To adapt to temperature and light fluctuations, sessile plants have evolved internal machineries to sense and acclimate to these environmental changes to maintain their homeostatic balance (Creux & Harmer, 2019; McClung, 2019) . Anticipation of external conditions have allowed the development of an internal timing mechanism, the circadian clock, to align their key biological processes to a 24-hour period. In the model plant Arabidopsis thaliana (Arabidopsis), the circadian clock is a complex gene regulatory network consisting of multiple intertwined feedback loops of transcriptional repressors that creates daily rhythms and influences a wide range of physiological processes, including flowering transition in plants (Harmon et al. 2018; Creux & Harmer, 2019; Ronald and Davis 2019) . LATE ELONGATED HYPOCOTYL ( LHY ) and CIRCADIAN CLOCK ASSOCIATED 1 ( CCA1 ) genes that encode the closely related MYB family transcription factors (Alabadí, Yanovsky, Más, Harmer, & Kay, 2002; Mizoguchi et al., 2002) are expressed in the morning and repress PSEUDO-RESPONSE REGULATOR ( PRR ) genes, including TIMING OF CAB2 EXPRESSION 1 ( TOC1 ), PRR5 , PRR7 , and PRR9 . These PRR genes are expressed in the afternoon and repress LHY and CCA1 genes in turn, creating a feedback loop (Nakamichi et al., 2010; Nakamichi et al., 2012) ). LHY and CCA1 also repress the components of the Evening Complex (EC): the transcriptional repressor EARLY FLOWERING 3 ( ELF3 ) , the nuclear protein ELF4, and the MYB transcription factor LUX ARRHYTHMO ( LUX ) (Huang & Nusinow, 2016) . The EC genes in turn repress LHY , CCA1 , and multiple PRR genes (Chow, Helfer, Nusinow, & Kay, 2012; Ezer et al., 2017; Mizuno et al., 2014) .

Our current knowledge in the molecular basis underlying the circadian clock in crop species is limited. In soybean ( Glycine max ), recent studies have reported that soybean’s clock gene homologs affect photoperiodic flowering transition by modulating the E1 gene, a legume specific flowering repressor (Xia et al., 2012) . The soybean LHY homolog LHY1a directly binds to the promoter of the E1 gene and inhibits its expression (Lu et al., 2020) , while GmPRR3 genes repress LHY/CCA1 homologs, derepressing E1 expression (Li et al., 2020; Lu et al., 2020) . In addition, components of the evening complex (EC), GmELF3 and GmLUX homologs, are shown to directly inhibit E1 expression (Bu et al., 2021; Lu et al., 2017) . In Arabidopsis, the EC is known to play a central role in the entrainment of the circadian clock and the coordination of plant growth and environmental signals (Ezer et al., 2017; Huang & Nusinow, 2016) . Similarly, the soybean ELF3 homolog GmELF3-1 , the causal gene for the J locus that confers a long juvenile trait, is involved in the latitudinal adaptation of soybean (Lu et al., 2017; Yue et al., 2017) . Soybean is a short-day flowering plant, and its yield is critically dependent on the photoperiod in each latitudinal zone. Recessive mutations at GmELF3-1 caused late flowering and higher yield under short day (SD) conditions via upregulation of E1 , providing the better adaption in tropic, low-latitude zones such as Brazil, the second largest producer of soybean. Therefore, GmELF3-1 is a crucial component of a regional expansion of soybeans.

In this study, we aimed to gain a better understanding of the regulatory roles of the EC gene GmELF3-1 in the circadian clock and flowering gene network of soybean. To test the regulatory effects of GmELF3-1 on downstream gene expression, soybean protoplast cells were used as a transient model. Previously, regulatory interactions among 82 soybean’s circadian clock and flowering genes were inferred using in-house time series RNA-seq data and the network inference algorithmic package CausNet (Wu et al., 2019) . Candidate target genes downstream of GmELF3-1 were identified: GmCOL1a, GmPRR7-1, and GmPRR7-2 (Alcantara et al., 2022) . GmPRR7-1 and GmPRR7-2 were predicted to be downregulated by GmELF3-1 with strong reliability weights of 0.62 and 0.72, respectively, in the photothermal condition long day (LD) at 25°C. GmCOL1a was upregulated by GmELF3-1 with a very strong reliability weight of 1.0, in LD at 16°C.

To verify these inferred regulatory interactions experimentally, the effects of GmELF3-1 in the candidate downstream genes were examined using soybean protoplast cells as a transient model in two ways. In the first approach, GmELF3-1 was overexpressed in protoplast cells by transfecting them with GFP-GmELF3-1 fused to the Cauliflower Mosaic Virus (CaMV) 35S promoter (Table 2) . By constitutively expressing GFP-GmELF3-1 , we can measure gene expression changes of inferred target genes. Protoplasts isolated from unifoliate leaves were transfected, incubated overnight, and harvested at the Zeitgeber time points ZT4, ZT8, ZT12, ZT16, ZT20, and ZT24. Cell viability and GFP expression were examined prior to harvest. Approximately 90% of protoplasts were viable and about 70% of transfected protoplasts were expressing GFP signals at all time points, while non-transfected protoplasts showed no fluorescence (Figures B-C) . Our RT-qPCR analysis confirmed that GmELF3-1 was expressed significantly higher in transfected protoplasts than in non-transfected control protoplasts by a range of 150,000-fold to 420,000-fold across the 6 time points (Figure E) . GmELF3-1 overexpression affected mRNA expression levels and patterns of the inferred target genes. GmPRR7-1 was upregulated at ZT16 and ZT20 by 7-fold and 4-fold, respectively, in transfected protoplasts than in non-transfected control protoplasts (Figure F) . GmPRR7-2 was upregulated at ZT16 and ZT20 by 5-fold and 2-fold, respectively, while it was downregulated at ZT24 by 1.5-fold (Figure G) . Significant upregulation of GmCOL1a was observed at ZT4 by 3,000-fold, while upregulation at other time points was marginable (Figure H) .

In the second approach, we examined the effects of GmELF3-1 knockout using CRISPR-Cas9 targeted mutagenesis as a counterpart of the overexpression approach. Protoplasts were transfected with the following ribonucleoprotein (RNP) complexes ( Tables 3 and 4) : single guide RNA (sgRNA) on the positive strand (g+) bound to Cas9 nuclease, RNP(g+); sgRNA on the negative strand (g-) bound to Cas9 nuclease, RNP(g-); or paired guide RNA (g+/-) bound to Cas9 nickase, RNP(g+/-). Protoplasts were harvested after over 48 hours incubation at the Zeitgeber time points ZT12 and ZT20. Approximately 60% of protoplasts were viable at the time of harvest based on FDA staining (Figure D) . RT-qPCR analysis showed that GmELF3-1 expression levels were moderately upregulated at ZT12: with RNP(g+) by 3-fold, with RNP(g-) by 1.5-fold, and with RNP(g+/-) by 2.5-fold compared with the negative control (Figure I) . This observation was puzzling because we expected our CRISPR experiments would induce a single nucleotide mutation or a small indel that would have little or no effects on GmELF3-1 mRNA abundance. However, it is possible that mutagenized GmELF3-1 may affect expression of its target gene(s) controlling GmELF3-1 , resulting in the observed upregulation of GmELF3-1 itself when GmELF3-1 normally is not expressed. Expression levels of the inferred target genes were observed in these CRISPR-treated protoplasts (Figures J-L) . GmPRR7-1 was significantly upregulated at ZT12: with RNP(g+) by 3-fold, with RNP(g-) by 5-fold, and with RNP(g+/-) by 7-fold compared with the negative control (Figure J) , while GmPRR7-2 showed varying expression levels at ZT12: downregulation with RNP(g+) by 1.2-fold, upregulation with RNP(g-) by 1.6-fold, and downregulation with RNP(g+/-) by 1.1-fold (Figure K) . GmCOL1b was significantly upregulated at ZT12 with RNP(g+) by 2.7-fold, RNP(g-) by 1.3-fold, and RNP(g+/-) by 2.1-fold (Figure L) . In the evening at ZT20, GmELF3-1 expression levels were generally maintained with minor down- or up-regulation: with RNP(g-) by 0.9-fold and RNP(g+/-) by 1.6-fold, respectively (Figure M) . GmPRR7-1 was moderately downregulated with RNP(g+) by 0.8-fold, RNP(g-) by 0.9-fold, and RNP(g+/-) by 0.9-fold at ZT20 (Figure N) . Similarly, GmPRR7-2 showed downregulation by 1.4-fold with RNP(g+), RNP(g-), and RNP(g+/-) (Figure O) . On the other hand, GmCOL1a showed significant upregulation with RNP(g+) by 7-fold, RNP(g-) by 6-fold, and RNP(g+/-) by 2-fold at ZT20 (Figure P).

These observations demonstrate critical regulatory interactions of GmELF3-1 with GmPRR7-1, GmPRR7-2, and GmCOL1a at varying times of the day. Two major implications of our results are a feedback regulation of PRR genes by the EC is likely conserved in the circadian clock of soybean, and that the EC controls GmCOL1 expression directly or indirectly. Limitations of this study include minimal sampling time points of the CRISPR experiment. Soybean’s circadian clock and flowering genes controlled by the clock exhibit daily oscillation patterns and their expression levels change quickly, thus using 2 time points may not capture an accurate regulation of rhythmic gene expression. In addition, experimentally determining whether a transcription factor/regulator is a transcriptional activator or a repressor is a difficult problem, especially when actions of a transcription factor change quickly within a short period of time. We cannot rule out a possible indirect regulation through multiple feedback loops within the circadian clock gene network that may mislead interpretations of gene regulatory interactions in our study. Moreover, our CRISPR approaches will require further verification for successful mutagenesis and for induced loss-of-function of GmELF3-1. Regardless, this work aids in characterizing the roles of GmELF3-1 and provides a framework in network inference and experimental verification of gene regulatory networks.

Plant Growth Condition and Sampling

The Glycine max accession PI 518671 (Williams 82) cultivar seeds were grown in Sunshine Mix #4 Professional Growing Mix with Mycorrhizae and Vermiculite in a 6:1 respective ratio. Plants were grown in a controlled growth chamber under long day conditions (LD, 14h light/10h dark) at 30°C/28°C. Soil conditions were monitored daily, establishing moisture was maintained between 40-50% and pH at 7. Soybean seedlings 4-5 days post-germination that had fully expanded unifoliate leaves were used for further analysis.

Plasmid Construction

The GmELF3-1 cDNA was cloned and subcloned into the p2GFW7 vector harboring the GFP gene, driven by the 35S promoter (VIB-UGENT Center for Plant Systems Biology; https://gatewayvectors.vib.be/collection/p2fgw7 ) as described previously (Alcantara et al., 2022) . Further information about the p2GFW7-ELF3-1 plasmid can be inquired by contacting the corresponding author Yoshie Hanzawa (yoshie.hanzawa@csun.edu).

RNP Assembly

This study used two gRNA, designed on Benchling ( https://www.benchling.com/ ), and used as follows: single guide RNA (sgRNA) on the forward strand of Exon 1 of GmELF3-1, sgRNA on the reverse strand of Exon 1 of GmELF3-1, and paired gRNAs on the forward and reverse strands of Exon 1 of GmELF3-1 spaced 20 nucleotides apart. Designed gRNAs were ordered from IDT (https://www.idtdna.com/) to be synthesized as sgRNAs. The RNP complex was produced by mixing the appropriate S.p. Cas9 enzyme (Cas9-nuclease for sgRNA and Cas9-nickase for paired gRNAs) with one or more sgRNAs in an equimolar amount (1:1 ratio) in Cas9 Dilution Buffer (30 mM HEPES, 150mM KCl, pH 7.5). The mixture was incubated for 10 minutes at room temperature for RNP complex formation and used for protoplast delivery in vivo.

Protoplast Isolation

The protoplast isolation followed the procedures described previously (Wu and Hanzawa, 2018; Alcantara et al., 2022) with minor modifications. Soybean mesophyll cells from unifoliate leaves (7d after sowing) were separated by the leaf-tape method (Wu et al. 2009; Alcantara et al., 2022) . The cells were incubated in 10mL of an enzymatic digestion solution (0.02M MES pH 5.7, 1.5% w/v Cellulase R-10, 0.50% w/v Macerozyme, 0.20% w/v Pectolyase Y-23, 0.1M CaCl ₂ , 0.1% BSA (7.5% stock), and 0.4M w/v D-Mannitol) under low light at 22 ^º C with gentle agitation of 100rpm. Digestion was quenched by the addition of W5 (154 mM NaCl, 125 mM CaCl ₂ , 5 mM KCl, 2 mM MES pH 5.7) and filtered by a 70mm nylon mesh. Protoplast cells were resuspended in prechilled W5 to a final concentration of 2 x 10 ⁵ mL ^-1 , predetermined by hemocytometer. Succeeding an ice bath incubation for 30 minutes, protoplasts were resuspended in MMg (4 mM MES pH 5.7, 400 mM D-Mannitol, 15 mM MgCl ₂ ) to a final concentration of 2 x 10 ⁵ mL ^-1 .

Protoplast Transfection for Overexpression

For overexpression transfection assay, 20 μg of pGFW7-ELF3-1 plasmid was added to 10 ⁵ mL ^-1 protoplasts, mediated by an equal volume of freshly prepared PEG solution (20% w/v PEG4000, 400 mM D-Mannitol, 100 mM CaCl ₂ ) and immediately mixed by inversion. The mixture was incubated for 12 minutes at room temperature, after which, 8 mL of W5 was slowly added to quench transfection. Pellets were collected after 100 x g centrifugation for 5 minutes and resuspended in WI solution (4 mM MES pH 5.7, 500 mM Mannitol, 20 mM KCl) to a final concentration of 10 ⁵ mL ^-1 . A 6-well tissue culture plate was coated with 50% v/v sterile calf serum where 1 mL of transfected protoplasts were transferred and incubated at 22 ^º C overnight in the dark. At roughly 24 hours post transfection, a sample of protoplast cells were stained with fluoresceine diacetate (FDA) and imaged under Leica confocal microscope. GFP signals were verified by confocal laser microscopy at the time of harvest.

Protoplast Transfection for Targeted Mutagenesis

For in vivo CRISPR/Cas9 transfection assay, 25 μL RNP complex (1 mM sgRNA: 1 mM Cas9) was added to 1 mL aliquot of 10 ⁵ mL ^-1 protoplasts, mediated by an equal volume of freshly prepared PEG solution, and immediately mixed by inversion. The transfection mixture was incubated for 30 minutes in the dark at room temperature, after which, 8 mL of W5 was slowly added to quench transfection. Pellets were then collected after 100 x g centrifugation for 5 minutes and resuspended in WI solution to a final concentration of 10 ⁵ mL ^-1 . Transfected cells were transferred to a 50% v/v sterile calf serum pre-coated tissue culture plate and incubated at 25 ^º C for 48 hours in the dark. At roughly 48 hours post transfection, a sample of protoplast cells were stained with FDA and imaged under confocal microscopy.

RNA Extraction, cDNA Synthesis, and Quantitative RT-PCR

Total RNA was extracted from harvested protoplast cells and invasive genomic DNA removed with DNAse using the Invitrogen RNA kit (Invitrogen, CA, USA), per the manufacturer’s protocol. First-strand cDNA was synthesized using the Reverse Transcriptase Kit, per the manufacturer’s protocol and diluted to 1:20 for RT-qPCR application. RT-qPCR was performed using the QuantStudio3 with samples containing the fluorogenic probe SYBR™ Green. The following amplification settings were performed: 2-minute hold at 50°C, 10 minutes at 95°C, a PCR of 15 seconds at 95°C, and ending with 1 minute at 60°C where amplification was captured. These settings were set to repeat for 50 cycles. The Ct values were then analyzed on Thermo Fisher server (https://apps.thermofisher.com/). Relative gene expression was normalized against the housekeeping gene GmPBB2 and calculated as 2 ^-DCt as described in Wu et al 2014 and Livak and Schmittgen 2001 (Livak & Schmittgen, 2001; Wu et al., 2014) .

Table 1. Soybean accession.

Table 2. Fusion construct of GmELF3-1 and Gene ID.

Table 3. Guide RNA sequences and targeted cut site.

Table 4. Cas endonucleases and their respective recognition site.

Table 5. RT-qPCR primers targeting CDS sequences of selected targeted genes using the obtained sequences from https://phytozome-next.jgi.doe.gov/ in Williams82.a2.v1 .