Genome-Wide Analysis of DA1-Like Genes in Gossypium and Functional Characterization of GhDA1-1A Controlling Seed Size

Cotton (Gossypium spp.) is an economically important crop grown for natural fiber and seed oil production. DA1 is a ubiquitin receptor that determines final seed and organ size by restricting the period of cell proliferation. In the present study, we identified 7 DA1-like genes each in cultivated tetraploid (AADD) G. hirsutum and G. barbadense, and 4 and 3 DA1-like genes in their ancestral diploid G. arboreum (A2A2) and G. raimondii (D5D5), respectively. The 7 GhDA1 genes were confirmed to be distributed on four At and three Dt subgenome chromosomes in G. hirsutum. GhDA1-1A showed a high sequence similarity to AtDA1 in Arabidopsis, and they possessed the same functional domains, suggesting conserved functions. The overexpression of GhDA1-1AR301K in Arabidopsis significantly increased seed size and seed weight, indicating that GhDA1-1A is a promising target for cotton improvement. This study provides information on the molecular evolutionary properties of DA1-like genes in cotton, which will be useful for the genetic improvement of cotton.


INTRODUCTION
Organ size is one of the most important features and is regulated by complex developmental processes involving both internal and external signals (Cai et al., 2020). Seeds represent the core of plant life cycle traits and are involved in the mechanisms of plant diffusion, germination, seedling survival and overall reproductive success (Cardinal-McTeague et al., 2019;Keren et al., 2020). In contrast to the reproductive advantage of small-seeded species, the key advantage of larger seeds appears to be their tolerance to abiotic stresses such as shade or drought, and seed size is also an important agronomic trait that greatly affects crop yield (Liu et al., 2020b). The signaling pathways that affect endosperm and/or maternal tissue growth to determine seed size include the HAIKU1 (IKU), ubiquitin-proteasome, G-protein signaling, and mitogen-activated protein kinase, plant hormone, and transcription regulator pathways (Li and Li, 2016).
Cotton is an economically important crop. In cotton breeding, focusing on a higher lint percentage (a ratio between lint weight and total seed cotton weight from seed and lint) inadvertently leads to the reduction in seed size or weight (i.e., the seed index, or the weight in g of 100 cotton seeds), which is an indicator of the quality of seeds (Zhao et al., 2015a). Generally, cottonseeds with larger volume and mass tend to content more storage material and have higher vigor (Zhao et al., 2019). Previous studies have shown that plants with large seeds exhibit better traits than those with small seeds on the basis of testing the effect of seed size on cotton seedling growth. Large seeds exhibit more nutrient accumulation than small seeds, which may affect seed germination and even the growth and development of plants .
There are many factors affecting the size of a seed, including genetic factors, environment (including pests and diseases) and genotype-by-environment interactions (Wang et al., 2016). In recent years, the completion of plant genome sequencing and the construction of plant transcription factor databases have greatly advanced research progress related to transcription factors involved in seed development and their molecular regulation mechanisms (Wang et al., 2013). It has been shown that multiple components of the ubiquitin pathway are involved in the regulation of seed size: for example, E3 ubiquitin ligases, the proteasome, and ubiquitination modification play important roles in regulating seed size. Several genes related to plant organ size have been cloned and functionally verified, such as EBPI (Horváth et al., 2006), DA1  and DA2 (Xia et al., 2013). There are three different protein domains in AtDA1, including two UIMs proximal to the N-terminus, one zinc-binding LIM domain and one DA1-like functional domain next to the C-terminus (Peng et al., 2015). The DA1 genes of Arabidopsis thaliana encode a ubiquitin receptor. When the conserved arginine (R) at the 358th position in the AtDA1 protein sequence is mutated to lysine (K), the resulted mutant produces larger seeds than wild-type plants, indicating that the DA1 gene negatively regulates both seed and tissue size . The overexpression of AtDA1 R358K can increase the rapeseed yield in Brassica napus (Wang et al., 2017). In addition, the overexpression of the mutant ZmDA1 (Zmda1) or ZmDAR1 (Zmdar1) gene improves sugar import in sink organs and starch synthesis in maize kernels (Xie et al., 2018).
However, DA1-like proteins without conserved mutation may play different roles in different plant species. Different cis-acting regulatory elements in the promoter sequences of Arabidopsis thaliana and rice respond to different hormones (such as abscisic acid and salicylic acid) and stress signals (such as heat stress and drought stress) (Li et al., 2009). As a receptor for E3 ubiquitin ligase, DA1-like genes also play important roles in regulating ABA signaling pathways to participate in drought stress . In Glycine soja, constitutive GsoDA1 expression can improve salt resistance with no effect on seed size (Zhao et al., 2015b). The overexpression of TaDA1 decreased the size and weight, while the downregulation of TaDA1 might be effective in improving grain yields in wheat (Liu et al., 2020a).
Till now, there is no report whether DA1-like genes from cotton regulate seed size. In the present study, sequence characteristics and expression patterns of DA1-like genes were analyzed in cotton. GhDA1-1A is the homologous gene of AtDA1 in cotton. Previous studies had demonstrated that seed size of Atda1-1 mutant increased . We wanted to know whether over-expression of mutated GhDA1-1A would have a similar phenotype. Then, GhDA1-1A R301K sequence was designed containing a single-nucleotide G-to-A transition at 902nd nucleotide site of GhDA1-1A (GH_A01G1154), which was predicted to cause an arginine-to-lysine change at the 301st amino acid site. GhDA1-1A R301K was transformed into Arabidopsis thaliana ecotype Col-0. The relationship between GhDA1-1A R301K and seed size was elaborated.

Sequence Retrieval and Identification of DA1-Like Genes in Gossypium
The DA1 genome sequences and protein sequences of Arabidopsis and Glycine max were retrieved from The Arabidopsis Information Resource (TAIR release 10) 1 and SoyBase 2 , respectively. At the CottonGEN website 3 , we downloaded the genome sequences of G. arboreum (A2, CRI_V1.0) (Du et al., 2018), G. raimondii (D5, JGI v2_a2.1) (Paterson et al., 2012), G. hirsutum acc. TM-1 (AD1, ZJU) (Hu et al., 2019), and G. barbadense acc.3-79 (AD2, ZJU) (Hu et al., 2019). The candidate DA1 protein sequences were used as the query sequences, and BlastP (E-value = 10 × 10 −5 ) searches were performed in the above genome databases. The default parameter settings were used. Then, the candidate sequences were submitted to Pfam 4 (Finn et al., 2014) and further verified in the SMART 5 (Letunic et al., 2015) database to determine whether the candidate sequence contained one zinc-binding LIM domain and one DA1-like functional domain next to the C-terminus. Multiple sequence alignments of all DA1 full-length protein sequences were performed using Clustal X2.0 software (Larkin et al., 2007) with the default values. Subsequently, the neighbor-joining (NJ) method was employed to construct phylogenetic trees by using MEGA v7.0 software (Tamura et al., 2013) with the pairwise deletion option, Poisson correction model and uniform rates. The statistical reliability of the phylogenetic tree was evaluated using the bootstrap method with 1000 repeats. Furthermore, the theoretical molecular weight (MW) and isoelectric point (pI) of the DA1like proteins were predicted using the online ExPASy tool 6 (Bjellqvist et al., 1994).

Chromosomal Location
All DA1-like genes of G. raimondii, G. arboreum, G. hirsutum, and G. barbadense were mapped on the corresponding chromosomes according to their positional information provided in the genome annotation document. The chromosome location of the cotton DA1-like genes was illustrated with MapChart v2.2 software (Voorrips, 2002).

Genetic Structure Analysis and Protein Domain Detection
Tbtools software (Chen et al., 2020) was used to predict DA1like gene structure. The NCBI database 7 was used for the identification of DA1-like protein domains.

Plant Materials and Growth Conditions
Upland cotton TM-1 was used for gene cloning and spatiotemporal quantitative real-time PCR (qRT-PCR) analysis and was grown at Anyang (AY), Henan, China. Roots, stems and leaves were collected at the seedling stage, and fibre and ovule samples were collected at 0, 5, 10, 20 and 30 days post-anthesis (DPA) for RNA extraction. Each experiment was independently repeated in triplicate. Arabidopsis thaliana ecotype Columbia-0 (Col-0) was used as the wild-type line. The RNA-seq data of G. hirsutum acc. TM-1 (Hu et al., 2019) were used to identify the expression levels of GhDA1-like genes. The expression profiles of AtDA1 and AtDAR1-7 genes were extracted from the Arabidopsis eFP Browser database 8 were used to identify the expression levels of AtDA1 and AtDAR1-7 genes. Seeds were surface-sterilized with 10% (v/v) household bleach for 10 min, washed at least three times with sterile water, stratified at 4 • C for 3 days in the dark, dispersed on Murashige and Skoog 7 https://www.ncbi.nlm.nih.gov/Structure/bwrpsb/bwrpsb.cgi 8 http://bar.utoronto.ca/efp/cgi-bin/efpWeb.cgi (MS) medium with 0.88% agar, and then grown at 22 • C. Plants were grown under long-day conditions (16-h light/8-h dark) at 22 • C.

Constructs and Transformation
The complete coding sequence of GhDA1-1A was amplified with gene-specific primers (Supplementary Table 1). GhDA1-1A R301K DNA sequencing revealed a single-nucleotide G-to-A transition in the GhDA1-1A gene (GH_A01G1154), which was predicted to cause an arginine-to-lysine change at the 301 st amino acid site. GhDA1-1A-F and mGhDA1-1Ar as well as mGhDA1-1Af and GhDA1-1A-R bridge PCR primer pairs were used in the first PCR amplification to introduce a G to A single-base mutant in two fragments. GhDA1-1A-F and GhDA1-1A-R were used to join the two fragments together. The complete coding sequence of GhDA2 was amplified with gene-specific primers (Supplementary Table 1).
The 35S:GhDA1-1A R301K construct was generated using a PCR-based system. The specific primers used to produce the 35S:GhDA1-1A R301K construct were 35S:GhDA1-1A R301K -F and 35S:GhDA1-1A R301K -R (Supplementary Table 1). The PCR products were subcloned into a pBI121 vector digested with BamHI and SacI using the ClonExpress R II One Step Cloning Kit (Vazyme, Nanjing, China). The 35S:GhDA1-1 plasmid was introduced into Col-0 plants using Agrobacterium GV3101, and transformants were selected on medium containing kanamycin (50 mg/L). The progeny of the transformants showed approximately 3:1 segregation of live and dead phenotypes, and homozygous lines of the T3 generation were used for further analysis.

Bimolecular Fluorescence Complementation Assays
The CDS of GhDA1-1A was cloned into the BamHI and SalI sites of the pSPYNE vector and the CDS of GhDA2 was cloned into the BamHI and SalI sites of the pSPYCE vector with a ClonExpress II One Step Cloning Kit (Vazyme, C112-01), respectively. The constructs were transferred into Agrobacterium GV3101 cells.

Luciferase Complementation Imaging Assay
The CDS of GhDA1-1A was cloned into the SacI and SalI sites of the pCAMBIA1300-nLUC vector with a ClonExpress II One Step Cloning Kit (Vazyme, C112-01). The CDS of GhDA2 was cloned into the KpnI and SalI sites of the pCAMBIA1300-cLUC vector with a ClonExpress II One Step Cloning Kit (Vazyme, C112-01). The constructs were transferred into Agrobacterium GV3101 cells. Agrobacterium cells were grown in LB medium containing 1% (m/v) peptone, 0.5% (m/v) FIGURE 1 | Phylogenetic relationships of DA1-like genes from Gossypium, Arabidopsis, and soybean. Phylogenetic analysis was performed using the neighbor-joining method with 1,000 replicates. The DA1-like genes from Gossypium, Arabidopsis, and soybean are marked with red, blue, and purple triangles, respectively. Each group is indicated with a specific color.

Genome-Wide Identification and Phylogenetic Analysis of DA1-Like Genes in Gossypium
To identify all the DA1-like proteins in cotton, BLASTP searches were performed against the diploid cotton (G. raimondii and G. arboreum) and tetraploid cotton (G. hirsutum and G. barbadense) protein databases using the AtDA1 and AtDAR1-7 protein sequences of Arabidopsis as queries. The candidate genes were further subjected to analysis in the NCBI database to identify their protein domains. After a strict two-step selection process, 4 deduced DA1-like genes were identified in G. arboreum, along with 3 in G. raimondii, 7 in G. barbadense, and 7 in G. hirsutum. More information about DA1-like genes, such as identifiers and predicted properties of DA1-like proteins, is listed in Table 1.
To assess the phylogenetic relationships of DA1-like genes among four cotton species, Arabidopsis and soybean, a comprehensive phylogenetic tree was constructed using the NJ method (Figure 1). In accordance with previous studies (Zhao et al., 2015b), the DA1-like genes could be divided into two groups. There were 15 members in DA1-like group 1: 5 from G. hirsutum, 5 from G. barbadense, 3 from G. arboreum, and 2 from G. raimondii. In addition, DA1-like group 2 consisted of 6 members: 2, 2, 1, and 1 from G. hirsutum, G. barbadense, G. arboretum, and G. raimondii, respectively.

Chromosomal Distribution of DA1-Like Genes
The mapping of 21 DA1-like genes to chromosomes based on the available genomic information on the four cotton species revealed that all the DA1-like genes were evenly distributed on chromosomes. In the G. arboreum genome, 4 GaDA1s were distributed on four chromosomes (A01, A05, A10, and A12)   Figure 1A). Three GrDA1 genes were mapped to 3 chromosomes of G. raimondii: chromosome 09 (D05), chromosome 11 (D10), and chromosome 08 (D12) ( Table 1 and Supplementary Figure 1B). In the G. hirsutum genome, we found that four GhDA1s were located on At subgenome chromosomes (A01, A05, A10, and A12), while three GhDA1s genes were located on three Dt subgenome chromosomes (D05, D10, and D12) ( Table 1 and Supplementary Figure 1C). In the G. barbadense genome, we also found that four GbDA1s were located on the four At subgenome chromosomes (A01, A05, A10, and A12). The other three GbDA1s were located on Dt subgenome chromosomes (D05, D10, and D12) ( Table 1 and Supplementary Figure 1D). We speculate that DA1-like genes were conserved during diploid to tetraploid evolution.

Gene Structure and Protein Domain Analyses of DA1-Like Genes
The analysis of gene structure is a very effective method for determining gene function and can reflect the phylogenetic relationships among DA1-like genes. By comparing the GFF3 files of each DA1-like gene family member in G. hirsutum, G. barbadense, G. raimondii, and G. arboreum with their corresponding coding sequences, we evaluated the gene structure of DA1-like gene family members (Figure 2A). A common feature that can be observed is that DA1-like genes may contain more than ten exons: eleven DA1-like genes contain 11 exons, eight genes contain 12 exons and two genes contain 13 exons (Figure 2A). To better understand the similarity and diversity of DA1 proteins, their putative protein domains were predicted using the NCBI database. Previous studies have shown that UIM domains are absent in some DA1-like proteins . Our results showed that most of the DA1-like proteins contained two UIM domains, one LIM domain, and one conserved DA1-like protein domain at the C-terminus. However, GrDA1-1, GhDA1-3D, GbDA1-3D, GaDA1-4, GhDA1-4A, and GbDA1-4A exhibited only one LIM domain and one DA1like protein domain (Figure 2B). In addition, ten different conserved motifs were indicated by the MEME online program 9 (Bailey and Elkan, 1994). It was demonstrated that most of these motif sare conserved in DA1-like proteins, except in GbDA1-3A, GhDA1-3A, GaDA1-3, GrDA1-3, GbDA1-2D, and GhDA1-2D, in which motif-10 was absent. It remains to be determined whether the lack of motif-10 alters protein function ( Figure 2C).

Tissue-Specific Expression Profiles of GhDA1 Genes
To reveal the tissue-specific expression profiles of DA1-like genes in cotton and Arabidopsis, published TM-1 expression data (Hu et al., 2019) and the public Arabidopsis expression data (See Text Footnote 8) were used for analysis ( Supplementary  Figure 2A,B). GhDA1-like genes exhibited different expression 9 http://meme.nbcr.net/meme/ patterns in different tissues of TM-1 as that in Arabidopsis, indicating that GhDA1-like genes have multiple biological functions in cotton growth and development. qRT-PCR results also showed that GhDA1-1A was highly expressed in roots, stems, and leaves but presented almost no expression in the early stage of ovule development (Figure 3A). GhDA1-2 (including the expression of GhDA1-2A and GhDA1-2D) was most highly expressed in roots ( Figure 3B). GhDA1-3 FIGURE 4 | Multiple sequence alignments of amino acid sequences. (including the expression of GhDA1-3A and GhDA1-3D) and GhDA1-4 (including the expression of GhDA1-4A and GhDA1-4D) expression was highest in 30 DPA fibers, suggesting the involvement of these genes in the late period of cotton fiber development (Figures 3C,D).
GhDA1-1A and AtDA1 showed similar expression patterns and were widely expressed, indicating that GhDA1-1A has similar biological functions during cotton growth and development. 476,477,520,549,477,522, and 537 amino acids, respectively, and they share 44.31% to 63.79% identity, indicating the high conservation of these homologs. The amino acid sequence of GhDA1-1A showed the closest similarity to that of AtDA1 and contained the same functional domain (Figure 4). The overexpression of AtDA1 R358K results in large seeds and organs Weng et al., 2008), therefore, we were interested in whether GhDA1-1A could exhibit similar functions. Hence, GhDA1-1A was selected for further functional analysis.

Generation of
Based on sequence alignment, mutation site 358 in AtDA1 is equivalent to conserved amino acid 301 in the DA1-like functional domain of GhDA1-1A (Figure 4). A corresponding single-nucleotide mutation was designed as a G-to-A transition in the GhDA1-1A gene to cause an arginine-to-lysine change in the conserved amino acid at position 301 (Figure 4). Arabidopsis plants overexpressing the sequence were generated and preliminarily identified by PCR (Figure 5A). qRT-PCR was performed to further assess relative expression levels using cDNA from three different transgenic lines and WT plants as templates. A total of three lines with high expression levels were obtained and used for further studies (Figure 5B).

Overexpression of GhDA1-1A R301K Increases Seed Size and Seed Weight
To evaluate the applicability of GhDA1-1A R301K to transgenic breeding for seed size, we characterized the phenotypes of GhDA1-1A R301K transgenic Arabidopsis at different developmental stages. The seed size of the transgenic lines was  examined and was shown to be significantly increased compared to that of Col-0 (Figures 6A-D). After the seeds germinated, the cotyledon area of 9-day-old seedlings was further measured. The results showed that the seedling size of the transgenic lines was greater than that of Col-0 (Figures 6E,F). Moreover, the transgenic lines produced large flowers (Figures 6G,H), indicating that the overexpression of GhDA1-1A R301K influenced flower development and the seed mass of line 2 was increased to 128% of the Col-0 seed mass (Figures 6I-K). However, there were no differences in flowering timing, frequency, or duration (data not shown). Therefore, the overexpression of GhDA1-1A R301K increased seed size, seed weight, cotyledon size and flower size.

GhDA1-1A Interacts With GhDA2
DA1 is a ubiquitin receptor that interacts with the E3 ubiquitin ligase DA2 to regulate seed and organ size in Arabidopsis and maize (Xia et al., 2013;Liu et al., 2020a). We were interested in the relationship between GhDA1-1A and GhDA2 (GH_D05G3532) in G. hirsutum.
To determine the subcellular localization of GhDA1-1A and GhDA2, the GhDA1-1A and GhDA2 construct fused with green fluorescent protein (GFP) were infiltrated into tobacco leaves, respectively. The epidermal cells of tobacco leaves were observed under a confocal microscope (Olympus). In contrast to the pattern observed in plants carrying the empty vector, the GFP signal of GhDA1-1A was found in membrane. Besides, the 35S-GhDA2-GFP signal was not only detected in the nucleus but also found in the membrane (Figure 7A).
BiFC assays were performed in tobacco leaves to test whether GhDA1-1A could interact with GhDA2. The results showed the fusion proteins of pSPYNE-GhDA1-1A and pSPYCE-GhDA2 co-localized in the membrane ( Figure 7B). Then, the firefly luciferase complementation imaging assay (Chen et al., 2008) was used to provide further evidence for the interactions between GhDA1-1A and GhDA2 using agroinfiltration ( Figure 7C). The results indicated that GhDA1-1A may interact with GhDA2.

DISCUSSION
Seed size is a key agronomic trait that strongly affects the grain yield of plants (Liu et al., 2020b). However, in cotton, fibers are the main economic product. Given that fiber cells develop from the cotton seed epidermis, seed development strongly influences fiber growth, yield, and quality (Ruan, 2013). Cotton seeds are also the sixth-largest source of vegetable oil worldwide . Large-seeded plants accumulate abundant nutrients to increase stress tolerance, whereas small-seeded plants flourish via dispersal and colonization (Moles et al., 2005). The mechanisms controlling seed size are thus of interest in both agriculture and biology.
In recent years, with the completion of various plant genome sequences, DA1-like genes in Arabidopsis, Brassica napus, Zea mays, and Triticum aestivum have been cloned and functionally verified Wang et al., 2017;Xie et al., 2018;Liu et al., 2020a). The overexpression of AtDA1 R358K increases seed size in Arabidopsis and B. napus. In addition, the expression of the Zmda1 or Zmdar1 mutant gene improves grain yield in maize. The overexpression of TaDA1 decreases the size and weight of wheat kernels, while RNA interference (RNAi) has the opposite effect. The above results may suggest that mutation is necessary to increase seed size.
In the present study, the DA1-like genes of four typical Gossypium cotton species were identified. Three, four, seven, and seven DA1 genes were identified in G. raimondii, G. arboreum, G. hirsutum, and G. barbadense. Previous studies have shown that DA1 proteins without the UIM and LIM domains exist in crop plants, including rice and maize . Our studies also identified some UIM domain-lacking DA1like genes, such as GrDA1-1, GhDA1-3D, GbDA1-3D, GaDA1-4, GhDA1-4A, and GbDA1-4A, as typical representatives of DA1like genes.
DA1-like genes play important roles in increasing seed yield and biomass during plant growth and development Wang et al., 2017;Liu et al., 2020a). Therefore, sequence alignment was used to identify protein sequence similarities to AtDA1. All GhDA1-like genes were shown to be conserved, and the GhDA1-1A homolog shared 63.79% identity to AtDA1. AtDA1 and GhDA1-1A contain the same functional domains, including two UIM domains, a LIM domain, and a DA1-like functional domain, suggesting their similar functions. Furthermore, GhDA1-1A and AtDA1 show similar expression patterns and are both widely expressed. Therefore, GhDA1-1A was cloned from G. hirsutum for further research. Previous studies have shown that OsDA1 is detectable around the plasma membrane of tobacco epidermal cells, while the fused TaDA1-A, -B, or -D proteins are distributed throughout the cytoplasm and nucleus (Liu et al., 2020a).
Because the overexpression of AtDA1 R358K was shown to increase seed size and seed weight Wang et al., 2017), the GhDA1-1A R301K mutant was transformed into Arabidopsis, which increased seed size and seed weight. Moreover, the transgenic plants produced larger seedlings and larger flowers than the wild-type plants, indicating that GhDA1-1A plays important roles in regulating seed and organ size. Researches had showed that the DA1 can interacts with DA2 in Arabidopsis and maize (Xia et al., 2013;Liu et al., 2020a). However, the relationship between DA1 and DA2 was unknow in cotton. Therefore, the BiFC assay and dualluciferase reporter assay were performed and verified that GhDA1-1A has an interaction with GhDA2, which showed that GhDA2 may function synergistically with GhDA1-1A to regulate seed size in cotton.
Previous studies showed that the seedling vigor and oil content are related to seed size. Seedlings from large seeds were more vigorous than the small ones in Brassica napus and cotton (Elliott et al., 2008;Pahlavani et al., 2009). In addition, there was a positive correlation between oil content and seed size in cotton (Pahlavani et al., 2009;Snider et al., 2014). Cotton ovules are composed of developing seeds and fibers. Developing seeds and fibers assimilate photosynthetic products through the same ovules. However, the relationship between seed size and fiber is complex and needs to be studied further. In our future work, we will generate transgenic cotton lines to explore the relationship between seed size and fiber.
In summary, we performed a genome-wide comprehensive analysis of DA1-like genes in the four sequenced Gossypium species and identified a total of 21 DA1-like genes. Larger seeds and organs present potential economic value in oil crop improvement and biomass energy production (Wang et al., 2016). The overexpression of GhDA1-1A R301K in Arabidopsis increased seed size and size weight, which may be useful in the improvement of seed size in cotton. Our detailed analysis of DA1-like genes in cotton has far-reaching significance for breeding work.

DATA AVAILABILITY STATEMENT
The datasets presented in this study can be found in online repositories. The names of the repository/repositories and accession number(s) can be found in the article/Supplementary Material.

AUTHOR CONTRIBUTIONS
JY and XZ directed the experiments. LH, JS, LL, YB, BJ, LW, YX, MW, and JZ participated in the study. SY conceived the study, performed the experiments, and wrote the manuscript. JY, XZ, and JZ revised the manuscript. All authors read and approved the final manuscript.