A deletion/duplication in the Ligon lintless-2 locus induces siRNAs that inhibit cotton fiber cell elongation

Abstract

Most cultivated cotton (Gossypium hirsutum L.) varieties have two types of seed fibers: short fuzz fiber strongly adhered to the seed coat, and long lint fiber used in the textile industry. The Ligon lintless-2 (Li₂) cotton mutant has a normal vegetative phenotype but produces very short lint fiber on the seeds. The Li₂ mutation is controlled by a single dominant gene. We discovered a large structural rearrangement at the end of chromosome D13 in the Li₂ mutant based on whole-genome sequencing and genetic mapping of segregating populations. The rearrangement contains a 177-kb deletion and a 221-kb duplication positioned as a tandem inverted repeat. The gene Gh_D13G2437 is located at the junction of the inverted repeat in the duplicated region. During transcription such structure spontaneously forms self-complementary hairpin RNA of Gh_D13G2437 followed by production of small interfering RNA (siRNA). Gh_D13G2437 encodes a Ran-Binding Protein 1 (RanBP1) that preferentially expresses during cotton fiber elongation. The abundance of siRNA produced from Gh_D13G2437 reciprocally corresponds with the abundance of highly homologous (68%–98% amino acid sequence identity) RanBP1 family transcripts during fiber elongation, resulting in a shorter fiber phenotype in the Li₂. Overexpression of Gh_D13G2437 in the Li₂ mutant recovered the long lint fiber phenotype. Taken together, our findings revealed that siRNA-induced silencing of a family of RanBP1s inhibit elongation of cotton fiber cells in the Li₂ mutant.

A deletion/duplication in the Ligon lintless-2 locus induces siRNAs from self-complementary transcripts of Ran Binding Protein 1 that inhibit cotton fiber cell elongation.

Introduction

Cotton (Gossypium hirsutum L.) fibers are seed trichomes that emerge from ovule epidermal cells before or on the day of anthesis (DOA). These trichomes are single-celled projections that elongate up to 3,000 times their initial length at maturation and are considered as a model system for studying cell elongation and cell wall biogenesis (Kim and Triplett, 2001). Development of cotton fiber consists of four distinct yet overlapping stages, including initiation, elongation, secondary cell wall biosynthesis, and maturation (Kim and Triplett, 2001). Long lint and short fuzz fiber cells emerge at different times from the epidermal layer: lint fibers initiate around 3 days before or on the DOA, whereas fuzz fibers initiate around 3–5 days postanthesis (DPA) (Steward, 1975; Zhang et al., 2007).

A naturally occurring cotton (G. hirsutum L.) fiber mutant was found in a cotton-breeding nursery of the Texas Agricultural Experiment Station in 1984 (Narbuth and Kohel, 1990). This mutant had normal vegetative morphology; however, its seeds had very short fibers similar to the Ligon lintless-1 (Li₁), therefore, this mutant was named as Ligon lintless-2 [Li₂] (Narbuth and Kohel, 1990). The Li₂ mutation is controlled by a single dominant gene, which is not allelic with the Li₁ (Narbuth and Kohel, 1990). Since its discovery, the Li₂ mutant has been used as a model system to study cotton fiber development (Kohel et al., 1992; An et al., 2010; Hinchliffe et al., 2011; Naoumkina et al., 2013, 2015; Gilbert et al., 2014; Patel et al., 2020).

Eventually, two near-isogenic lines (NILs) of the Li₂ with the Upland cotton variety DP5690 as recurrent parent were developed (Hinchliffe et al., 2011). These NILs were used for morphological evaluation, gene expression, and metabolite analyses of developing ovules and fibers cells from mutant and wild-type plants (Hinchliffe et al., 2011; Naoumkina et al., 2013, 2015; Gilbert et al., 2014). Evaluation of developing ovules of the Li₂ by scanning electron microscopy determined that plants produce seeds with fuzz and short lint fibers since fiber initials observed on the DOA (Hinchliffe et al., 2011). Earlier mapping studies linked the Li₂ locus to chromosome (Chr.) 18 [also called Chr. D13] by phenotype association with cotton aneuploid stocks (Kohel et al., 2002), and linkage analysis by restriction fragment length polymorphism (RFLP) markers (Rong et al., 2005). More recent mapping studies placed the Li₂ locus toward the end of the Chr. D13 (Liu et al., 2010; Hinchliffe et al., 2011; Thyssen et al., 2014). A genetic study of the Li₂ mutation indicated that it may have incomplete penetrance since both long and short types of fibers (mixed phenotype) were observed on the same plant (An et al., 2010). The most recent mapping study detected the Li₂ mutation locus within a terminal deletion of Chr. D13, which was suggested to be responsible for the Li₂ short fiber phenotype (Patel et al., 2020).

Small noncoding RNAs have emerged as key regulators of gene expression, genome stability, and defense against foreign genetic elements (Moazed, 2009). They play important roles in plant development and stress responses. For example, microRNA (miR)828 and miR858 regulate homoeologous MYB2 gene functions in Arabidopsis (Arabidopsis thaliana) trichome and cotton fiber development (Guan et al., 2014). The (miR164–G. hirsutum Cup-shaped Cotyledon 2–GhBRANCHED1 (miR164–GhCUC2–GhBRC1) module regulates cotton plant architecture through abscisic acid (Zhan et al., 2021). A recently discovered miRNA gar-miRN44, identified in Gossypium arboreum, regulates cotton ovule growth by targeting a Zn²⁺ ion transporter gene, GaZIP1L (Wen et al., 2021). An evolutionary study of the miR482/2118 superfamily, which modulates the expression of nucleotide-binding site leucine-rich repeat disease resistance genes, reported that this superfamily expanded in cotton by a class-II transposable element (Shen et al., 2020). A recent study revealed that small interfering RNAs (siRNAs) from bidirectional transcripts of GhMML3_A12 (MYB-MIXTA-like [(MML] transcription factor 3 on chromosome A12) regulate cotton fiber development (Wan et al., 2016). Numerous studies have demonstrated that small RNAs play an essential role in cotton ovules development and fiber elongation (Abdurakhmonov et al., 2008; Liu et al., 2014; Naoumkina et al., 2016; Sun et al., 2017).

Transport of macromolecules between the cell nucleus and the cytoplasm occurs through the nuclear pore complexes and is mediated by RanGTPase (Bischoff and Görlich, 1997). Ras-related nuclear protein (Ran) is a member of the Ras superfamily of small guanosine triphosphatases (GTPases), which bind and hydrolyze GTP (Moore, 2013). Ran cycles between GTP- and GDP-bound states, which is linked to transport into or out of the nucleus (Moore, 1998; Nielsen, 2020). By the actions of the guanine nucleotide exchange factor (RanGEF) found in the nucleus (Ohtsubo et al., 1989) and the GTPase-activating protein (RanGAP1) found in the cytoplasm (Hopper et al., 1990), a steep gradient of RanGTP and RanGDP is maintained and directs transport across the nuclear membrane, with RanGTP inside the nucleus and RanGDP inside the cytoplasm (Izaurralde et al., 1997). Ran-binding protein 1 (RanBP1) is an abundant cytosolic protein that binds RanGTP and increases its affinity for RanGAP1 to catalyze GTP hydrolysis on Ran in cytoplasm (Bischoff et al., 1995; Moore, 1998; Merkle, 2011; Nielsen, 2020).

In this study, we report our discovery of a structural rearrangement at the end of Chr. D13 in the Li₂ mutant that includes a 177-kb deletion and a 221-kb duplication positioned as a tandem inverted repeat. We found that siRNAs were likely produced from the self-complementary transcripts of Gh_D13G2437 (located at the junction of the tandem inverted repeat) in the Li₂ mutant but not in the wild-type. Gh_D13G2437 gene encodes an ortholog of an Arabidopsis RanBP1. We examined expression of the genes from the Chr. D13 deleted and duplicated regions in developing fibers of a mixed phenotype plant. Only Gh_D13G2437 showed substantially lower expression in the short fiber cells relative to the long fibers from the same plant, which suggested that this gene could be responsible for the short fiber phenotype in the Li₂ mutant. Introduction of the highly expressed transgenic copy of Gh_D13G2437 into the Li₂ mutant complemented and partially recovered the longer lint fiber cells phenotype.

Results

Li₂ phenotype

The vegetative tissues of the Li₂ mutant plant do not have visible differences compared to its wild-type NIL DP5690. The lint fiber on the Li₂ seeds is very short, <6-mm length (Figure 1). Occasionally, the Li₂ plants produce two types of fibers (mixed fiber phenotype) on the same plant or in the same boll (Figure 1, C and D). We observed different levels of phenotypic penetrance, such as most of seeds on plant have short fiber and a few with longer fiber; or seeds have shorter than wild-type fiber but not uniformly short as on the Li₂. The cases where the plant produces clearly wild-type and uniformly short Li₂ type cotton bolls are rare.

Figure 1.

Seed phenotype of wild-type and Li₂ plants. A, Open boll of wild-type DP5690; (B) open boll of Li; (C) two types of long and short fiber on the same branch of Li₂ plant; and (D) two types of fiber on the same developing boll from an Li₂ plant.

Mapping studies suggested that the Li₂ mutation is monogenic and dominant (Narbuth and Kohel, 1990; Rong et al., 2005). Our evaluation of phenotype approximately matched 3–1 wild-type to short fiber segregation ratio from two populations grown in two different years (Table 1). However, progeny plants with wild-type fiber phenotype were enriched in both populations producing a 2.6–1.4 ratio in 2012 and a 2.8–1.2 ratio in 2013.

Table 1.

Evaluation of phenotypic ratio by goodness of fit using chi-square test of short fiber seed segregation of F₂ populations

F2 population cross	Year	F2 plants	Seed phenotype observed		Expected for 3 : 1 ratio		Chi-square (probability)
			Short fiber no.	Wild-type no.	Short fiber no.	Wild-type no.
DP5690 × Li2	2012	515	336	179	386	129	11.6 (P = 6.7e-4)
DP5690 × Li2	2013	1185	837	348	889	296	5.8 (P = 0.016)

Mapping the Li₂ locus

Using a large F₂ mapping population, we had previously reported the Li₂ mutation locus within a 0.3-cM interval on Chr. D13 (Thyssen et al., 2014). However, due to the use of an earlier G. raimondii diploid reference genome (Paterson et al., 2012), the mapping prediction had a certain degree of error. Previously we used super bulked segregant analysis sequencing (sBSAseq) of two pools of samples derived from 100 short fiber and 100 wild-type fiber F₂ plants for single nucleotide polymorphism (SNP) discovery and mapping-by-sequencing (Thyssen et al., 2014). We re-aligned the Illumina short reads from the two bulks of sequences to G. hirsutum TM-1 reference genome (Zhang et al., 2015). The alignment of short sequence reads from the Li₂ plant revealed that the end of Chr. D13 is missing a genomic section of about 177 kb (according to the assembled version of TM-1 reference genome used in this study), while a region of ∼221 kb before the deleted region was aligned with double sequence read depth (Figure 2A). Careful, manual inspection of alignment files and de novo assembly of the end of Chr. D13 from Li₂ and its wild-type revealed a large structural rearrangement at the end of Chr. D13, including a deletion and a tandem inverted duplication in the Li₂ mutant.

Figure 2.

Genetic mapping of the Li₂ locus. A, Structural re-arrangement at the end of Chr. D13 identified in Li₂ mutant by sBSAseq comparison of two pools of F₂ samples. B, Informative recombinant plants from F₂ mapping population. Letter “B” indicates wild-type homozygous SNP, whereas letter “H” is heterozygous SNP. C, Physical map of Li₂ locus on Chr. D13 of G. hirsutum (only fragment of Chr. D13 depicted).

To confirm whether or not the duplicated region is arranged as a tandem inverted repeat and was linked to the short fiber phenotype we designed a HairPin marker: forward primer placed in 3′UTR of Gh_D13G2437 gene and reverse primer placed partially in the short unique sequence at the junction that was revealed by de novo assembly and partially in the inverted repeat (Supplemental Figure S1). The HairPin primers confirmed the tandem inverted repeat and produced a PCR product only in Li₂ homozygous or heterozygous plants. We analyzed this HairPin marker in the entire F₂ populations. The marker genotype is completely co-segregated with the short fiber phenotype of the progeny plants, indicating the region CFB5849-HairPin is very tightly linked to the causative mutation, which includes the deleted and duplicated regions on Chr. D13 (Figure 2, B and C).

Small RNA analysis

Sometimes, a Li₂ plant produces mixed fiber phenotype. This observation indicates incomplete penetrance of the Li₂ mutation. Therefore, there is some factor in that structural rearrangement region that inconsistently triggers the short fiber phenotype in the Li₂ plant. To assess the involvement of small RNA-induced silencing, we evaluated small RNAs from developing fiber (8 DPA) of wild-type and Li₂ mutant. We found that siRNA was produced from Gh_D13G2437 (located at the junction of the structural variation) in the Li₂ but not in the wild-type NIL (Figure 3, A and B). Figure 3B shows distribution of siRNAs through the Gh_D13G2437, including introns, with largest peaks in the first and last exons, and 3′-untranslated region (UTR). The wild-type had negligible amount of siRNAs through the coding sequence of the gene compared to Li₂. None of the other genes in the structural rearrangement region of Chr. D13 produced small RNA.

Figure 3.

siRNA production from Gh_D13G2437 and its expression analysis. A, Simplified cartoon of structural rearrangement in Li₂ mutant, including large deletion and inverted duplication of part of Chr. D13. B, Graph of alignment of siRNAs to gene model of Gh_D13G2437 of wild-type and Li₂. C, RT–qPCR analysis of expression level of Gh_D13G2437 in developing fiber cells and vegetative tissues in wild-type and Li₂. Error bars indicate standard deviations from three biological replicates with two technical replicates.

We evaluated expression of Gh_D13G2437 by reverse transcription–quantitative polymerase chain reaction (RT–qPCR) in developing fibers from DOA to 20 DPA and in leaf, stem, and root tissues. The gene was preferentially expressed in developing fibers during the peak of elongation at 5–12 DPA in the wild-type fiber cells, though the expression of this gene was scarcely detected in elongating fibers of the Li₂ mutant (Figure 3C). This observation suggests that inverted repeat of the Gh_D13G2437 (located at the junction of the tandem inverted duplication) induced production of siRNAs that silences the gene in Li₂ during fiber elongation.

Expression analysis of genes from structural rearrangement region

We used RNAseq data obtained from developing fibers at the peak of elongation (8 DPA) from wild-type and Li₂ plants grown in greenhouse and field conditions. Two growth conditions were used to mitigate the effects caused by environments. Analysis of variance (ANOVA) of RNAseq samples was set in two comparisons between wild-type and Li₂ grown in greenhouse and field. Supplemental Data Set 1 provides ANOVA of expression of all predicted coding sequences in TM-1 genome (Zhang et al., 2015), whereas Supplemental Table S1 provides an extraction from this data, the expression of the genes from structural rearrangement. The deleted region harbors 27 genes according to the version of TM-1 reference genome used in this study. All of these, except the Gh_D13G2464 (possible errors in alignment of short reads due to high identity between homeologous copies or errors in genome assembly), showed no expression in the Li₂ samples. Fourteen of the 27 genes from deleted region showed substantial (higher than background) expression in wild-type developing fibers at 8 DPA.

The duplicated region harbors 29 genes; 20 of them were expressed above the background level in wild-type or Li₂ at 8 DPA. The Gh_D13G2437 was the most highly expressed gene in wild-type developing fibers at 8 DPA (Supplemental Table S1). Fifteen genes in the duplicated region were significantly (false discovery rate (FDR) < 0.05; at least two-fold) differentially expressed between wild-type and Li₂. Only Gh_D13G2437 was significantly downregulated in the Li₂, whereas the 14 remaining genes were significantly upregulated in the Li₂.

Evaluation of expression of genes from a plant with both short and long fiber cells

To evaluate expression levels of genes from the structural rearrangement in the Li₂ plant with a mixed fiber phenotype, we collected developing fibers at 16 DPA from a single heterozygous Li₂ plant (showing mixed phenotype) grown in greenhouse. As mentioned earlier, the appearance of clear mixed fiber phenotype is neither predictable nor controllable, as also reported by others (An et al., 2010). We were able to obtain by chance only a single Li₂ plant showing clear both types of fiber phenotype on the same plant. The purpose of the experiment with a single plant was to refine the list of Li₂ candidate genes for further functional analysis. The 16 DPA is approximately the end of cotton fiber elongation stage; however, at this time point, we can clearly distinguish long-fiber and short-fiber bolls collected from the same plant. At the same time, developing bolls at 16 DPA were collected from three wild-type and three homozygous Li₂ plants for comparison. Figure 4 shows RT–qPCR analysis of four genes each from duplicated and deleted regions, which were substantially expressed in wild-type at 8 DPA according to RNAseq data (Supplemental Table S1).

Figure 4.

Expression analysis of genes from duplicated and deleted regions on Chr. D13 by RT–qPCR. Developing fibers at 16 DPA were collected from greenhouse grown plants: from three homozygous plants each for wild type (WT) and Li₂, and from a single heterozygous plant with mixed phenotype for long fiber (Mix-long) and short fiber (Mix-short) samples. Error bars indicate standard deviations from three biological replicates for WT and Li₂, and three technical replicates for Mix-long and Mix-short samples. Fold change values of expression Mix-long/Mix-short fiber samples are shown above bars.

The genes from the Chr. D13 deleted region were not expressed in the Li₂ samples, which is consistent with the RNAseq data (Supplemental Table S1). However, in the heterozygous plant with a mixed fiber phenotype, the expressions of these genes were higher than in Li₂ and lower than in wild-type (the wild-type copy in heterozygous plant contributed to expression), and not much different between Mix-long and Mix-short fiber samples (Figure 4). The genes from the duplicated region, including Gh_D13G2411, Gh_D13G2423, and Gh_D13G2425, which were significantly (FDR < 0.05; at least two-fold) upregulated in Li₂ at the peak of elongation at 8 DPA were less than two-fold differently expressed at 16 DPA between wild-type and Li₂ and mixed long and short fiber samples (Figure 4). Only Gh_D13G2437 showed 15.3-fold different expression between Mix-long and Mix-short fiber samples. Therefore, taken together, the results from small RNA analysis and expression in Mix-long and Mix-short fiber samples indicate the Gh_D13G2437 is the most promising candidate for triggering the short fiber cell phenotype in the Li₂ mutant.

VIGS of Gh_D13G2437 in cotton

The Gh_D13G2437 is annotated as RanBP1. The predicted mRNA is 2,134-bp long and contains four exons. The coding sequence (651 bp) produces a small protein 216 amino acids and a predicted molecular mass of 24.1 kDa.

First, we tested suppression of Gh_D13G2437 in cotton by virus-induced gene silencing (VIGS). We used a tobacco rattle virus (TRV) vector system (Gao et al., 2011) for the VIGS experiment. The plants infiltrated with the experimental TRVg2437 construct did not develop leaf tissue after infiltration and eventually died (Supplemental Figure S2). The strong VIGS suppression of Gh_D13G2437 in vegetative tissues was possibly lethal to the plant; this also suggested a possible VIGS suppression of an important family of homologous genes. Therefore, we searched for homologous genes to Gh_D13G2437 in the G. hirsutum reference genome and found seven more genes with high sequence similarity to Gh_D13G2437. All eight G. hirsutum RanBP1s (Gh_RanBP1s) fell into the same phylogenetic clade along with Arabidopsis and human RanBP1s (Supplemental Figure S3). The AtRanBP1a was the closest homolog to GhRanBP1s, showing from 65% to 71% of amino acid sequence identity (Supplemental Figure S4). Supplemental Figure S5 shows alignment of G. hirsutum and Arabidopsis RanBP1s. The deduced amino acid sequences of the eight G. hirsutum genes showed very substantial sequence homology, from 68% to 98% (Supplemental Figures S4 and S5). Each G. hirsutum deduced protein consists of 216–227 amino acid residues, except Gh_D08G2142, which consists of 271 amino acid residues. Supplemental Figure S5 also shows that the Ran-binding domain (involved in the interaction with Ran GTPases) and a leucine-rich nuclear export signal (NES) are highly conserved among Arabidopsis and cotton RanBP1s. The NES in the C-terminus of the protein confines RanBP1 in the cytoplasm (Richards et al., 1996; Haasen et al., 1999). We isolated RNA from stem and root tissue of VIGS plants (since there was no leaf tissue to collect from experimental plants) and evaluated expression of Gh_D13G2437 and its homologs. RT–qPCR analysis confirmed reduction of expression level in TRVg2437 infiltrated plants of all tested homologous genes of this family (Supplemental Figure S6).

Expression analysis of RanBP1 family members

High sequence similarity among G. hirsutum RanBP1s suggests that siRNA produced from hairpin structure of Gh_D13G2437 might also target other members of RanBP1 family. Therefore, we mapped the siRNA reads to the alignment of coding sequences of eight RanBP1s to see their sequence identity for possible siRNA off-target effect (Supplemental Figure S7). The siRNA reads were range of sizes from 20 to 24 nucleotides that overlapped the area about 64% of coding sequences. Sites of high siRNA production are highlighted by red asterisks (Supplemental Figure S7). Analysis of the sequence alignment confirmed high sequence identity among RanBP1s and possible off-target effect. We also checked whether or not the siRNA of Gh_D13G2437 in the Li₂ suppressed expression of the family of homologous RanBP1 genes. Supplemental Table S2 provides identification of G. hirsutum genes involved in nucleocytoplasmic transport based on homology to functionally characterized genes in Arabidopsis, whereas Supplemental Table S3 provides differently expressed genes among those identified. The RNAseq analysis of elongating fibers at 8 DPA showed reduced expression of eight Gh_RanBP1s in the Li₂ (Supplemental Table S3).

This raised the question: why was VIGS suppression of family Gh_RanBP1s lethal to the plant, whereas siRNA suppression of the same family genes in the Li₂ was not? Therefore, we tested expression of members of RanBP1 family in developing fibers and vegetative tissues. The genes of this family were higher expressed in developing fibers during elongation 5–16 DPA than in vegetative tissues in wild-type (Supplemental Figure S8). An independent expression data from ccNET database (http://structuralbiology.cau.edu.cn/gossypium/; You et al., 2016) showed similar results—the highest expression of eight Gh_RanBP1s in developing fibers at 5 and 10 DPA (Supplemental Figure S9). In the Li₂ the expression of these genes reduced substantially in elongating fibers, but not in the vegetative tissues (Supplemental Figure S8).

Transformation of Gh_D13G2437 in cotton

To reproduce the Li₂ phenotype we used RNAi silencing constructs for stable agrobacterium-mediated transformation of the Gh_D13G2437 gene under 35S and fiber-specific promoters into cotton. We were unable to regenerate positive transgenic plants from RNAi silencing constructs for five years of transformation. The next approach was introduction of an overexpression (OE) construct of Gh_D13G2437 into cotton. We obtained eight positive OE transgenic plants, expressing the Gh_D13G2437 with different degrees of intensity (Supplemental Figure S10). Three positive OE plants showed at least six-fold higher expression level of the Gh_D13G2437 than transformed plants with pBI121 vector itself or untransformed Coker-312 plants (Supplemental Figure S10).

The transgenic OE cotton plants show no visible phenotypic changes in vegetative tissues or seed fiber (Supplemental Figure S11). The high volume instrument (HVI) measurement of fiber samples from OE plants did not detect significant changes in fiber quality characteristics compared to controls (Supplemental Table S4; Supplemental Figure S12). Therefore, producing more Gh_D13G2437 transcripts in cotton does not phenotypically affect the plant. However, VIGS suppression of the family of Gh_RanBP1s was lethal to the plant, while siRNA suppression of the same family during fiber elongation only caused shorter fiber length. Our next approach to prove that the Gh_D13G2437 is the causative gene was to cross the transgenic line with the highest levels of transgene expression (OE-2) with the Li₂ mutant for complementation analysis. The purpose of this cross was to introduce the highly expressed transgenic copy of Gh_D13G2437 to partially compensate for the siRNA-induced suppression of this gene in the Li₂ mutant.

From the crosses of the transgenic line with the highest levels of transgene expression OE-2 (hemizygous) and three Li₂ (homozygous) plants, we obtained 12 F₁ plants. PCR screening for presence of transgene copy detected six positive independent plants that also harbored the Li₂ mutant allele (Supplemental Figure S13). The seeds from positive F₁ plants produced longer fibers compared to the seeds of PCR-negative F₁ plants, which also contained the dominant Li₂ mutant allele but lacked the complementary Gh_D13G2437 (Figure 5A). The RT–qPCR analysis of F₁ plants confirmed the higher expression of Gh_D13G2437 gene in positive F₁ plants (Figure 5B). We have tested siRNA level in developing fibers of F₁ plants. The amount of detected siRNA was not substantially changed in F₁ plants comparing to the Li₂ (Supplemental Figure S14). However, due to OE of additional copy transcript level of the Gh_D13G2437 was higher in positive transgenics comparing to Li₂ and negative F₁ plants (Figure 5C). To ensure that the restored phenotype is not caused by other genes in the structural rearrangement region we examined their expression level in the F₁ plants (Supplemental Figure S15). The expression level of genes from deleted region were not detectable in the Li₂ and F₁s. Genes from duplicated region were not substantially differently expressed between the Li₂ and F₁s, except Gh_D13G2437, which showed significantly (P < 0.0001) higher expression in positive F₁s then negative F₁s (Supplemental Figure S15). This result provides additional evidence that partial restoration of phenotype caused by OE of Gh_D13G2437.

Figure 5.

Seed fiber phenotype and Gh_D13G2437 expression analysis of F₁ plants. A, Seed fiber phenotype of F₁ plants obtained from crosses of Li₂ with OE2. (−) indicates that the F₁ plant does not have a transgenic copy, whereas (+) does have. RT–qPCR analysis of expression level of Gh_D13G2437 in leaf tissue (B) and fiber at 8 DPA (C) of F₁ plants; WT is Coker-312 which was used for transformation; OE-2, the transgenic line with the highest levels of transgene expression. Error bars indicate standard deviations from three technical replicates.

Transcriptional analysis of genes involved in Ran GTPase cycle in the Li₂ fiber cells

Next, we examined how siRNA suppression of Gh_RanBP1s affected expression of the genes involved in nucleocytoplasmic transport in the Li₂ developing fibers. Currently, only a handful of genes that are involved in nucleocytoplasmic transport have been characterized in G. hirsutum; therefore, we searched homologous genes to identify related genes that have been characterized in Arabidopsis, reviewed in Merkle (2011) and Nielsen (2020). We have identified 185 genes in G. hirsutum genome that are homologous to Arabidopsis genes involved in nucleocytoplasmic transport (Supplemental Table S2). Among 185 such genes, 155 are expressed in developing fiber cells at detectable levels (Supplemental Data Set 1), and 42 are significantly (P < 0.01) differentially expressed between wild-type and the Li₂ mutant at least 1.6-fold (Supplemental Table S3).

The siRNA-induced silencing of Gh_RanBP1s in Li₂ fiber cells remarkably perturbed expression of genes involved in RanGTPase cycle. The RanGTPase family is currently represented by six genes in G. hirsutum, which are highly expressed in developing fiber cells. Four out of these six RanGTPases increased their expression from 1.6- to 2.8-fold in response to deficiency of Gh_RanBP1s in the Li₂ fiber cells. The conversion of RanGTP into RanGDP requires RanGAP1 with its activating factor RanBP1 (Pay et al., 2002). Three out of four G. hirsutum RanGAP1 genes were upregulated (Supplemental Table S3). An AtWIP1 homolog (Gh_D13G1341) that anchors RanGAP1 to nuclear envelope (Xu et al., 2007) had higher transcript abundance in the Li₂ than in the wild-type. Six G. hirsutum homologs to Arabidopsis EXPORTIN1A [XPO1A] (Wu et al., 2010), were upregulated in the Li₂ fiber cells. Seven genes annotated as importin β were downregulated, whereas three were upregulated in the Li₂. Gh_A05G1683 homolog of Arabidopsis BZR1 that functions as a nucleocytoplasmic shuttling protein (Ryu et al., 2007) was downregulated in the Li₂. In G. hirsutum BZR1 (BRASSINAZOLE RESISTANT 1) is involved in regulation of fiber initiation and elongation by modulating brassinosteroid signaling (Zhou et al., 2015).

Discussion

In this study, we detected a large structural rearrangement at the end of Chr. D13 in the Li₂ short cotton fiber mutant, including a deletion and a tandem inverted duplication. An independent study has also detected a terminal deletion of Chr. 18 (Chr. D13) and concluded that the missing part of chromosome might be responsible for the Li₂ phenotype (Patel et al., 2020). At the junction of the terminally duplication of the Li₂ mutant Chr. D13, Gh_D13G2437 is positioned just 409 bp from the coding sequence of its the tandem inverted repeat, which follows a small spacer (23 bp) between the inverted genes (Figure 3;Supplemental Figure S1). During transcription, such a structure may form a long self-complementary hairpin RNA, which is highly effective at inducing RNAi silencing (Waterhouse et al., 1998; Wang and Waterhouse, 2000). As evidence we observed production of siRNAs in developing fiber cells at 8 DPA from the Gh_D13G2437 only in the Li₂ and not in wild-type, while none of the other genes from the deleted and duplicated regions produced small RNAs. We do not think that deleted part itself can cause the phenotype in allotetraploid species (though it possible in diploids), because the functions of deleted genes will be complemented by homoeologous genes from second set of chromosomes. Also, deleted part cannot explain the mixed fiber phenotype. However, siRNA is very efficient silencer that can target highly identical sequences of homeologous and homologous genes. Analysis of the sequence alignment confirmed high sequence identity among RanBP1 family members, which can be targeted by siRNA induced from inverted repeat of Gh_D13G2437 (Supplemental Figure S7).

We observed that the Li₂ mutant sometimes shows a mixed short/long fiber phenotype on the same plant or in the same boll. Evaluation of gene expression in these two types of fibers from the same plant showed that only Gh_D13G2437 was substantially less expressed in short fiber than in long fiber sample. Still, expression of Gh_D13G2437 in the long fiber sample from the mixed fiber plant was much lower than in wild-type which implied there might be some threshold of expression that can produce the wild-type phenotype. This has been well documented from studies on the nematode Caenorhabditis elegans that variability in gene expression underlies incomplete penetrance (Raj et al., 2010). Originally small noncoding RNAs were considered merely regulate gene expression at the transcriptional/posttranscriptional level. Now, there is a growing body of evidence that regulatory noncoding RNAs play an important role in epigenetic control. Studies in Arabidopsis demonstrated that posttranscriptional gene silencing, and the accompanying DNA methylation of target loci, associated with the production of siRNAs and therefore linked RNA-directed DNA methylation to the RNAi pathway (Hamilton and Baulcombe, 1999; Jones et al., 1999; Mette et al., 1999; Dalmay et al., 2000; Matzke et al., 2001).

The randomness of fiber phenotype was observed in another cotton fiber mutant. The N₁ is the dominant fuzzless-linted seed mutant that has a wide range of fiber densities, even on the bolls of a single plant (Wan et al., 2016). The N₁ mutation was linked to a gene on Chr. A12 that was annotated as an MML transcription factor GhMML3_A12 (Wan et al., 2016). siRNA originated from bidirectional transcripts of GhMML3_A12 reduced its transcript level and therefore caused the fuzzless seed phenotype (Wan et al., 2016). Wan et al. (2016) suggested that the randomness of the small RNA distribution pattern and the epigenetic modifications might be associated with the randomness of the fiberless phenotype in the N₁ mutant. Therefore, factors such as epigenetic modifications or other mechanisms that modulate penetrance should not be ruled out in the Li₂ mutant.

We observed that VIGS suppression of the family of Gh_RanBP1s was lethal to the plant, while siRNA suppression of the same family in the Li₂ was not. The Gh_D13G2437 was more highly expressed in elongating fiber cells than in vegetative tissues of cotton (Figure 3C). Therefore, following the transcriptional program of Gh_D13G2437, the tandem inverted repeat structure of the gene in the Li₂ produces self-complementary hairpin RNAs with higher rate during fiber elongation, which consequently leads to higher rate of siRNA silencing of a family of Gh_RanBP1s during fiber elongation. The higher rate of silencing of Gh_RanBP1s specifically during fiber elongation corresponds with shorter fiber phenotype in Li₂. Expression of Gh_RanBP1s in vegetative tissues was not substantially different between wild-type and Li₂ (Figure 3C;Supplemental Figure S8) that coincides with normal vegetative morphology.

It is more difficult to explain occurrence of two types of fiber on a single boll (Figure 1D). We can assume that siRNA/epigenetic regulations have occurred differently among ovules in the boll. The fertilized ovules are separated by the locules in the ovary of the flower and eventually in developing boll. Ovules are attached to placenta by funiculus through which nutrients are received. We know that the siRNA production coincides with higher expression of Gh_D13G2437, mainly during rapid fiber elongation time (Figure 3C). So, majority of siRNAs are housed in the elongating fiber cells of the Li₂. Molecules can be transported through plasmodesmata in plant cells. It is possible that some amount of siRNA is transported thought plasmodesmata from epidermal fiber cells through funiculus to placenta and to other ovules, but it is not enough to silence fiber development on ovules in adjacent locules.

Exchanges of proteins and other macromolecules across the nuclear envelope are essential during cotton fiber as well as other cells development. RanGTPases play a pivotal role in driving and regulating the active transport of cargos between the nucleus and cytoplasm (Moore, 1998; Floch et al., 2014; Moore, 2021). However, the role of the RanBP1, that is currently believed enhances the hydrolysis of GTP on Ran (Bischoff et al., 1995), might be underestimated. We observed that strong VIGS suppression of the RanBP1 family in G. hirsutum plants was lethal (Supplemental Figure S2). Similar results were observed in mice where knockout of RanBP1 exhibited male infertility due to a spermatogenesis arrest, while siRNA-mediated depletion of RanBP2 caused severe cell death only in RanBP1-deficient mouse embryonic fibroblasts, indicating that simultaneous depletion of RanBP1 and RanBP2 severely affects normal cell viability (Nagai et al., 2011).

Expression of the genes involved in RanGTPase cycle was perturbed remarkably in the Li₂ fiber cells (Supplemental Table S3). The Ran gradient, or asymmetry in the cellular distribution of RanGTP and RanGDP bound forms, is the key to establish the transport directionality (Moore, 1998; Floch et al., 2014; Nielsen, 2020). During nuclear import, RanGDP/importin/cargo complexes dissociate upon entry into nucleus, as a result of conversion of RanGDP to RanGTP by nuclear-localized RanGEF action. In case of nuclear export, RanGTP/exportin/cargo complexes dissociate upon conversion of RanGTP to RanGDP through the action of RanGAP1 and RanBP1 activity. It is tempting to speculate that deficiency of RanBP1 leads to accumulation of the undissociated RanGTP/exportin/cargo complexes and, therefore a disturbance of RanGTPase cycle in the Li₂. It has been shown in vitro that Ran mutant with ∼20-fold reduced affinity to RanBP1 promotes nuclear export and inhibits nuclear import (Kehlenbach et al., 2001). Results of Kehlenbach et al.’s (2001) study provide physiological evidence that release of RanGTP from the nuclear transport receptors (NTRs) by RanBP1 is critical for the recycling of NTRs to a transport-competent state. In agreement with Kehlenbach et al.’s (2001) study, we observed simultaneous upregulation of six out of seven identified exportins (XPO1A) and downregulation of seven importins β (Supplemental Table S3). Therefore, expression changes of genes involved in nuclear transport suggest that suppressed RanBP1’s activity perturbs the Ran GTPase cycle and prevents the functioning of the Ran signaling system during the cotton fiber cells development in the Li₂.

Materials and methods

Plant materials

Detailed descriptions of the development of the cotton (G. hirsutum L.) wild-type and mutant Li₂ NILs used in this study have been reported earlier (Hinchliffe et al., 2011). For development of the mapping populations the parental NILs, the Li₂ and DP5690, were crossed to produce F₁ seeds. The resulting F₁ plants were grown in a greenhouse and were self-pollinated to produce F₂ seeds. Two F₂ populations were grown in Stoneville, MS, USA for two consecutive years. The first population of 515 F₂ plants was grown during the summer of 2012, and the second population of 1,185 F₂ plants was grown during summer of 2013. Standard conventional field practices were applied during the plant growing season.

Li₂ and wild-type NILs (∼100 plants per each line) were grown in the field in New Orleans, LA, USA in 2013 for fiber samples collection. The individual mutant and wild-type cotton plants were labeled into three pools representing three biological replicates. All samples of the same developmental stage were tagged and collected on the same day. Cotton bolls were harvested at DOA, 3, 5, 8, 12, 16, and 20 DPA. The number of bolls per bulked sample varied according to developmental time point, with a greater number of bolls required for the earliest time point to ensure sufficient biological material. For example, approximately 30 bolls were bulked for each DOA sample, and ∼10 bolls were bulked for each 20 DPA sample.

Transgenic plants were grown in greenhouse in five gallons pots filled with Metromix-360 soil (Sun Gro Horticulture, Quincy, MI, USA) under natural light and 2 times per day watering schedule. Plants were fertilized with Osmocote Classic 14-14-14 (ICL Specialty Fertilizers, North Charleston, SC, USA).

sBSAseq

F₂ plants from the first mapping population of 515 individuals were selected for sequencing by a bulked segregant approach (Michelmore et al., 1991; Takagi et al., 2013). sBSAseq of this mapping population was previously reported (Thyssen et al., 2014). Briefly: two DNA pools were constructed, including a pool of DNA from 100 short fiber (Li₂/Li₂ and Li₂/li₂) and a pool of DNA from 100 wild-type (li₂/li₂) plants. DNA was Illumina sequenced with paired 101-bp reads flanking 150-bp inserts by Data2Bio LLC (Ames, IA, USA). In this study, the Illumina short reads from the two bulks of sequences were aligned to the tetraploid G. hirsutum TM-1 reference genome (Zhang et al., 2015).

RNA isolation and RT–qPCR analysis

Cotton fiber cells were separated from ovules at early developing time points as described before (Taliercio and Boykin, 2007). Developing fibers from 8 to 20 DPA ovules were manually separated using forceps. Total RNA was isolated from detached fibers using a Sigma Spectrum Plant Total RNA Kit (Sigma-Aldrich, St Louis, MO, USA) with the optional DNase1 digestion according to the manufacturer’s protocol. The concentration of each RNA sample was determined using a NanoDrop 2000 spectrophotometer (NanoDrop Technologies Inc., Wilmington, DE, USA). The RNA quality for each sample was determined using an Agilent Bioanalyzer 2100 and the RNA 6000 Nano Kit Chip (Agilent Technologies Inc., Santa Clara, CA, USA) with 250 ng of total RNA per sample. The cDNA synthesis reactions were performed using the iScript cDNA Synthesis Kit (Bio-Rad Laboratories, Hercules, CA, USA) according to the manufacturer’s instructions with 1 μg of total RNA per reaction. The RT–qPCR reactions were performed with iTaq SYBR Green Supermix (Bio-Rad Laboratories Inc., Hercules, CA, USA) in a Bio-Rad CFX96 real-time PCR detection system. The endogenous G. hirsutum 18S rRNA (Genbank U42827) was used as a reference for normalization. Primers were designed using Primer-BLAST software (Ye et al., 2012). The reference G. hirsutum cDNA sequences (Zhang et al., 2015) were used to design specific primers. The relative expression levels were calculated using the standard 2^−ΔΔCt method (Pfaffl, 2001). Primer sequences are listed in Supplemental Table S5.

Measurements of siRNA transcripts were conducted as previously described (Cirera and Busk, 2014). In brief, 100 ng of total RNA was incubated with 1 μL of 10× reaction buffer of Escherichia coli poly(A) polymerase (New England BioLabs Inc., Ipswich, MA, USA), 0.1-mM dNTP, 0.1-mM ATP, 1-μM universal RT primer (5′-CAGGTCCAGTTTTTTTTTTTTTTTVN), 1 U of E. coli poly(A) polymerase (New England BioLabs Inc.), and 100 U of M-MuLV reverse transcriptase (New England BioLabs Inc.) in 10-μL reaction mixture for 1 h at 42°C. The reaction was inactivated by heating at 95°C for 5 min and cDNA was diluted 50 times before being used in RT–qPCR. Small RNA-specific primers were designed with free available software (Busk, 2014). The RT–qPCR reactions were performed with iTaq SYBR Green Supermix (Bio-Rad Laboratories Inc., Hercules, CA, USA) in a Bio-Rad CFX96 real time PCR detection system. 5.8S rRNA was used as the internal reference gene for normalization of RT–qPCR data. Primer sequences are listed in Supplemental Table S5.

Small RNA sequencing and processing

RNA samples were extracted from developing fibers at 8 DPA (in three biological replicates) and pooled together for preparation of wild-type and Li₂ small RNA libraries. The small RNA libraries preparation and sequencing were conducted by LC Science (Houston, TX, USA) using 1 μg of total RNA according to the TruSeq Small RNA sample preparation guide (Illumina, San Diego, CA, USA). The libraries were sequenced using the Illumina Hiseq 2,500 platform. The small RNA sequencing sequence data were deposited to NCBI-SRA (PRJNA307581). The global alignment and analysis of these reads were described previously (Naoumkina et al., 2016); however, in this study the existence of siRNA from Gh_D13G2437 and other genes in the structural rearrangement near the telomere of D13 were carefully and manually inspected in new alignment files generated using HISAT2 software (Kim et al., 2019) to align the short reads to the reference genome (Zhang et al., 2015).

RNAseq and data processing

Sequencing of RNA samples from greenhouse and field grown plants were previously reported (Naoumkina et al., 2014, 2015). Briefly, libraries preparation and sequencing were performed by Data2Bio LLC (Roy J. Carver Co-Laboratory, Ames, IA, USA). The libraries were sequenced using 101-bp paired-end reads. The RNA samples from developing fibers at 8 DPA of greenhouse-grown plants were sequenced in two biological replicates, whereas for the field grown plants were sequenced in three biological replicates. RNAseq data from both sets of plants are available in NCBI-SRA (PRJNA209459 and PRJNA273732).

Mapping the reads to the G. hirsutum TM-1 coding DNA sequences (Zhang et al., 2015) was performed using GSNAP software (Wu and Nacu, 2010). Default parameters were used, but with the flags “-n 1 -Q” which means that only a single mapping was reported for each read, and reads with multiple equally good hits were discarded rather than randomly mapped.

VIGS assay

A cDNA was synthesized from RNA isolated from leaf tissue of the wild-type NIL (G. hirsutum cv. DP5690). One microgram of total RNA was used in a first-strand synthesis using SuperScript III reverse transcriptase (Invitrogen, Carlsbad, CA, USA) in a 20-µL reaction with oligo(dT) primers according to the manufacturer’s protocol. A 610-bp fragment of Gh_D13G2437 cDNA, beginning from its ATG start and covering 94% of the cDNA, was amplified from cDNA by PCR using Q5 high fidelity DNA polymerase (New England BioLabs, Ipswich, MA, USA) and specific primers (Supplemental Table S5). Specific primers were design by using Primer-BLAST software (Ye et al., 2012). The primers contained over-hang sequences (15–18 bp) for Gibson cloning into the pTRV2 vector. The pTRV2 vector was digested with EcoRI/KpnI, gel purified and the Gh_D13G2437 gene fragment was inserted using HiFi DNA assembly cloning kit (New England BioLabs, Ipswich, MA, USA). The sequence accuracy of the insert was confirmed by Sanger sequencing. This construct was named TRVg2437. Two positive control constructs were used in experiment, including magnesium chelatase subunit I (Chl1) and Cloroplastos alterados 1 (Cla1). The Chl1 fragment (501 bp) plus over-hang sequences (26–35 bp) for Gibson cloning into the pTRV2 vector was synthesized by IDT (Integrated DNA Technologies, Inc., Coralville, IA, USA). The sequence of the fragment is provided in Supplemental Figure S16. The same cloning strategy was used as described above, except that the pTRV2 vector was digested with EcoRI/SacI. The cloning of the Cla1 positive control was previously described (Gao et al., 2011).

The pTRV1 helper plasmid, along with either TRVg2437, the VIGS positive controls Chl1 or Cla1, and a negative control empty vector pTRV0 were introduced into the Agrobacterium tumefaciens strain GV3101 by electroporation (Bio-Rad, Hercules, CA, USA). The VIGS experiment followed a previously published protocol (Gao et al., 2011). Seedlings with mature cotyledons but without a visible true leaf (7 days after germination) were infiltrated by inserting the Agrobacterium suspension into the cotyledons using a syringe. The plants, including nine in the experimental group and six each in of the controls, were grown in pots at 22°C in a growth chamber under a 12-h light and 12-h dark cycle. RNA was isolated from stem and root tissues 3 weeks after infiltration (4-week old plants). Expression of Gh_D13G2437 and its homologs in the VIGS plants were analyzed using RT–qPCR (primer sequences shown in Supplemental Table S5).

Cotton transformation for gene complementation

A full-length cDNA sequence of Gh_D13G2437 plus over-hanging sequences for Gibson cloning into the pBI121 vector was synthesized by IDT (Integrated DNA Technologies, Inc., Coralville, IA, USA). The sequence of the insert is provided in Supplemental Figure S17. The pBI121 vector was digested with BamHI/SacI, gel purified and the cDNA of Gh_D13G2437 was inserted using HiFi DNA assembly cloning kit (New England Biolabs, Ipswich, MA, USA). The construct was named “OE” for OE of the Gh_D13G2437 cDNA in pBI121 vector and was expressed in the transformed plants from the constitutive S35 promoter in the T-DNA of pBI121.

The construct was transformed into A. tumefaciens strain GV3101 by electroporation followed by the standard hypocotyl infiltration method for cotton transformation (Trolinder and Goodin, 1987). Selection of transgenic plants was carried out on media with 50 μg L⁻¹ of kanamycin. The positive transgenic lines were identified by PCR using one primer from the 35S promoter (pBI121-35S) and a second primer from the coding sequence of Gh_D13G2437 (Supplemental Table S5).

Fiber quality measurements

Cotton bolls were collected from the middle part of the plant. The cotton samples were ginned using a laboratory roller gin. Fiber quality traits of transgenic and control plants were measured using a High Volume Instrument (HVI, Uster Technologies Inc, Charlotte, NC, USA). The standard HVI fiber quality traits, described below, were measured by the Cotton Fiber Testing Lab, USDA-ARS-SRRC, New Orleans, LA, USA. Upper half mean length (UHML) is the average length (mm) of the 50% longest fibers by weight. Micronaire is a measure of the air permeability of compressed cotton fibers, and it is often used as an indication of fiber fineness and maturity. Uniformity index is the ratio between the mean length and the UHML of the fibers, expressed as a percentage. Short fiber content is the percent of fibers <12.7 mm in a fixed weight of fibers. Maturity ratio indicates the fiber maturity in terms of the degree of thickening of the secondary cell wall relative to the diameter or fineness of the fiber.

Phylogenetic analysis

Phylogenetic analysis was conducted in MEGA version 6 (Tamura et al., 2013) using the neighbor-Joining method (Saitou and Nei, 1987). The bootstrap test was performed in 1,000 replicates (Felsenstein, 1985). The evolutionary distances were computed using the Poisson correction method (Zuckerkandl and Pauling, 1965) and are in the units of the number of amino acid substitutions per site. The analysis involved 23 amino acid sequences. All positions containing gaps and missing data were eliminated. There were a total of 151 positions in the final dataset. The deduced proteins of G. hirsutum genes (placed at the end of tree branches) used in phylogenetic analysis can be found in the CottonGen database (https://www.cottongen.org). Human RanBPs used in analysis can be accessed at National Center for Biotechnology Information database (https://www.ncbi.nlm.nih.gov/protein/) and Arabidopsis RanBPs can be accessed at The Arabidopsis Information Resource (https://www.arabidopsis.org/).

Statistical analyses

The JMP/Genomics version 9.0 software (SAS, Cary, NC, USA) was used for RNAseq data normalization and statistical analysis. The RNAseq data were normalized using the trimmed mean of M component method (Robinson and Oshlack, 2010). An ANOVA process was conducted as previously described (Naoumkina et al., 2014). The liner model was used to test the null hypothesis that expression of a given gene was not different. Specifically, two comparisons were made between wild-type and Li₂ mutant grown in greenhouse and field environments. We identified genes for which the difference in expression between wild-type and the Li₂ mutant were significantly different with false discovery rate <0.05 (Benjamini and Yekutieli, 2001).

A one-way ANOVA statistical test with Tukey correction for multiple comparisons (GraphPad Prism version 7, San Diego, CA, USA) was applied to identify significantly differentially expressed samples in RT–qPCR data.

Accession numbers

The G. hirsutum RanBP1 genes discussed in this research can be found in the CottonGen database (https://www.cottongen.org) as follows: Gh_D13G2437; Gh_A01G1959; Gh_A05G0932; Gh_A08G1793; Gh_A13G2035; Gh_D01G2219; Gh_D05G1017; and Gh_D08G2142. The small RNA sequence data are available in the NCBI Sequence Read Archive (SRA), accession PRJNA307581. RNAseq data are available in the NCBI SRA, accessions PRJNA209459 and PRJNA273732.

Supplemental data

The following materials are available in the online version of this article.

Supplemental Figure S1. Partial sequence of inverted repeat of the Gh_D13G2437 detected in the Li₂ mutant.

Supplemental Figure S2. VIGS of Gh_D13G2437.

Supplemental Figure S3. Phylogenetic analysis of RanBP-like proteins of G. hirsutum.

Supplemental Figure S4. Amino acid sequence similarity between Arabidopsis and G. hirsutum RanBP1s.

Supplemental Figure S5. Amino acid alignment of G. hirsutum and Arabidopsis RanBP1s.

Supplemental Figure S6. RT–qPCR analysis of Gh_D13G2437 and its homologous genes from VIGS infiltrated plants.

Supplemental Figure S7. Mapping of siRNAs produced from Gh_D13G2437 to alignment of coding sequences of G. hirsutum RanBP1s.

Supplemental Figure S8. RT–qPCR analysis of homologous genes to Gh_D13G2437 in developing fibers and vegetative tissues of wild-type and Li₂.

Supplemental Figure S9. Heat map of expression of G. hirsutum RanBP1 genes.

Supplemental Figure S10. RT–qPCR analysis of expression level of Gh_D13G2437 gene in transgenic cotton lines over-expressing this gene.

Supplemental Figure S11. Seed fiber phenotype of transgenic and control plants.

Supplemental Figure S12. Statistical analysis of fiber length trait between transgenic OE and control groups of samples.

Supplemental Figure S13. PCR screening of F₁ plants from crosses of the Li₂ with the transgenic line with the highest levels of transgene expression (OE-2).

Supplemental Figure S14. RT–qPCR analysis of siRNA produced from inverted repeat of Gh_D13G2437 in developing fibers of F₁ plants.

Supplemental Figure S15. Expression analysis of genes from structural rearrangement in F₁ plants.

Supplemental Figure S16. DNA sequence of magnesium chelatase subunit I (Chl1) with overhangs for Gibson cloning used in VIGS assay.

Supplemental Figure S17. DNA sequence of Gh_D13G2437 coding sequence with overhangs for Gibson cloning into pBI121 binary vector.

Supplemental Table S1. RNAseq analysis of genes from the Li₂ structural rearrangement.

Supplemental Table S2. Identification of G. hirsutum genes involved in Ran GTPase cycle.

Supplemental Table S3. Expression of Ran GTPases and their interacting proteins in Li₂ and wild-type developing fibers at 8 DPA

Supplemental Table S4. HVI measurements of fiber traits of transgenic lines overexpressing Gh_D13G2437 and controls.

Supplemental Table S5. Sequences of primers.

Supplemental Data Set 1. ANOVA of RNAseq samples.

 

D.D.F. and M.N. conceived the research. M.N. designed expression, VIGS, and transformation experiments, and wrote the manuscript. G.N.T. performed bioinformatics analysis, identified the structural rearrangement, and designed molecular markers. D.D.F. oversaw mapping experiments. C.B.F. conducted cotton transformation and VIGS experiments. P.L. conducted mapping experiments. All authors read and approved the manuscript.

The author responsible for distribution of materials integral to the findings presented in this article in accordance with the policy described in the instructions for authors (https://academic.oup.com/plphys/pages/general-instructions) is: Marina Naoumkina (marina.naoumkina@usda.gov).