In this study, soybean [G. max (L.) Merrill cv. Williams 82] and Agrobacterium rhizogenes strain K599 were used for the hairy root transformation. The hairy root transformation procedure was previously described (Wang et al., 2014) with some modifications. The positive transformed composite plants were cultured in a low-nitrogen nutrient solution for 5 days for recovery; after that, the plants were transferred to vermiculite for inoculation with a suspension of B. japonicum strain USDA110 (30 ml, OD600 = 0.08). Nodule numbers were evaluated at 28 DAI (days after inoculation).

Genome data, protein sequence, and genome annotation files of Glycine max were downloaded from the Phytozome database (https://phytozome-next.jgi.doe.gov/). Protein families in the PANTHER database (http://www.pantherdb.org) were applied to download the Hidden Markov model (HMM) of the NOT transcription complex related family (PTHR12603). The soybean genome was searched using hmm search in HMMER (http://hmmer.org/) to identify NOT4 candidate genes (the screening threshold was 1.0e−10). The resulting candidate sequences were submitted to Intel Pro Scan (https://www.ebi.ac.uk/interpro/search/sequence-search) to check the PTHR12603 structure domain. Soybean GmNOT4 family members were named according to their location distribution on soybean chromosomes.

NOT4 protein sequences from Medicago truncatula, Phaseolus vulgaris, Arabidopsis thaliana, soybean [G. max (L.)], and rice (Oryza sativa) NOT4 were downloaded from the Phytozome database (https://phytozome-next.jgi.doe.gov/). The phylogenetic tree of the GmNOT4 gene family members of the five species was constructed by MEGA-X software and calculated by the NJ (neighbor-joining) method. The parameters were set to self-expanding and repeated 1,000 times. Chromosome distribution was analyzed using TBtools.

The MEME Suite online software (https://meme-suite.org/meme/) was used for conserved motif analysis, and the parameter was set to 10. The gene structure information of GmNOT4s was extracted from the soybean gene information GFF file, and the gene structures were visualized by TBtools.

RNAprep Plant Plus Trizol Kit was used to extract RNA from collected samples, and the first-strand cDNA was synthesized using the Super Mix Kit (Hifair II 1 strand cDNA Synthesis SuperMix, gDNA digester plus) (Yeasen Biotech Co. Ltd., Shanghai, China). qPCR was performed using SYBR Green JumpStart Taq ReadyMix (Sigma-Aldrich). GmCYP2 was used as an internal control (Jian et al., 2008). The primers used in this study are shown in Table S1 .

For the GmNOT4-1 overexpression construct, the GmNOT4-1 CDS fragment was inserted into the pCAMBIA1300-GFP vector through seamless cloning using the BamH1 restriction site; for the GmNOT4-1-RNAi construct, the GmNOT4-1 CDS fragment was ligated into the pDONOR207 entry vector, and then the target sequence was cloned into the pK7GWIW-GFP vector through the gateway LR reaction. For the GmNOT4-1 CRISPR/Cas9 knock-out construct, sgRNAs were designed using the software Crispr-P (http://cbi.hzau.edu.cn/crispr/), and the top two reliable sgRNAs, CAAGGTGCGGTGAAGAGCA and TCGTCCTCTTCGCCTCTGC, were selected. Then, vector pCBC-DT1T2 was used as a template to clone the two CRISPR fragments, and the two obtained products were inserted into vector pKSE401-GFP.

GraphPad Prism 7 (GraphPad Software, La Jolla, CA, USA) was used to analyze the data in this study. A Student’s t-test was performed to generate P-values. The statistical differences are marked as follows: *, P <0.05; **, P <0.01; ***, P <0.001.

Seven GmNOT4s were identified from the soybean genome using BLAST and PANTHER searches and named GmNOT4-1–GmNOT4-7 according to their positions on six chromosomes ( Figure S1 ). The chromosomal distribution of NOT4s indicated that they were dispersed on chromosomes 5, 10, 12, 15, and 17. Except for GmNOT4-4 and GmNOT4-5, which are on chromosome 13, each of the other five chromosomes contains one family member. The physical and chemical properties of GmNOT4s were analyzed, and it was found that the amino acid residues encoded by seven GmNOT4s ranged from 232 (GmNOT4-1) to 1,046 (GmNOT4-6), and the corresponding molecular weight ranged from 35,386.84 (GmNOT4-1) to 115,070.36 (GmNOT4-6) Da. The theoretical isoelectric points of seven GmNOT4 family members ranged from 4.76 (GmNOT4-7) to 6.40 (GmNOT4-4), all of which belonged to weakly acidic proteins ( Table S2 ). The total mean hydrophobic index was less than 0, indicating that they were hydrophilic proteins. To understand the functional characteristics and evolutionary relationship of the NOT4 gene family, NOT4 protein sequences in soybean, M. truncatula, common bean, A. thaliana, and rice were retried to construct the phylogenetic tree. The results showed that the members of the NOT4 gene family were divided into three subgroups (Groups I–III), in which groups I and III contained 10/11 members, and group II had three members. In the same subgroups, NOT4 proteins cluster together within species ( Figure 1 ). Soybean NOT4 family members were evenly distributed in three groups, with three members in Group I, two members in Group II, and two members in Group III.

To get a hint about the function of NOT4s, we first performed conserved motif analysis and gene structure analysis. There are 10 conserved motifs obtained from the GmNOT4 protein, and the number of motifs of GmNOT4s in different subgroups varied significantly, ranging from 2 to 10. Three members of Group I contain 10 motifs; there were eight motifs in GmNOT4-2 and seven for GmNOT4-4 in Group II and the members of Group III contain only two motifs. Among the 10 motifs, Motif1 was the most conserved one and was present in all GmNOT4 proteins ( Figure 2 ). The variation in the types and amounts of conserved motifs in GmNOT4s reflects the functional diversity of these proteins. In addition, we have confirmed the conservation of motifs in NOT4 proteins in other plant species; we found these motifs are consistently present among Arabidopsis, rice, Medicago, and the common bean ( Figure S2 ).

Diagram of GmNOT4s gene structures shows the exon number variation of GmNOT4s ranged from 3 to 13 ( Figure 2 ). Combined with phylogenetic tree analysis, we found that genes with close genetic relationships mostly had similar gene structures. For example, all Group I GmNOT4s contained 12 exons, while Group III GmNOT4s contained three exons.

Most transcriptional factors function in gene expression regulation by binding specific cis-acting elements in gene promoters to modulate gene expression. To predict the functions of GmNOT4 genes, 2-kilobase pairs upstream of the translational initiation site were selected as the promoter sequence and submitted to PlantCARE online software to predict the cis-acting elements ( Figure S3 ). A total of 63 cis-acting elements were identified in seven GmNOT4 promoters, while 54 cis-acting elements were related to plant hormones, stress, growth, and development. Firstly, core cis-acting elements exist in almost all promoters and consist of AT-TATA-box, CAAT-box, TATA-box, and TATA. The second group contains 26 cis-acting elements related to plant growth and development; all GmNOT4 promoters contain photoreactive elements, which coincide with the CCR4–NOT complex response to far-red light. However, only GmNOT4-1 and GmNot4-7 contain cis-regulatory elements for flavonoid biosynthesis, which play a vital role in rhizobia attraction and symbiosis construction. Plant hormones are important for legume nodulation; 13 kinds of plant hormone related cis-acting elements were harbored in GmNOT4 promoters; all GmNOT4 promoters contained abscisic acid and ethylene-related cis-acting elements; and auxin response elements (TGA-elements), gibberellin response elements (P-box, GARE-motif) existed in most GmNOT4s promoters. Low oxygen conditions were very important for nitrogen fixation in nodules, and we found that among the stress response cis elements, GmNOT4s harbored the most anaerobic induction and stress-related cis elements. The specific cis-acting element distributions in GmNOT4 promoters suggest their different biological functions.

To investigate the potential function of NOT4s during soybean nodulation, we examined the temporal and spatial expression patterns of GmNOT4s by qRT-PCR.

We found that all but NOT4-7 were highly expressed in root nodules ( Figures 3A–F ). Further study has shown that the expression of GmNOT4-1, GmNOT4-2, GmNOT4-3, GmNOT4-4, and GmNOT4-7 was dramatically stimulated by rhizobium infection at 1 DAI ( Figures 3H–K, N ). The expression of GmNOT4-1 and GmNOT4-2 was induced by rhizobium USDA110 while GmNOT4-3 was repressed at 3 DAI ( Figure 3H–J ). The expression of GmNOT4-1, GmNOT4-5, and GmNOT4-6 was induced by rhizobium, while GmNOT4-4 and GmNOT4-7 were repressed at 6 DAI ( Figures 3H, L, M ). These results show that all GmNOT4s responded to rhizobium at least at one point. Combining the above tissue expression pattern to prove the function of GmNOT4s in legume root nodule symbiosis, we selected GmNOT4-1 for further study, whose expression was induced at all checkpoints within a short period of inoculation (1, 3, and 6 DAI).

To genetically explore whether GmNOT4-1 is involved in the regulation of soybean nodulation, overexpression and knockout/knockdown analyses of GmNOT4-1 were performed using the hairy root transformation system. Firstly, we constructed 35S:GmNOT4-1 and obtained overexpressing GmNOT4-1 (GmNOT4-1-OX) roots by qPCR analysis ( Figure 4A ). The effects of GmNOT4-1 overexpression on the early and late stages of nodulation were evaluated at 1, 6, and 28 days after inoculation (DAI) ( Figure S4 ). Firstly, the expression of rhizobial infection-related genes, including GmRPG, GmNPL, GmVPY, GmCYCLOPS, and GmSCARN, was validated. As shown in Figure S5 , all rhizobial infection-related genes we checked in this study were significantly inhibited in GmNOT4-1 overexpressing roots. Then, we found the number of root hairs showing deformation was markedly decreased in GmNOT4-1-OX hairy roots at 6 DAI ( Figure S6 ). Finally, the nodule numbers were quantified at 28 DAI. The average nodule number per GmNOT4-1 overexpressed root was about 3.6, while empty vector (EV) control roots produced an average of 11.8 nodules per root, with an approximately 69% reduction by GmNOT4-1 overexpression. These data suggest that GmNOT4-1 plays a negative role in regulating soybean nodulation ( Figures 4B, C ).

To further validate the function of GmNOT4-1 during nodulation, we constructed the NOT4-1-RNAi to analyze the soybean nodulation phenotype when the expression of GmNOT4-1 was reduced ( Figure S4 ). As shown in Figures 5B, C , transgenic roots harboringGmNOT4-1-RNAi produced fewer nodules (6.9 nodules per root) than empty vector (EV) control roots (19.5 nodules per root), so that the nodule number per GmNOT4-1-silenced root was reduced by approximately 64.6%. Further, the CRISPR/Cas9 system was also applied to knock out GmNOT4-1 in soybean hairy roots ( Figures S8 , S9 ), and the gene editing was verified by sequencing ( Table S3 ). The expression of rhizobial infection-related genes, including GmRPG, GmNPL, GmVPY, GmCYCLOPS, and GmSCARN, was significantly inhibited in GmNOT4-1-KO roots. The number of deformed root hairs was markedly decreased in GmNOT4-1 edited knock-out hairy roots at 6 DAI compared to the vector control ( Figure S6 ). Finally, we found that the GmNOT4-1 edited roots produce functional nodules ( Figure S7 ) with a significantly reduced nodule number, which showed the same result as the GmNOT4-1-RNAi roots ( Figures 5B, C ).

Combined with the overexpression results, it is suggested that homeostasis of the expression of GmNOT4-1 is critical for the regulation of soybean nodulation.

Nodule number was mainly modulated by both the NF signaling and AON signaling pathways. Given the phenotype that either overexpression or knockdown (out) of GmNOT4-1 significantly decreased soybean nodule number, it is worth checking whether GmNOT4-1 regulates nodulation through the above pathways. Thus, we examined the expression level of several nodulation marker genes in soybean, including NF signaling pathway genes ENOD40, GmNINa, NSP1, HAP2-1, and HAP2-2, and AON signaling pathway genes GmRIC1/2. In general, we found that either overexpression or silence of GmNOT4-1 resulted in reduced expression of nodulation marker genes compared with the empty vector control roots ( Figure 6 ).

Legumes plants can specifically interact with the phylogenetically diverse group of soil bacteria-rhizobia to form nodules. However, symbiotic nitrogen fixation is a highly energy-intensive biological process; thus, host legumes have evolved a root–shoot–root long-distance auto-regulation of nodulation (AON) system to refine the number of nodules (Patra et al., 2017; Ren et al., 2019). NODULE INCEPTION (NIN) induces the expression of CLE ROOT SIGNAL1 (CLE-RS1) and CLE-RS2 to activate AON to inhibit excessive nodulation (Chun et al., 2005). NIN also modulates almost all nodulation processes, including nodule initiation, nodule organogenesis, and nitrogen fixation. Thus, NIN can act as a bifunctional transcription factor, fine-tuning legume nodulation. Interestingly, in this study, we observed that GmNOT4-1 also function as a biofunctional regulator in nodulation, both overexpression and CRISPR/Cas9 knock out of GmNOT4-1 inhibited the number of nodules in soybean. Except for GmNOT4-1, we found all GmNOT4s responded to rhizobium at least at one time point ( Figure 3 ). We do not exclude the possibility that different family members might have a role in a particular stage of root nodule symbiosis.

In previous studies, a series of transcriptional factors and microRNAs were identified (NIN, NSP1/2, miR172c, etc.) that modulate the expression of some key nodulation regulators (ERN1, ENOD40, RICs, etc.) and finally legume nodule number. In eukaryotic cells, except for transcriptional factors and miRNAs, the multiprotein complex CCR4–NOT also plays a vital role in regulating gene expression via the shortening of poly(A) tails of messenger RNA. In this study, we identified GmNOT4-1, one component of the soybean CCR4–NOT complex, whose expression homeostasis is important for soybean nodule initiation and subsequent nodule number ( Figures 4 , 5 ; S4 - S6 ). We found that either overexpression or silencing of GmNOT4-1 resulted in reduced expression of nodulation marker genes compared with empty vector control roots ( Figure 6 ). RIC1/2 was induced by nodule primordial formation to inhibit excessive nodulation; GmNOT4-1 may function before nodule primordial formation; thus, the resulting RIC1/2 induction and AON signaling were blocked by the overexpression or silencing of GmNOT4-1. Further study is needed to clarify the other components of the CCR4–NOT complex, including GmNOT4 family members in nodulation, and construct a direct link between the CCR4–NOT complex and nodulation signaling pathway genes. In addition, GmNOT4-1 encodes a RING domain containing an E3 ligase; it may function by adjusting substrate protein levels. Identification of the target of GmNOT4-1 can further unveil its role in the process.