 mutants were sensitive to photoperiod, we examined whether genetic compensation response of E1 and its homologs exists. The soybean genome harbors two E1 homologs, E1-La and E1-Lb (Xia et al., 2012). The expression of E1-La and E1-Lb in leaves of the e1-2 mutant and its wild-type soybean cultivar Tianlong1 were analyzed. In e1-2 mutant plants, the expression levels of two E1-Ls were significantly higher than those in the wild type ( Figures 2A, B ), indicating that a genetic compensation response of E1 and its homologs exist.

In our previous study, E1 overexpression repressed the expression of GmFTs (GmFT2a and GmFT5a), two homologs of Arabidopsis FT (Xia et al., 2012). To clarify the correlation between the expression of E1 and GmFT2a/5a in flowering time regulation, we analyzed the transcript levels of GmFT2a and GmFT5a in the leaves of e1-2 mutant and wild-type ( Figures 2C, D ). GmFT2a and GmFT5a expression increased in e1-2 mutant compared to the wild-type Tianlong 1.

The plant architecture of two E1 mutants was investigated. Mutation of E1 altered the stem growth habit of the mutants. The stem growth habit of the wild-type TianLong1 tended to be indeterminate under natural LD conditions and determinate under natural SD conditions ( Figure 3F ). Under a longer light period (16h:8 h light/dark), TianLong1 entirely exhibited indeterminate stem growth habit ( Figure 3A ). Notably, the two E1 mutants exhibited determinate stem growth habits under both LD and SD conditions, characterized by early terminal flowering ( Figures 3B, F ), reduced plant height and decreased node number along the main stems ( Figures 1D , 3C ).

Soybean stem growth habit is regulated by two major genes, Dt1 and Dt2. Dt1 specifies the indeterminate growth habit, which prevents terminal flowering, leading to taller plants (Liu et al., 2010; Tian et al., 2010; Liu et al., 2016). Dt2 functions as a direct repressor of Dt1, promoting terminal flowering and leading to shorter plants (Liu et al., 2016). Here, we hypothesized that E1 might regulate stem growth habit by modulating Dt1 and Dt2 genes. We thus measured the expression levels of Dt1 and Dt2 in the e1-2 mutant and wild-type Tianlong1 plants. Dt1 expression was higher in the stem tips of wild type plants than in e1-2 mutant, while Dt2 expression was significantly higher in the stem tips of e1-2 mutant than in wild-type stem tips ( Figures 3D, E ). These results indicate that E1 regulates the stem growth habit via the Dt2- Dt1 signaling pathway.

The branch number of wild type plants and two E1 mutants were investigated. Mutation of E1 altered the branch number. Wild type plants produced much more branches than the two E1 mutants ( Figures 3F, G ). TianLong1 produced significantly more branch numbers under LD conditions than under SD conditions; however, the branch number of E1 mutants was less different between SD and LD conditions ( Figures 3F, G ), indicating that mutation of E1 reduces photoperiod sensitivity in branch number.

Transcriptome sequencing (RNA-seq) and expression analysis of differentially expressed genes (DEGs) were performed further to understand the molecular differences between TianLong1 and e1-2. Similar gene expression levels were observed among three biological replicates ( Figure 4A ), WT and e1-2 were distinguished via principal component analysis (PCA). A total of 1161 DEGs were obtained by comparing the RNA-seq datasets (P< 0.05). Volcano plots were used to visualize the significant DEGs in e1-2 ( Figure 4B ). The red points in the graphic represent significantly upregulated genes, while the blue points in the graphic represent downregulated genes. Furthermore, we analyzed the gene ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment of the differentially expressed mRNAs between E1 mutants and controls. The most enriched terms GO terms included cell part (GO: cellular component), metabolic process (GO: biological process), binding (GO: molecular function), cellular process (GO: biological process), catalytic activity (GO: molecular function) and organelle (GO: cellular component) ( Figure 4C ). The most significantly enriched KEGG pathway was phenylpropanoid biosynthesis, Cutin, suberin and wax biosynthesis ( Figure 4D ).

The expressions of several flowering related genes were altered in the E1 knockdown lines, consistent with the e1-2 early flowering phenotype, with most of the genes up-regulated. Among the 32 MADS-box genes, 28 were significantly up-regulated and 4 were down-regulated. Specifically, E1 knockdown significantly up-regulated floral meristem identity genes, LEAFY (LFY) and APETALA1 (AP1), and most floral organ identity genes. Furthermore, the expression of SEPALLATA(SEP), CAULIFLOWER(CAL) and WUSCHEL(WUS) genes were up-regulated. E1 knockdown also affected the expression of multiple key factors in auxin and gibberellin signaling pathways. In particular, the expression of PIN, encoding auxin efflux carrier protein and GA2OX, the key enzymes in Gibberellin (GA) synthesis were upregulated. Also, 12 differentially expressed NAC transcription factors that mediate SAM formation were up-regulated ( Figure 5 and Supplementary Table 4 ).