A Modified Actin (Gly65Val Substitution) Expressed in Cotton Disrupts Polymerization of Actin Filaments Leading to the Phenotype of Ligon Lintless-1 (Li1) Mutant

Dynamic remodeling of the actin cytoskeleton plays a central role in the elongation of cotton fibers, which are the most important natural fibers in the global textile industry. Here, a high-resolution mapping approach combined with comparative sequencing and a transgenic method revealed that a G65V substitution in the cotton actin Gh_D04G0865 (GhACT17D in the wild-type) is responsible for the Gossypium hirsutum Ligon lintless-1 (Li1) mutant (GhACT17DM). In the mutant GhACT17DM from Li1 plant, Gly65 is substituted with valine on the lip of the nucleotide-binding domain of GhACT17D, which probably affects the polymerization of F-actin. Over-expression of GhACT17DM, but not GhACT17D, driven by either a CaMV35 promoter or a fiber-specific promoter in cotton produced a Li1-like phenotype. Compared with the wild-type control, actin filaments in Li1 fibers showed higher growth and shrinkage rates, decreased filament skewness and parallelness, and increased filament density. Taken together, our results indicate that the incorporation of GhACT17DM during actin polymerization disrupts the establishment and dynamics of the actin cytoskeleton, resulting in defective fiber elongation and the overall dwarf and twisted phenotype of the Li1 mutant.


Introduction
Cotton is an important economic crop, providing most of the natural fibers used in the textile industry. High fiber yield and good quality are always the ultimate goals in the conventional breeding of upland cotton, which is a challenge for scientific researchers, because of the complex genetic basis of these plants. Gossypium hirsutum Ligon lintless-1 (Li 1 ) is a fiber mutant [1], which has been used frequently to study the genetic mechanisms associated with fiber development in cotton. Li 1 not only has seeds with very short lint fibers, but also displays other abnormal phenotypes such as twisted stems, curled leaves, and short plants, which have been confirmed to be caused by a dominant gene and result in a low level of survival in the homozygote [2,3]. The Li 1 locus has been mapped to Chr. 22 (D) [4][5][6], and the candidates for the Li 1 gene vary among different studies [4,[7][8][9] or lack supportive evidence of their roles in causing the mutation [10][11][12].
Actin proteins are required to construct the cell skeleton, maintain normal development, and undertake almost all activities in eukaryotic cells [13]. In higher plants, actin is encoded by a multigene family, whose members are involved in many aspects of plant growth and development. In poplars, the actin family consists of eight members that are differentially expressed in various tissues, but preferentially expressed in the stem phloem and xylem, suggesting that poplar actins are involved in wood formation [14]. In Arabidopsis, the actin gene family contains eight functional members. They are grouped into two ancient classes, a vegetative class that is expressed in leaves, stems, roots, petals, and sepals and a reproductive class that is expressed in pollen, ovules, and embryonic tissues [15][16][17]. The mis-expression of the Arabidopsis ACT1 (AtACT1) iso-variant in vegetative tissues affects the dynamics of actin and actin-associated proteins, in turn, disrupting the organization of the actin cytoskeleton and normal development in Arabidopsis [18]. The actin cytoskeleton, through interaction with the ARP2/ARP3 complex, plays a crucial role in controlling the shape of several cell types, especially trichome cells in Arabidopsis [19,20]. Arabidopsis ACT2 (AtACT2) functions in cell elongation and root formation, and an AtACT2 dominant-negative mutation disturbs F-actin polymerization, causing aberrant cell morphology [21]. Another actin mutant (AtACT8) with a single nucleotide substitution also has dominant-negative effects on actin polymerization, resulting in functional defects of actin filaments, including frizzy filaments containing many kinks and severe curling and coiling-up of lateral shoots [22]. Taken together, the actin cytoskeleton is essential for cell elongation and tip growth and may be involved in the transportation of organelles and vesicles carrying membrane and cell wall components to the site of cell growth, as is the case in root hairs, trichomes, and fibers [23].
In cotton, 15 actin genes (GhACT1-15) have been cloned and characterized. Some evidence indicates that actin is important in maintaining the normal morphology of cotton fibers [23]. Upland cotton actin genes are differentially expressed in different tissues, and GhACT1 is known to play a critical role in fiber elongation [23]. In other studies, a number of cytoskeletal genes have been found to be expressed specifically or preferentially in developing fiber cells. A comparative proteomic analysis between Li 1 and its wild-type revealed that numerous cytoskeleton-related proteins are decreased remarkably in the fibers of Li 1 . Accordingly, the architecture of the actin cytoskeleton is severely deformed and the microtubule organization is moderately altered, while vesicle trafficking is dramatically disrupted [24]. Using sequencing-mapping and virus-induced gene silencing (VIGS), Thyssen et al. [11] and Sun et al. [12] recently indicated that the actin molecule encoded by Gh_D04G0865 could be the best candidate for producing the Li 1 mutant. Thyssen et al. [11] proposed "Poison subunit" model based on structural modeling of actin monomer and polymer, whereas Sun et al. [12] experimentally proved that model by in vitro polymerization of defective actin unit. However, there is no direct transgenic evidence available. More genetic and cellular evidence still needed to revealing the mechanism for F-actin polymerization in vivo.
We previously reported fine mapping of the Li 1 locus in a gene-sparse region of about 0.3 cM and 1.2 Mb in length, where 36 genes have been annotated [10]. As candidate gene expression studies indicated that Gh_D04G0865 is a promising candidate, we immediately constructed a larger F 2 population to further fine map this gene and carried out transgenic verification of Gh_D04G0865 in Arabidopsis and cotton. In the present work, the Li 1 gene was further mapped to a narrower region, which co-segregated with Gh_D04G0865. Based on comparative sequencing, analysis of expression patterns, transgenic evidence, and filament array analysis, Gh_D04G0865, named GhACT17D in our experiments, mutated in Li 1 plant. We named this mutant as GhACT17DM, the G65V substitution of which was found to affect F-actin elongation, causing the abnormal development of the Li 1 mutant.

Fine Mapping and Sequence Analysis Indicated That GhACT17D Was a Promising Candidate Gene for the Li 1 Mutation
Compared with the wild-type, the Li 1 mutant not only exhibited short-lint on the seeds but also presented obvious differences in all morphological characteristics throughout its development ( Figure 1D). The mutant demonstrated delayed growth with smaller and  . (C), The genetic map of Li1 and the candidate genes constructed using all segregating data from current research. The lines connect the same DNA sequence in different maps. (D), The varied phenotypes from different genotypes of Li1. Scale Bars, 1 cm. (E), Gene structure of GhACT17D. Boxes represent exons; filled parts of boxes represent the coding region. The position of the point mutation in GhACT17DM is indicated by a black arrow.
The GhACT17D ORF was 1134 bp in length and encoded a peptide containing 377 amino acid residues, comprising 38 strongly basic (+), 50 strongly acidic (−), 129 hydrophobic, and 86 polar amino acids. In a search against the NCBI protein sequence databases, the putative protein sequence of GhACT17D showed homology with actin proteins from other plants and contained the conserved nucleotide-binding domain (NBD) of the sugar kinase/HSP70/actin superfamily, which forms a deep cleft in which the nucleotide sits [26]. Eleven residues that compose this conserved feature were mapped to GhACT17D (Figure 2A). The GhACT17DM protein from the Li1 mutant had a single residue substitution (Gly to Val) at the 65th amino acid in the NBD region, which was the result of a single nucleotide mutation (G to T) at the 194th base in the ORF (Figure 1).

Gly65 Is Highly Conserved and Crucial to GhACT17D
We investigated the sequence variation of GhACT17D in eight cultivars of five tetraploid cotton species. The coding regions of GhACT17D from these cultivars were sequenced, aligned, and used to construct a phylogenetic tree. All sequences were divided into two subgroups, A and D, and each group included eight cultivar sequences. GhACT17D sequences varied slightly either between different species or different cultivars, but no single nucleotide polymorphism (SNP) was found at the mutation site of GhACT17DM in the genome ( Figure S2). We also analyzed the G65 position in GhACT17D of 36 cotton actin families with complete NBDs in the TM-1 genome [25] and eight Ara- We had previously mapped the Li 1 locus to a 0.3 cM region using two populations consisting of 1166 F 2 plants, which was about 1.2 Mb in the published G. raimondii genome [10]. When the previously mapped markers around Li 1 were aligned in the TM-1 reference sequence [25] (Figure 1B), most of the homologous loci of the single-strand conformation polymorphism (SSCP) markers were found, except for a few including P095, P257, and P191. They also appeared in the same order as before, except for P251 ( Figure 1A,B). To further narrow the Li 1 region, a total of 80 microsatellite markers (Table S2) were selected from the TM1 genome sequence to screen for the polymorphic ones. After analyzing 2316 F 2 plants, including 1111 used in our previous paper, 603 new F 2 plants from the same cross combination and 602 F 2 plants from E24-3583 × Li 1 , a genetic map containing 19 markers was constructed. Following this effort, Li 1 was found to co-segregate with D04_1111, D04_1117, and D04_1118, which narrowed the search to a 0.04 cM region flanked by D04_1108 and D04_1124, corresponding to 392 kb of the TM-1 reference genome ( Figure 1C, Supplemental data 1). According to the gene annotation, this region contained eight genes, in which a beta-tubulin 7 (Gh_D04G0862) and an actin (Gh_D04G0865) gene ( Figure 1C) were predicted to be two of the most-promising candidates to be the Li 1 mutation . However, no expression of the beta-tubulin 7 gene was found in the stem and ovules of 0 DPA wild-type plants or Li 1 mutants [10]. Therefore, we isolated the Gh_D04G0865 sequences from the Li 1 mutant and wild-type plants, and a single nucleotide substitution of G by T at the 194th base of the ORF from Li 1 was detected ( Figure 1E). According to these results, we deduced that Gh_D04G0865 was the most-promising candidate gene causing the Li 1 mutation and named it GhACT17D. The mutant from Li 1 was named as GhACT17DM.
Using the SMART-RACE approach, the full-length cDNA of GhACT17D containing 1532 was cloned, and deposited in GenBank (accession numbers: MG132060). Sequence analysis revealed that the cDNA of GhACT17D contained a single ORF from nucleotides 63 to 1196, which was flanked by a 62-bp 5 -UTR and 336-bp 3 -UTR. The gene structure of GhACT17D showed that three introns split the coding region between the start and stop codons ( Figure 1E). Furthermore, a 397-bp intron interrupted the 5 -UTR at -10 nucleotides upstream of the ATG initiation codon ( Figure 1E). The splicing sites of all introns obeyed the GT-AG rule.
The GhACT17D ORF was 1134 bp in length and encoded a peptide containing 377 amino acid residues, comprising 38 strongly basic (+), 50 strongly acidic (−), 129 hydrophobic, and 86 polar amino acids. In a search against the NCBI protein sequence databases, the putative protein sequence of GhACT17D showed homology with actin proteins from other plants and contained the conserved nucleotide-binding domain (NBD) of the sugar kinase/HSP70/actin superfamily, which forms a deep cleft in which the nucleotide sits [26]. Eleven residues that compose this conserved feature were mapped to GhACT17D (Figure 2A). The GhACT17DM protein from the Li 1 mutant had a single residue substitution (Gly to Val) at the 65th amino acid in the NBD region, which was the result of a single nucleotide mutation (G to T) at the 194th base in the ORF ( Figure 1).

Gly65 Is Highly Conserved and Crucial to GhACT17D
We investigated the sequence variation of GhACT17D in eight cultivars of five tetraploid cotton species. The coding regions of GhACT17D from these cultivars were sequenced, aligned, and used to construct a phylogenetic tree. All sequences were divided into two subgroups, A and D, and each group included eight cultivar sequences. GhACT17D sequences varied slightly either between different species or different cultivars, but no single nucleotide polymorphism (SNP) was found at the mutation site of GhACT17DM in the genome ( Figure S2). We also analyzed the G65 position in GhACT17D of 36 cotton actin families with complete NBDs in the TM-1 genome [25] and eight Arabidopsis actin sequences, which showed that none of them varied in the position of G65 ( Figure S3, Table S5). These results indicated that the G65 in GhACT17D was highly conserved in paralogs and orthologs.
The physicochemical properties of the putative proteins showed that GhACT17D and GhACT17DM were stable, but the instability index and molecular weight of GhACT17DM were slightly higher than those of GhACT17D (Table S6). The analysis of intolerant amino acid substitutions sorted by SIFT revealed that the change of G65 to V65 was deleterious (p = 0.00), which indicated that the GhACT17DM protein was altered phenotypically and functionally. The three-dimensional structure of the GhACT17D/GhACT17DM monomer was modeled with Swiss-Model Workspace using a major actin (SMTL id: 3ci5.1.A) with a Global Model Quality Estimation (GMQE) of 0.99 as the template, and G65 was found to be at the entrance of the NBD cleft ( Figure 2B, labeled by box 1). When using another actin (SMTL id: 1nmd.1.A) with a GMQE of 0.98 as the template, a site for ligand SO2, in contact with R64, T204, T205, and A206, was found to be adjacent to the 65th amino acid ( Figure 2C). Furthermore, the GhACT17D/DM homo-dimer was modeled with a cytoplasmic actin (SMTL id: 2oan.1.A) as the template ( Figure 2D). These simulated protein structures indicated the importance of the position of G65 in GhACT17D. Compared to G65 in GhACT17D, the V65 in GhACT17DM emerged as a derivative of three branches ( Figure 2C,D, labeled by box 1), which was attributed to the fact that valine is a non-polar branched-chain amino acid, while glycine is a small amino acid. The substitution of G65V would result in spatial change in the structure of the GhACT17DM molecule ( Figure 2D), most likely affecting the polymerization of actin filaments. 2D). These simulated protein structures indicated the importance of the position of in GhACT17D. Compared to G65 in GhACT17D, the V65 in GhACT17DM emerged derivative of three branches ( Figure 2C,D, labeled by box 1), which was attributed to fact that valine is a non-polar branched-chain amino acid, while glycine is a small am acid. The substitution of G65V would result in spatial change in the structure of GhACT17DM molecule ( Figure 2D), most likely affecting the polymerization of actin ments.  Table S5. TcACT11 and AtACT1 are actin genes from other plant (Table  S5) Table S5. TcACT11 and AtACT1 are actin genes from other plant (Table S5)

Over-Expression of GhACT17DM in Transgenic Upland Cotton Results in the Li 1 Mutant Phenotype
Considering that actin is a constitutive cytoskeletal protein, the heterologous expression of GhACT17DM in Arabidopsis seemed insufficient to prove its function, although the over-expression of GhACT17DM resulted in more severely deformed morphology than the over-expression of GhACT17D (data not shown). Therefore, over-expression transgenic cotton lines were produced (Figure 3 and Table S7). GhACT17DM sequences, containing an 1134-bp ORF along with a 48-bp 3 -UTR, were inserted into over-expression vectors driven by CaMV35s or GbEXPA2 promoters (pGbEXPA2), respectively. CaMV35s is a constitutive promoter and pGbEXPA2 is the preferential promoter for quickly elongating cotton fibers [27]. Twenty-five independent transgenic over-expression lines of pGbEXPA2::GhACT17DM were obtained and all the lines generated short fibers, without obvious changes in plant morphology. After screening by Southern blotting, the transgenic lines with low copy numbers (EM4, EM10, and EM24) were selected for further analysis ( Figure S4). With GhACT17D-specific primers, the GhACT17DM expression levels were found to be significantly increased in fibers (10 DPA) of EM4, EM10, and EM24, compared to the wild-type and nulls ( Figure S6). The expression level was closely related to the mature fiber lengths of the transgenic lines and the fibers were shorter when the expression level was higher ( Figure 3A, Figures S5 and S6A). Only two CaMV35s::GhACT17DM transgenic lines (3M3-1 and 3M3-6) were obtained, because the constitutive strong expression of GhACT17DM affected somatic embryogenesis, which resulted in the same phenotype as the Li mutant, with dwarf plants, curly leaves, and much shorter fibers ( Figure 3A,C,D). Similar to the result obtained from the pGbEXPA2::GhACT17DM transgenic lines, the GhACT17DM transcript levels of 3M3-6 were significantly higher than those of 3M3-1, and the leaves of 3M3-6 were more heavily crinkled than those of 3M3-1.
genic lines with low copy numbers (EM4, EM10, and EM24) were selected for further analysis ( Figure S4). With GhACT17D-specific primers, the GhACT17DM expression levels were found to be significantly increased in fibers (10 DPA) of EM4, EM10, and EM24, compared to the wild-type and nulls ( Figure S6). The expression level was closely related to the mature fiber lengths of the transgenic lines and the fibers were shorter when the expression level was higher ( Figure 3A, Figures S5 and S6A). Only two CaMV35s::GhACT17DM transgenic lines (3M3-1 and 3M3-6) were obtained, because the constitutive strong expression of GhACT17DM affected somatic embryogenesis, which resulted in the same phenotype as the Li mutant, with dwarf plants, curly leaves, and much shorter fibers ( Figure 3A,C,D). Similar to the result obtained from the pGbEXPA2::GhACT17DM transgenic lines, the GhACT17DM transcript levels of 3M3-6 were significantly higher than those of 3M3-1, and the leaves of 3M3-6 were more heavily crinkled than those of 3M3-1.
In order to further confirm that GhACT17DM over-expression, and not that of GhACT17D, resulted in the mutant phenotype, transgenic lines (3W7 and 3W46) over-expressing GhACT17D driven by CaMV35 were also obtained. There was no obvious phenotype difference, although the expression levels were much higher than those in the wild-type ( Figure 3C, Figure S6F).  In order to further confirm that GhACT17DM over-expression, and not that of GhACT17D, resulted in the mutant phenotype, transgenic lines (3W7 and 3W46) over-expressing GhACT17D driven by CaMV35 were also obtained. There was no obvious phenotype difference, although the expression levels were much higher than those in the wild-type ( Figure 3C, Figure S6F).

An Abnormal F-actin Structure Results in Defective Growth and Development of the Li 1 Mutant
To gain further insight into the cellular mechanisms underlying the severely dysfunctional growth phenotype of the Li 1 mutant, we performed live cell imaging of the actin cytoskeleton in both elongating cotton fibers at 2 DPA and expanding leaf epidermal cells. Remarkably, F-actin architectures in both the Li 1 fiber cells and leaf epidermal cells had shorter F-actin fragments, along with long F-actin cables, compared with the wild-type control. The defective F-actin organization in the Li 1 mutant resembled that of the dominant-negative Arabidopsis act2-2D mutant, which displays actin polymerization defects [21]. Moreover, we monitored and measured the parameters of the F-actin dynamics in rapidly elongating fiber cells at 2 DPA. Both the growth and shrinkage rates of F-actin in Li 1 fibers were about twice those of the wild-type control (Figure 4 and Supplemental movies 1 and 2). The F-actin density was significantly higher in the Li 1 fibers than the wild-type control. On the contrary, the skewness and parallelness of F-actin decreased in the Li 1 fibers. Taken together, these results indicated that actin polymerization may be disturbed in the Li 1 mutant, resulting in defective F-actin assembly with more short fragments offering less stability.
Remarkably, F-actin architectures in both the Li1 fiber cells and leaf epidermal cells had shorter F-actin fragments, along with long F-actin cables, compared with the wild-type control. The defective F-actin organization in the Li1 mutant resembled that of the dominant-negative Arabidopsis act2-2D mutant, which displays actin polymerization defects [21]. Moreover, we monitored and measured the parameters of the F-actin dynamics in rapidly elongating fiber cells at 2 DPA. Both the growth and shrinkage rates of F-actin in Li1 fibers were about twice those of the wild-type control (Figure 4 and Supplemental movies 1 and 2). The F-actin density was significantly higher in the Li1 fibers than the wildtype control. On the contrary, the skewness and parallelness of F-actin decreased in the Li1 fibers. Taken together, these results indicated that actin polymerization may be disturbed in the Li1 mutant, resulting in defective F-actin assembly with more short fragments offering less stability.

Dominant-Negative Mutation of the Gh_D04G0865 Gene Causes the Li1 Phenotype
The Li1 mutant has attracted extensive attention since its discovery in 1929 [1] and much effort has been put into cloning the gene responsible [2,4]. Many possible candidate genes and related pathways have been suggested, but most of the results, especially those from gene expression analyses, vary considerably [28][29][30]. In 2015, we proposed that Black arrows represent the elongation direction. Scale bars, 10 µm. The two asterisks represent highly significant difference.

Dominant-Negative Mutation of the Gh_D04G0865 Gene Causes the Li 1 Phenotype
The Li 1 mutant has attracted extensive attention since its discovery in 1929 [1] and much effort has been put into cloning the gene responsible [2,4]. Many possible candidate genes and related pathways have been suggested, but most of the results, especially those from gene expression analyses, vary considerably [28][29][30]. In 2015, we proposed that Gh_D04G0865 (numbered as P258) was one of the most promising candidate Li 1 mutant genes, based on fine genetic mapping and gene expression analysis [10]. Thyssen et al. also identified Gh_D04G0865 as the Li 1 candidate, labeled by a SNP marker, CFB11927 (23,783,515 bp in G. hirsutum Chr D04), using sequencing-mapping and VIGS [11]. The experiment carried out by Thyssen et al. used a 354-bp fragment from the conservative domain of actin C-terminal sharing high similarity among actin members that would not discriminate the actin families in VIGS, which resulted in a Li 1 -like phenotype but in a different mechanism [12]. Benefiting from the whole genome sequencing of upland cotton [25], we now know that 37 actin genes with complete NBDs sharing high identity occur in the TM-1 genome, including Gh_D04G0865 ( Figure S3, labeled by black square). Therefore, Sun et al. used a 3 UTR of Gh_D04G0865 for VIGS to show that it caused little change in wild-type seedlings and partially recovered the curled leaves to be normal in Li 1 seedlings after specifically inhibiting Gh_D04G0865 expression [12]. Sun et al. also discovered that the G65V substitution in Gh_D04G0865 disturbed filament polymerization in vitro but failed to generate gene-overexpression cotton.
In this study, the Li 1 locus was further narrowed to a region of 392 kb, in which only eight genes exist. According to nucleotide variation and gene expression studies, Gh_D04G0865 was believed to be the most promising candidate for the Li 1 mutation. Further stable transgenic validation confirmed that the overexpression of GhACT17DM, driven by either the CaMV35 or the pGbEXPA2 promoter in cotton caused the Li 1 -like phenotype. Therefore, our results directly support the conclusion that Gh_D04G0865 is the mutant gene associated with Li 1 and its G65V substitution results in the abnormal phenotype of the Li 1 mutant.

Incorporation of the Mutated Version of GhACT17DM Impaired Actin Polymerization and Cotton Growth with a Dosage-Related Effect
Actin is one of the most conserved proteins in eukaryotes. Structural predictions suggested that the G65V substitution would affect the polymerization capability of GhACT17DM, as confirmed by previous in vitro assays [12]. Importantly, our live-cell observations revealed that Li 1 cells showed defective dynamic behavior of the actin cytoskeleton, compared with that of the wild-type control (Figure 4). Actin is one of the most abundant proteins in cells and numerous actin encoding genes show redundant roles. Thus, knocking out the function of a single actin gene usually gives rise to no obvious change in phenotype. However, a point mutant in an important domain of the actin molecule that often produces a dominant-negative effect. It should be noted that F-actin networks could still form in Li 1 cells, because GhACT17DM only accounts for a small portion of actin proteins in the abundant actin pool. The incorporation of the mutated version of GhACT17DM impaired actin polymerization, thus generating a defective actin cytoskeleton with short F-actin fragments in Li 1 cells (Figure 4 and Supplemental movies 1 and 2). Integrating more defective actin variants resulted in more severe defects of the actin cytoskeleton, showing a dosage effect. Accordingly, the homozygous Li 1 presents a more twisted and dwarf phenotype than the heterozygote and does not often survive, which causes a lower number of mutant plants than expected in the F 2 offspring [3,6,10]. In our transgenic cotton lines, the higher the GhACT17DM expression, the shorter the mature fiber length or more crinkled the plant. These results indicated that more copies of GhACT17DM, or more transcripts, made plants more severely abnormal, which revealed the positive dosage effect of GhACT17DM.

Plant Materials and Fine Genetic Mapping
A total of 12 varieties of five tetraploid cotton species, G. barbadense, G. darwinii, G. hirsutum, G. mastelinum, and G. tomentosum, were used in this study (Table S1)  . G. hirsutum acc. YZ1 was used to produce transgenic cotton. The plants were grown in a greenhouse at Huazhong Agricultural University in Wuhan using standard farm management practices, according to relevant approvals for biotechnology research. The other four tetraploid cotton species were used to study gene expression, sequence similarity, and copy number.
The generation of ABD2-GFP marker lines was produced using G. hirsutum (accession R15) as the recipient of the 35S::ABD2-GFP construct [31]. Crosses were made between Li 1 mutants and ABD2-GFP marker lines. In the segregated F 2 progeny, Li 1 homozygous lines and wild-type lines expressing the ABD2-GFP marker were used for live cell imaging, and subsequent comparative analysis on F-actin dynamics was performed between Li 1 fibers and the wild-type control. Images of elongating fibers were obtained from cotton fibers at 2 days-post-anthesis (DPA), and images of leaf epidermal cells were obtained from the first true leaf of 2-week-old cotton seedlings.
Genomic DNA was extracted from fresh leaves using the CTAB method [32], which was slightly modified, as described by Ding [7]. DNA from the F 2 populations was used to map Li 1 . In order to narrow the map of Li 1 , 19 simple sequence repeat (SSR) markers were selected from those reported in the targeted region [25], based on a previously published map [10] (Table S2). The bordering SSRs were used to screen the recombinant F 2 plants. Further, SSR markers were applied to determine the location of the recombinant events and finally to narrow the location of the Li 1 mutation.

Gene Cloning and Sequence Analysis
The DNA extracted from different genotypes and species was used to amplify the candidate gene for sequence analysis. Total RNA was extracted from fresh young leaves with a TianGen RNAprep pure kit (Tiangen, China). The full-length cDNA was obtained using SMART-Rapid Amplification of cDNA Ends (RACE) approaches (SMARTer ® RACE 5 /3 Kit, Clontech, USA). The gene-specific primers (Table S3) for 5 and 3 RACE were designed from the coding regions. Approximately 2 µg of high-quality total RNA was used for the 5 and 3 SMART-RACE protocols, strictly following the manufacturer's instructions. The PCR products obtained from the cDNA and genomic DNA amplification were ligated into the pMD 18-T vector (Takara, China) and transformed into E. coli cells (DH5α), before sequencing at BGI (Shenzheng, China).

cDNA Preparation and RT-PCR
To conduct reverse transcription (RT)-PCR analysis, approximately 1 µg of total RNA from each sample was used to synthesize the first-strand cDNAs in a 20-µL reaction volume using Moloney-Murine Leukemia Virus reverse transcriptase (Takara) with Oligo d(T)18 primer. The synthesized cDNAs were used as templates in the following PCR reactions.
For each target gene, PCR amplification was performed using Taq polymerase (Takara) with gene-specific primer pairs (Table S3). The PCR was conducted in a thermal cycler (Bio-RAD, Singapore), using the following procedure: pre-denaturation at 94 • C for 3 min, followed by 30 cycles of 30 s at 94 • C, and 60 s at 68 • C, for each gene. The cotton gene ChUB7 (DQ116411) was used as an internal control [33].

Quantitative Real-Time PCR (qRT-PCR)
Quantitative PCR was carried out to analyze GhACT17D/GhACT17DM expression in the Li 1 mutant and wild-type, the primers were listed in Table S3. The cDNAs used to detect gene expression were the same as those used in the RT-PCR analysis. The reaction was conducted on a BioRad CFX connect™ real-time PCR System (BioRad, USA) using Takara Taq TM (Takara) with EvaGreen (Biotium), according to the manufacturer's instructions. The amplification of the target gene was monitored with an EvaGreen fluorescence signal on every cycle. The relative expression level was standardized to the expression level of ChUB7 (DQ116411) ( Table S3). The cycle threshold value (Cq) was used as a measure of the starting copy number of the target gene, and gene relative expression level =1/2ˆ(CqGene-CqUbq7), error value = S/sqrt(3), S=standard deviation [33].
Quantitative PCR was carried out to analyze gene expression in the transgenic cotton. Flower buds were marked on the day of anthesis (0 DPA), 10 DPA bolls were harvested, hulls were stripped, and fibers and ovules were frozen immediately in liquid nitrogen and stored at −70 • C until use. Total RNA was extracted using the thiocyanate method and 3 µg of total RNA were reverse-transcribed to cDNA using SuperScript II reverse transcriptase (Invitrogen, USA). Quantitative RT-PCR was performed as previously described [34] on a 7500 Real-Time PCR system (Applied Biosystems).

Phylogenetic Analysis
The sequence data used here were retrieved from GenBank databases using BLASTN/P programs (http://blast.ncbi.nlm.nih.gov (accessed on 1 February 2021)) on the National Center for Biotechnology Information (NCBI) database. Multiple alignments of the nucleotide and deduced amino acid sequences were performed using ClustalW [35], and the results were manually adjusted with GeneDoc 2.7 [36]. A phylogenetic tree was constructed by the neighbor-joining method with 1000 bootstrap replications, using MEGA 5 [37]. The ExPASy ProtPara program (http://web.expasy.org/protparam/ (accessed on 1 February 2021)) was used to analyze the physicochemical properties. The SIFT tool was used to predict the effects of amino acid substitutions on proteins, with probabilities of <0.05 considered deleterious [38]. The Swiss-Model Workplace was used to carry out protein structure homology modeling [39].

Vector Construction and Genetic Transformation in Cotton
To obtain an over-expression construct, an 1134-bp ORF fragment, along with the 48-bp 3 -untranslated region (UTR) of GhACT17DM and GhACT17D was amplified by the primers with attB1 and attB2 adaptors and ligated into the over-expression vectors pGWB408 (driven by the CaMV35s promoter) and pGWB407-GbEXPA2 (driven by the PGbEXPA2 promoter) using Gateway technology [40].

Southern Blotting in Transgenic Cotton
The genomic DNA was extracted using a Plant Genome Extracting kit (TIANGEN, China) and 20 µg DNA were digested completely with HindIII, separated by 0.8% gel electrophoresis, transferred into a nylon membrane (Amersham, UK), and hybridized with digoxin-labeled DNA fragments of NPII at 42 • C overnight. The membrane was incubated with anti-AP in 20 mL blocking solution at 37 • C for 40 min and washed three times. The CSPD was added and signaling was detected in a darkroom. Detail procedures were as described in the DIG High Prime DNA Labelling and Detection Starter Kit II (Roche, Switzerland).

Analyzing the Dynamic Behavior of F-actin
Live-cell imaging and video were carried out under a spinning disk confocal microscope (UltraView VoX, Perkin Elmer) equipped with the Yokogawa Nipkow CSU-X1 spinning disk scanner, the Hamamatsu EMCCD 9100-13 and the Nikon TiE inverted microscope containing the Perfect Focus System. Acquired images were analyzed using Volocity (Perkin Elmer), ImageJ (http://rsbweb.nih.gov/ij (accessed on 1 February 2021)), as described previously [31]. The F-actin severing frequency was defined as the number of severing events per 100 µm 2 per second. The elongation and shrinkage rates (µm/s) were measured as previously described [43,44]. F-actin skewness and density (%) were measured according to an established approach [45]. Parallelness, which was defined as the mean angular difference, was applied to evaluate the F-actin orientations as described previously [46].

Conclusions
In conclusion, our findings revealed that the G65V substitution in GhACT17DM is responsible for the Li 1 mutation, and incorporation of GhACT17DM during actin polymerization disturbs the establishment and dynamics of the actin cytoskeleton, resulting in defective fiber elongation and the overall dwarf and twisted phenotype of the Li 1 mutant.