A dominant negative mutation of GhMYB25-like alters cotton fiber initiation, reducing lint and fuzz

Abstract

Cotton (Gossypium hirsutum) fibers, vital natural textile materials, are single-cell trichomes that differentiate from the ovule epidermis. These fibers are categorized as lint (longer fibers useful for spinning) or fuzz (shorter, less useful fibers). Currently, developing cotton varieties with high lint yield but without fuzz remains challenging due to our limited knowledge of the molecular mechanisms underlying fiber initiation. This study presents the identification and characterization of a naturally occurring dominant negative mutation GhMYB25-like_At^hapT, which results in a reduced lint and fuzzless phenotype. The GhMYB25-like_At^hapT protein exerts its dominant negative effect by suppressing the activity of GhMYB25-like during lint and fuzz initiation. Intriguingly, the negative effect of GhMYB25-like_At^hapT could be alleviated by high expression levels of GhMYB25-like. We also uncovered the role of GhMYB25-like in regulating the expression of key genes such as GhPDF2 (PROTODERMAL FACTOR 2), CYCD3; 1 (CYCLIN D3; 1), and PLD (Phospholipase D), establishing its significance as a pivotal transcription factor in fiber initiation. We identified other genes within this regulatory network, expanding our understanding of the determinants of fiber cell fate. These findings offer valuable insights for cotton breeding and contribute to our fundamental understanding of fiber development.

A dominant negative mutation of GhMYB25-like inhibits fuzz initiation but only slightly reduces lint initiation by compensating with higher wild-type protein level during lint cell differentiation.

IN A NUTSHELL.

Background: Cotton fibers, vital natural textile materials, are single-cell trichomes that differentiate from the ovule epidermis. It is also a good model for elucidation mechanisms of cell fate determination. Cotton fiber can be categorized into two types, namely, “lint” and “fuzz.” Lint serves as a significant natural textile material, while the presence of fuzz leads to lint losses during the ginning process. Breeding cultivars that are fuzzless and have a higher lint percentage can profoundly benefit cotton production. Despite several loci controlling fiber initiation have been reported, fuzzless is always tightly linked to low lint percentage in these fuzzless mutants. Therefore, the genetic basis of these loci controlling lint and fuzz initiation needs to be clearly clarified before these loci can be applied for breeding.

Question: What are the specific molecular mechanisms underlying lint and fuzz initiation in fiber mutants?

Findings: A dominant negative mutation GhMYB25-like_At^hapT was identified to be responsible for the mutant phenotype with less lint and no fuzz (referred to as the LL mutant). The single-cell atlas of wild-type and LL mutant revealed the wild-type protein GhMYB25-like is a central hub in the regulatory network governing fiber initiation, regulating the expression of key genes, such as GhPDF2, CYCD3;1 and PLD. The dominant negative mutation GhMYB25-like_At^hapT not only lost its own ability to bind promoters of downstream genes but also exerted a dominant negative effect by repressing the transcriptional activation of the wild-type protein. While GhMYB25-like_At^hapT exerted a dominant negative effect on fuzz initiation, its impact on lint initiation was mitigated by the high doses of GhMYB25-like during lint initiation.

Next Steps: The exact function of GhMYB25-like needs further clarification, whether it participates in fiber cell differentiation or maintains the differentiation state of fiber cells. Besides, identifying upstream factors of GhMYB25-like is necessary, which should be the key determinant for cell fate. Specifically decreasing the expression of GhMYB25-like during fuzz initiation could lead to the development of fuzzless cotton varieties with a high lint percentage, thereby benefiting cotton production.

Introduction

Trichomes are unique outgrowths that differentiate from epidermal cells in plants, displaying a wide range of morphology and functions (Serna and Martin 2006). In particular, cotton (Gossypium spp.) fibers are specialized single-cell trichomes that originate from the ovule epidermis (Stewart 1975). Cotton fibers have the remarkable ability to elongate up to 60 mm while maintaining a diameter of only 11 to 20 μm, making them distinct seed trichomes in the plant kingdom (Mansoor and Paterson 2012). Originally, the primary function of cotton fiber was to aid in the dispersal of seeds (Wang et al. 2022). However, through prolonged domestication and selection, the fiber has undergone substantial changes, resulting in increased length and transforming it into a valuable material for the textile industry (Wang et al. 2019b).

Cotton (Gossypium hirsutum) fibers can be classified into two types, lint and fuzz (Stewart 1975). Lint is the primary source of textile material, while the presence of fuzz leads to lint loss during the ginning process (Haigler et al. 2012; Bechere et al. 2014). One of the major goals in cotton breeding is to develop varieties with a higher percentage of lint while eliminating fuzz. However, the lint percentage (LP) of fuzzless cotton is substantially lower than that of normal cotton with fuzz (Ware 1940; Ware et al. 1947; Turley and Kloth 2002; Wan et al. 2016). Therefore, it is imperative to investigate the underlying molecular mechanism that regulates the differentiation and initiation of lint and fuzz cotton fibers.

Trichome formation has been extensively studied as a model system for cell fate determination in Arabidopsis (Arabidopsis thaliana). In contrast to the well-understood trichome development in Arabidopsis, which is primarily controlled by a complex of MYB-bHLH (basic helix-loop-helix)-WD40 (tryptophan-aspartic acid (WD)-repeat protein) transcription factors (TFs) (Oppenheimer et al. 1991; Walker et al. 1999; Payne et al. 2000; Zhang et al. 2003), the initiation and differentiation of cotton fiber cells is a complex process regulated by multiple genes involved in cell differentiation (Walford et al. 2012), hormone signaling (Zhang et al. 2011), sugar signaling (Wang et al. 2014), and other pathways. Specifically, 2 important TF families, MYB and homeodomain leucine-zipper (HD-ZIP), are known to play crucial roles in the fiber-initiation process. Suppression of GhMYB25 or GhHD-1 in cotton has been shown to inhibit fiber initiation, while CRISPR-mediated knockout of GhPDF2 results in a substantially reduced density of fuzz (Machado et al. 2009; Walford et al. 2012; Qin et al. 2022).

Notably, among these TF genes, GhMYB25-like has been identified as the central gene in fiber cell differentiation, as revealed by single-cell RNA sequencing (scRNA-seq) (Qin et al. 2022). This is further supported by the development of transgenic lines with RNAi-silenced GhMYB25-like, which produce seeds with only a few strands of lint and no fuzz, as well as fully fiberless mutants generated through CRISPR-mediated knockout targeting both copies of GhMYB25-like in the At and Dt subgenomes (Walford et al. 2011; Qin et al. 2022). While the functions of these TFs have been well documented, robust experimental evidence is needed to confirm the regulatory relationships between GhMYB25-like and other TFs.

Cloning of fiber mutant genes has proven to be effective in identifying key specific alleles that control fiber initiation. However, the fine mapping of cotton fiber mutants remains challenging due to the functional redundancy and dosage effects of genes in allotetraploid cotton. Currently, only three loci conferring fiber initiation have been successfully cloned in tetraploid cotton: the dominant fuzzless locus N₁ (Wan et al. 2016), the recessive fuzzless locus n₂ (Zhu et al. 2018), and the recessive lintless locus li₃ (Wu et al. 2018; Chen et al. 2020). Furthermore, the complex genetic variations observed in natural fiberless mutants further complicate the accurate detection of their genotypes. An example of this complexity is the fiberless mutant Xu142 fl (Xuzhou142 fuzzless-lintless mutant). To date, four different genotype models have been proposed to explain the fiberless phenotype of Xu142 fl: the n₂n₂li₃li₃ model (Zhang and Pan 1991), the n₂n₂li₃li₃n₃n₃ model (Turley and Kloth 2008), the n₁n₁n₂n₂li₃li₃li₄li₄ model (Du et al. 2001) and the n₂^Xun₂^Xu n₃n₃n₅n₅ model (Chen et al. 2020). Although preliminary results have been obtained in the cloning of the loci controlling fiber initiation, the mechanistic details behind these loci and their role in fiber initiation, as well as the molecular mechanism underlying lint and fuzz initiation, remain inadequately documented.

Therefore, this study aims to reveal the molecular mechanism of natural fiber mutant loci controlling lint and fuzz fiber initiation. We found that a key mutation, GhMYB25-like_At^hapT, simultaneously controlled lint and fuzz initiation. GhMYB25-like_At^hapT is a dominant negative mutation and the encoded protein inhibits the function of the normal protein GhMYB25-like, resulting in down-regulation of the expression of a series of genes, including key fiber-initiation-related genes (GhPDF2, GhHD-1), cell cycle regulatory genes such as the cyclinD3 gene CYCD3; 1, and phospholipid signaling genes such as the Phospholipase D gene (PLD). Moreover, our findings indicate that lint and fuzz initiation is influenced by both the dominant negative effect of GhMYB25-like_At^hapT and the dosage effect of GhMYB25-like. By elucidating the molecular mechanism behind this dominant negative mutation, GhMYB25-like_At^hapT, our study sheds light on its control over lint and fuzz initiation and provides insights into cell fate determination.

Results

Genetic analysis reveals that the lint and fuzz traits are determined by a common locus

This study aimed to investigate the genetic basis of the lint and fuzz traits in the LM (linted—fuzzless, seeds with only lint attached) selfing population. The LM line was derived from a cross between Xu142 and Xu142 fl, and its lint and fuzz phenotypes were still segregating even after ten generations of self-crossing (Hu et al. 2018). To reveal the genetic mechanisms behind the initiation of lint and fuzz in the LM line, we grew and evaluated the LM selfing population in two different geographical regions, Wuhan and Sanya. Three distinct phenotypes of mature fiber were observed: LF (linted—fuzzy, seeds with both lint and fuzz attached), LM, and LL (less linted—fuzzless, seeds with a few strands of lint mainly on the chalazal end and no fuzz) (Fig. 1, A to I). Further analysis showed that LF plants had slightly higher lint percentages compared to LM plants without fuzz while LL plants had significantly lower lint percentages than both LF and LM plants (Supplementary Fig. S1). Additionally, the LM selfing population of Sanya had a higher LP than that in Wuhan, possibly due to the more stable climate in Sanya (Supplementary Fig. S1).

Figure 1.

Characterization of cotton seed fibers in the LM selfing population. A to C) Phenotype of mature bolls in LF, LM, and LL (from left to right). Bar = 1 cm. D to F) Phenotype of mature fiber in LF, LM, and LL (from left to right). Bar = 1 cm. G to I) Phenotype of fuzz in LF, LM, LL (from left to right). Bar = 1 cm. J to L) SEM photographs showing the ovules of LF, LM, and LL (from left to right) at 0 DPA. Bar = 50 μm. M to O) Paraffin sections of ovules of LF, LM, and LL (from left to right) at 4 DPA. Arrows in LF indicate fuzz cell. Bar = 50 μm. P) Frequency distribution graph of the LM selfing population for LP per plant in Wuhan. Q) Number of fuzzless and fuzzy plants of the LM selfing population in Wuhan.

Microscopic examination of the ovules confirmed consistent phenotypic details in both initiating fiber cells and mature fiber. Scanning electron microscopy (SEM) analysis of 0-day-postanthesis (DPA) ovules revealed normal protrusion of fiber cells on the surfaces of LF and LM ovules, whereas only a few fiber cells protruded on the chalazal end of LL ovules (Fig. 1, J to L). Similarly, paraffin sections of 4-DPA ovules showed the presence of normal lint and fuzz on LF ovules, lint without fuzz on LM ovules, and a few shorter lint strands on the chalazal end of LL ovules (Fig. 1, M to O).

The genetic characteristics of lint and fuzz traits were analyzed. In the LM selfing population in Wuhan, we observed a distribution of 86 LF plants, 180 LM plants, and 86 LL plants, which fit a segregation ratio of 1:2:1 (χ²_1:2:1 = 0.1818, Supplementary Table S1). Similarly, in the LM selfing population in Sanya, we observed 104 LF plants, 214 LM plants, and 98 LL plants, also conforming to a ratio of 1:2:1 (χ² = 0.5192, Supplementary Table S1). Furthermore, we examined the phenotypes of 379 LF plants and 380 LL plants derived from the LM selfing population in the next generation and these phenotypes did not exhibit any segregation. Moreover, all F₁ plants resulting from a cross between LF and LL plants produced lint-fuzzless seeds, resembling the LM phenotype. Analysis of the F₂ population derived from this cross revealed 359 LF plants, 659 LM plants, and 346 LL plants, mirroring a segregation ratio of 1:2:1 (χ² = 1.7991, Supplementary Table S1). These findings provide compelling evidence that the lint and fuzz traits in the LM line are controlled by the same locus.

The same locus decides the recessive trait for lint and the dominant trait for fuzz

To evaluate the genetic effects of mutation on lint and fuzz initiation, a thorough investigation was conducted on the genetic basis of lint and fuzz in the LM selfing population. In the LM selfing population in Wuhan, out of the 352 plants analyzed, 86 plants were classified as less-linted (with lint percentages lower than 12%) and 266 plants as linted (with lint percentages ranging from 21% to 34%). This segregation pattern fit a ratio of 1:3 (χ²_1:3 = 0.1136, Fig. 1P). Similarly, in the LM selfing population in Sanya, out of the 416 plants analyzed, 98 were less-linted (with lint percentages lower than 27%) and 318 were linted (with lint percentages ranging from 28% to 40%). This segregation pattern also aligned with a ratio of 1:3 (χ²_1:3 = 0.4806, Supplementary Fig. S2A). These inheritance analyses strongly indicate that the less-linted trait operates as a recessive trait. Regarding the fuzz trait, two distinct types were observed in the LM selfing population: fuzzy and fuzzless. In the LM selfing population in Wuhan, out of the 352 plants analyzed, 266 plants were fuzzless and 86 plants were fuzzy. This segregation pattern fit a ratio of 3:1 (χ²_3:1 = 0.0606, Fig. 1Q). Similarly, in the LM selfing population in Sanya, out of the 416 plants analyzed, 312 plants were fuzzless and 104 plants were fuzzy. This segregation pattern also matched a ratio of 3:1 (χ²_3:1 = 0, Supplementary Fig. S2B). These findings strongly suggest that the fuzzless trait operates as a dominant trait. Overall, it can be concluded that the locus controlling lint and fuzz initiation simultaneously exerts different effects on lint and fuzz.

A single amino acid change in the DNA-binding domain of GhMYB25-like_At causes the less-linted and fuzzless trait in LL mutant

To clone the gene responsible for the fiber phenotype, Bulked Segregant Analysis (BSA) was conducted using DNA pools from LF plants and LL plants in the LM selfing population. Calculating the Euclidean distance (ED) of single nucleotide polymorphisms (SNPs) and insertion–deletion mutations (Indels) between the LF and LL pools revealed a noticeable bias near the end of chromosome Ghir_A12 (Fig. 2, A and B; Supplementary Fig. S3). Consequently, the candidate region (Ghir_A12: 91.84—94.08 Mb) spanning 2.24 Mb was identified, containing a total of 117 genes (Supplementary Table S2).

Figure 2.

Fine-mapping of the mutant locus in LL plants. A) ED association analysis of SNPs between the LF pool and LL pool. The ED graph shows chromosome A12. Each spot on this figure represents the ED⁵ value of a SNP, the solid line represents the smoothed ED⁵ value, and the dashed line represents the significance threshold. B) ED association analysis of Indels between the LF pool and LL pool. The ED graph shows chromosome A12. Each spot represents the ED⁵ value of an Indel, the solid line represents the smoothed ED⁵ value, the dashed line represents the significance threshold. C) Genotypes and phenotypes of recombinants in LF × LL F₂ population. Dash lines indicate genotyping markers. D) Gene structure of GhMYB25-like_At in LF and LL. E) The predicted protein structure of GhMYB25-like_At and GhMYB25-like_At^hapT on Phyre2 website. F) The close-up view of typical MYB TFs (C-Myb, TvMyb2, TvMyb3, AtWER) interacted with DNA. G) Schematic presentation of GhMYB25-like_At and its truncated proteins, GhMYB25-like_At^CR1 and GhMYB25-like_At^CR2. DBD: DNA binding domain. H) Phenotypic analysis of GhMYB25-like_At CRISPR/Cas9-engineered mutations in cotton. Bar = 1 cm.

To fine-map the mutations of the LL plants, an F₂ population consisting of 1,364 individuals was constructed by crossing LF plants and LL plants (Supplementary Table S1). Due to the highly homogenous genetic background of LF and LL, only 4 KSAP markers were developed for genotyping within the candidate region. Genotyping analysis of the F₂ population revealed that the genotype of the KSAP marker K1745 coseparated with fiber phenotype in all 1,364 individuals (Fig. 2C). Considering the recombination events on both sides of marker K1745, a 114-kb interval between the markers R01 and R02 was identified.

This interval contains two genes, Ghir_A12G017450 (GhMYB25-like_At) and Ghir_A12G017460 (GhMML4_At), which had been previously implicated in fiber initiation (Fig. 2C; Walford et al. 2011; Wu et al. 2018). Notably, expression analysis showed that GhMYB25-like_At was highly expressed in the outer integument of ovules at 0 DPA in LF, LM, and LL plants, with a decrease at 5 DPA (Supplementary Fig. S4A). In contrast, GhMML4_At showed almost no expression (Supplementary Fig. S4B). Comparing the full-length genomic sequence and the 3-kb upstream/downstream sequence of GhMYB25-like_At between LF and LL plants, we only discovered a nonsynonymous mutation (SNP A314T) that resulted in an amino acid change from lysine to methionine (K105M) (Fig. 2D; Supplementary Fig. S5). Therefore, it is likely that the SNP A314T in GhMYB25-like_At is the causative mutation for the reduced lint and fuzz traits observed in LL plants.

To elucidate the impact of GhMYB25-like_At^hapT, we utilized the Phyre2 website to predict the protein structures of GhMYB25-like_At and GhMYB25-like_At^hapT (Fig. 2E). The substitution of residue K105 with M105 in GhMYB25-like_At^hapT occurred on the third α-helix of R3 motif and did not disrupt the secondary structure of α-helix (Fig. 2E). The residue K105 of GhMYB25-like_At was conserved among plant R2R3 MYB TFs and MYB TFs with known structures, such as c-Myb (Tahirov et al. 2002), TvMYB2 (Jiang et al. 2011), TvMYB3 (Wei et al. 2012), and WEREWOLF (AtWER) (Wang et al. 2020) (Supplementary Fig. S6, A and B). Moreover, the lysine residues (K105) were directly involved in hydrogen bonding with guanine (G), further suggesting the critical role of this lysine residue in DNA recognition across different species (Fig. 2F). Therefore, it can be concluded that the mutation GhMYB25-like_At^hapT is the causative mutation for the lint and fuzz traits in the LL mutant.

Dominant negative effect of GhMYB25-like_At^hapT determines lint and fuzz phenotype

To confirm that GhMYB25-like_At^hapT is the causative mutation, transgenic cotton lines were developed. Due to limitations in selecting the protospacer-adjacent motif (PAM) domain, it was impossible to create CRISPR/Cas9 transgenic cotton that fully mimics the GhMYB25-like_At^hapT mutation through a single base editor. Additionally, the high identity (98.1%) of DNA sequences between GhMYB25-like_At and GhMYB25-like_Dt made it difficult to specifically target the subgenome of tetraploid cotton. Therefore, we devised a different strategy to create a series of mutants. In a previous study, we successfully obtained GhMYB25-like_At/Dt CRISPR lines (a₁a₁a₂a₂), called MYB25-like_CR#1, which produced fiberless seeds (Supplementary Fig. S7, A and B; Qin et al. 2022). In this study, using CRISPR/Cas9 technology, we engineered a mutation with a different target in GhMYB25-like_At/Dt (a₁a₁a₂a₂) named MYB25-like_CR#3, resulting in seeds without both lint and fuzz (Supplementary Fig. S7, C and D). Moreover, to obtain the mutants only targeting GhMYB25-like_At, we conducted crosses between the wild-type Jin668 and Cas9-free MYB25-like_CR#1 and CR#3 mutants. Through screening the progenies of these two crosses, two potential GhMYB25-like_At mutants, CR1 and CR2, were obtained.

The GhMYB25-like_At_CR1 mutant displayed a 1-bp deletion in the first exon of GhMYB25-like_At, causing a frame-shift mutation at the 9th amino acid and premature termination of translation at the 11th amino acid (Fig. 2G; Supplementary Fig. S8, A and B). This GhMYB25-like_At_CR1 mutant (a₁^CR1a₁^CR1A₂A₂) led to almost complete knockout of the endogenous GhMYB25-like_At gene. Interestingly, despite a slight reduction in fuzz, the GhMYB25-like_At_CR1 mutant showed no substantial alteration in fiber phenotype compared to the wild-type (Fig. 2H). This indicated that two copies of GhMYB25-like could result in sufficient protein to initiate fiber growth, albeit insufficient for full fuzz initiation. Additionally, the GhMYB25-like_At_CR2 mutant had a 1-bp deletion in the third exon of GhMYB25-like_At, resulting in a frame-shift mutation at the 84th amino acid and early termination of translation at the 111th amino acid (Fig. 2G; Supplementary Fig. S8, A and B). Therefore, the GhMYB25-like_At_CR2 mutant represented a loss-of-function allele with a non-full-length protein, closely mimicking the natural mutation GhMYB25-like_At^hapT. However, the GhMYB25-like_At_CR2 mutant (a₁^CR2a₁^CR2A₂A₂) caused a slight reduction in the number of lint fibers and almost complete absence of fuzz compared to wild-type (Fig. 2H). The loss-of-function allele GhMYB25-like_At^CR2 brought about more pronounced changes in the phenotype than the complete knockout mutation GhMYB25-like_At^CR1, suggesting that GhMYB25-like_At^CR2 acts as a dominant negative mutation. In other words, the GhMYB25-like_At^CR2 mutant protein not only lost its function but also interfered with the function of the wild-type GhMYB25-like_Dt protein. In comparison to GhMYB25-like_At^CR2, the loss-of-function mutation GhMYB25-like_At^hapT in the LL natural mutant (a₁^hapTa₁^hapTA₂A₂) led to a notable decrease in LP and the absence of fuzz (Fig. 1). The evident phenotypic alterations implied that the GhMYB25-like_At^hapT mutation may have a stronger effect as a dominant negative mutation. Thus, these findings show that the dysfunctional GhMYB25-like_At is like “the dog in the manger”, resulting in seeds with less lint and no fuzz.

The down-regulation of fiber initiation-related TFs inhibits the initiation of both lint and fuzz cells

To investigate the discrepancies in lint cells between LF and LL plants, single-cell RNA sequencing (scRNA-seq) was conducted on the ovule outer integument of LL plants at various developmental stages (−1.5, −1, −0.5, and 0 DPA). This scRNA-seq data was then integrated with LF line data from the same four stages (Qin et al. 2022). Protoplasts released from the enzymatically digested ovule cell walls were captured for sequencing using the 10× Genomics platform. Following quality control, a total of 2,825 high-quality cells from LL plants were obtained, comprising 734, 1214, 487, and 390 cells from ovules at −1.5 DPA, −1 DPA, −0.5 DPA, and 0 DPA, respectively. The combined LF and LL data were grouped into three clusters based on different developmental stages, as visualized using the t-distributed stochastic neighborhood embedding (t-SNE) method (Fig. 3A). These clusters were annotated using marker genes identified in a previous study (Supplementary Fig. S9; Qin et al. 2022).

Figure 3.

Downregulating expression of fiber-initiation-related genes causes the arrest of fiber cell initiation in the LL mutant. A) The t-SNE projection plot shows major clusters of individual cell transcriptomes between LF and LL at −1.5, −1, −0.5, and 0 DPA, respectively. B) Dynamic changes of ovule epidermis between LF and LL during fiber initiation observed by SEM. Bars = 50 μm. C) Reclustering and UMAP projection plot of fiber cells between LF and LL at −1, −0.5, and 0 DPA, respectively. D) The differentially expressed TFs identified from scRNA-seq dataset between LF and LL fiber clusters at −1 and −0.5 DPA. The size of spots represents the gene expressed percentage in fiber clusters, and the expression level was normalized. E) Expression heatmap of −0.5 DPA ovule integument RNA-seq (left) and scRNA-seq (right) for DEGs in fiber cells.

At −1.5 DPA, both LF and LL plants exhibited only two cell types: epidermal (Epi) cells and outer pigment layer (OPL) cells. From −1 to 0 DPA, in addition to Epi and OPL cells, the fiber (Fib) cluster was observed in both LF and LL plants (Fig. 3A). The t-SNE projection plots revealed that the percentage of Fib cluster cells remained relatively consistent during the developmental process in LF plants (9.77% at −1 DPA, 10.95% at −0.5 DPA, and 9.51% at 0 DPA, respectively), but gradually decreased in LL plants (13.18% at −1 DPA, 11.91% at −0.5 DPA, and 0.26% at 0 DPA, respectively) from −1 to 0 DPA (Fig. 3A). The gene expression profiles and Gene Ontology (GO) analysis of the three cell clusters (Epi, OPL and Fib) at the four stages also showed that only the Fib cluster exhibited progressive differences during ovule development between LF and LL (Supplementary Fig. S10). These findings were further supported by SEM observations of ovule morphology, which confirmed the disparities in fiber development between LF and LL plants (Fig. 3B).

Although the fiber cell cluster was not identified at −1.5 DPA, it is possible that fiber precursors exist and can provide insights into the differences in potential fiber cells. Considering that the development of fiber precursors at −1.5 DPA and early fiber cells at −1 DPA is a continuous process, we hypothesized that these two cell types might have similar expression profiles. Therefore, we reclustered the cells obtained from LF and LL plants at −1.5 DPA and −1 DPA together and redefined 3 clusters: Epi, OPL, and Fib (Supplementary Figs. S11 and S12A).

Importantly, we identified a new fiber cluster that represented potential fiber cells at −1.5 DPA and early fiber cells at −1 DPA (Supplementary Fig. S12A). Using the Uniform Manifold Approximation and Projection (UMAP) method, we visualized 3 subclusters within the new fiber cluster (Supplementary Fig. S12B). Marker genes for Epi, OPL, and Fib were used for subclusters annotation (Supplementary Fig. S13). The Epi markers were scarcely enriched in the subclusters while the OPL markers were mainly enriched in subcluster 0 and subcluster 1 (Supplementary Fig. S13, A and B). Fib markers, on the other hand, were mainly enriched in subcluster 1 and subcluster 2 (Supplementary Fig. S13C). Based on the annotation of marker genes, we defined subcluster 0 as the OPL cells located near the fiber cells, subcluster 1 as the precursor fiber cells, and subcluster 2 as the early fiber cells (Supplementary Fig. S12B). The GO analysis with marker genes of each cluster further confirmed the correctness of the definition of these three subclusters (Supplementary Fig. S12C). We then examined the distribution and proportion of precursor fiber cells in LF and LL plants (14.87% in LF and 11.58% in LL). However, no substantial difference between LF and LL was observed at −1.5 DPA (Supplementary Fig. S12B). Moreover, the high similarity in Pearson's correlation coefficient (PCC) (r = 0.9) between LF and LL fiber precursors also supported this finding. These results suggest that precursor fiber cells are not the key factor underlying the difference between LF and LL plants.

Our analysis indicated that there was no apparent distinction between LF and LL precursor fiber cells (−1.5 DPA) (Supplementary Fig. S12). Consequently, the period spanning from −1 to 0 DPA might be the critical stage at which fiber cell differentiation occurs between LF and LL plants. By reclustering the fiber clusters of LF and LL plants at −1 DPA, −0.5 DPA, and 0 DPA, we were able to visualize them using the UMAP method. The disparity in the quantities and distributions of LF and LL fiber cell subclusters progressively increased with development (Fig. 3C). These scRNA-seq results imply that fiber cells differentiate in both LF and LL plants, however, the subsequent progression of fiber cells was hindered in LL plants.

To explore the underlying molecular causes behind the stasis of fiber cells in LL plants, we identified 367 differentially expressed genes (DEGs) in the fiber clusters between LF and LL plants at −1 DPA and −0.5 DPA (Supplementary Data Set 1). Among these DEGs, we focused on genes encoding TFs, as they play important roles in fiber cell initiation. Alongside the key fiber cell initiation TF genes (such as GhHD-1, GhPDF2, GhMYB25), other previously unreported TF genes (such as Ethylene Response Factor 105 (ERF105), WRKY40) were also found to be downregulated in LL fiber cells. Furthermore, genes encoding proteins involved in cell cycle regulation (CYCD3; 1), phospholipid signaling (PLDα1, PLDα2, LTPG1), and the cell wall (EXPA4) were enriched in these DEGs (Supplementary Data Set 1). We also verified the expression of the DEGs in the fiber cells of outer integument of ovules at −0.5 DPA using RNA-seq. In these data, it was observed that the expression of these DEGs, particularly GhHD-1, GhPDF2, and GhMYB25, slightly decreased in the GhMYB25-like_At complete knockout line GhMYB25-like_At_CR1, and substantially decreased in the dominant negative mutant line GhMYB25-like_At_CR2. The changes in gene expression patterns of these DEGs in the GhMYB25-like_At knockout lines were consistent with the fiber phenotype of these lines. Analysis of LF and LL 5-DPA ovule outer integument RNA-seq data revealed that lint initiation-related TF genes, including GhHD-1, GhPDF2, and GhMYB25 were also downregulated during fuzz initiation (Supplementary Fig. S14). These results suggest that GhMYB25-like serves as a central protein in the regulatory network of fiber initiation, controlling both lint and fuzz initiation through the regulation of the expression of various initiation-related genes.

Dominant negative mutation GhMYB25-like_At^hapT/CR2 inhibits the transcriptional activation activity of wild-type protein GhMYB25-like

To illustrate the underlying cause of the dominant negative effects of GhMYB25-like_At^hapT/CR2, protein characterization was conducted. Using the LF and LL single-cell atlas, differentially expressed TF genes, namely GhHD-1 and GhPDF2 were identified as potential downstream genes of GhMYB25-like. Therefore, we examined the transcriptional activation activity of GhMYB25-like on GhHD-1 and GhPDF2. In yeast one-hybrid (Y1H) assays, yeast strains containing both the bait (proGhHD-1/proGhPDF2) and prey (GhMYB25-like) successfully activated resistance to AbA on SD-leu-ura medium (Fig. 4A; Supplementary Fig. S15A). Conversely, the yeast strain harboring an empty pGADT7 as prey failed to induce resistance to AbA under the same conditions (Fig. 4A; Supplementary Fig. S15A). Additionally, luciferase reporter system (LUC) experiments showed that cotransfection of the reporter (proGhHD-1/proGhPDF2) and effector (GhMYB25-like) in Nicotiana benthamiana leaves resulted in stronger luminescence compared to the negative control (Fig. 4B; Supplementary Fig. S15B). These findings strongly suggest that GhMYB25-like directly interacts with the promoters of both GhHD-1 and GhPDF2.

Figure 4.

GhMYB25-like^hapT interacts with GhMYB25-like and inhibits its function. A) Y1H assay of GhMYB25-like binding to the promoter of GhPDF2. The Y1H assay used the promoter fragment of GhPDF2 as bait and GhMYB25-like as the prey, with the negative control transformed with empty pGAD7 vector. Growth of Y1HGold[pGADT7/proGhPDF2-AbAi] and Y1HGold[GhMYB25-like_At-pGADT7/proGhPDF2-AbAi] was examined on the SD/-Leu/Ura media or SD/-Leu/Ura media supplemented with AbA. B) LUC assay showing the activation activity of GhPDF2 expression by wild-type GhMYB25-like protein. Nicotiana benthamiana leaves were injected and luciferin signals were captured using a whole-body fluorescent imaging system. Bar = 2 cm. C) EMSA of GhMYB25-like/GhMYB25-like_At^hapT/GhMYB25-like_At^CR2 binding to the promoter of GhPDF2. Biotin-labeled probes were incubated with GhMYB25-like, GhMYB25-like_At^hapT, and GhMYB25-like_At^CR2 protein. Unlabeled probes were used as competitors to verify binding specificity. The DNA concentration was fixed at 0.2 nM and the protein concentration was increased from 0 to 0.3 μM. The gray level was analyzed by ImageJ. D) LCI assay verifying the interaction of GhMYB25-like-GhMYB25-like, GhMYB25-like-GhMYB25-like_At^hapT and GhMYB25-like-GhMYB25-like_At^CR2. Bar = 2 cm. E) BiFC assay verifying the interaction of GhMYB25-like-GhMYB25-like, GhMYB25-like-GhMYB25-like_At^hapT and GhMYB25-like-GhMYB25-like_At^CR2. Bar = 50 μm. F) Effects of GhMYB25-like_At^hapT and GhMYB25-like_At^CR2 on the activation activity of GhPDF2 promoter by GhMYB25-like. Error bars represent mean SD of three biological replicates. Adjusted P values were calculated by 1-way analysis of variance with Dunnett's multiple comparison test. Adjusted P < 0.05 (*); Adjusted P < 0.01 (**), Supplementary Data Set 4. G) EMSA of GhMYB25-like, GhMYB25-like_At^hapT, GhMYB25-like_At^CR2, and their substrate DNA. The DNA concentration was fixed at 0.2 nM and the protein concentration was increased from 0 to 0.3 μM. The gray level was analyzed by ImageJ. H) A model assumed the protein amount of GhMYB25-like in LF, LM, and LL. Each white square represents the protein amount translated by 1 copy of GhMYB25-like, each diagonal square represents the protein amount translated by 1 copy of GhMYB25-like^hapT, each grey square represents the protein amount of GhMYB25-like inhibited by GhMYB25-like^hapT, and the dotted line represents threshold value of fiber cell initiation.

Electrophoretic mobility shift assays (EMSA) were used to compare the affinity of GhMYB25-like, GhMYB25-like_At^hapT, and GhMYB25-like_At^CR2 to the promoters of downstream genes. Results revealed that GhMYB25-like exhibited specific binding to the promoters of GhHD-1 and GhPDF2 (Fig. 4C; Supplementary Fig. S15C, lines 1 to 5). In contrast, the binding nearly disappeared when incubated with GhMYB25-like_At^hapT and GhMYB25-like_At^CR2 protein (Fig. 4C; Supplementary Fig. S15C, lines 6 to 9). These findings indicate that GhMYB25-like_At^hapT and GhMYB25-like_At^CR2 experience a partial loss of their transcriptional activation activity.

To elucidate the dominant negative effect of the mutant proteins GhMYB25-like_At^hapT/CR2, their interaction with the wild-type GhMYB25-like protein was examined. Yeast 2-hybrid (Y2H) assays could not be conducted due to the self-activation of full-length GhMYB25-like protein. Therefore, the interaction was validated using a luciferase complementation imaging (LCI) assay and bimolecular fluorescence complementation (BiFC) assay in N. benthamiana leaves. The LCI assay showed that cotransfection of GhMYB25-like/GhMYB25-like_At^hapT/GhMYB25-like_At^CR2-nLUC and GhMYB25-like-cLUC into N. benthamiana leaves resulted in strong luminescence signals (Fig. 4D). Conversely, cotransfection of the empty vector (nLUC) with GhMYB25-like-cLUC did not produce luminescence signals (Fig. 4D). Furthermore, in the BiFC assay, positive interactions were observed between GhMYB25-like and GhMYB25-like, between GhMYB25-like_At^hapT and GhMYB25-like, and between GhMYB25-like_At^CR2 and GhMYB25-like in both the nucleus and cytoplasm (Fig. 4E). These findings suggest that the normal GhMYB25-like protein interacts with GhMYB25-like_At^hapT and GhMYB25-like_At^CR2.

We further investigated the potential dominant negative effect of GhMYB25-like_At^hapT/CR2 on the transcriptional activation activity of GhMYB25-like. In the dual-luciferase reporter assay, cotransfection of the effector GhMYB25-like and the reporter containing GhHD-1 or GhPDF2 promoter in N. benthamiana leaves resulted in clear activation of luciferase reporter activity (Fig. 4F; Supplementary Fig. S15D). However, when GhMYB25-like was co-expressed with GhMYB25-like_At^hapT or GhMYB25-like_At^CR2, the luciferase activity significantly decreased (Fig. 4F; Supplementary Fig. S15D), indicating that GhMYB25-like_At^hapT or GhMYB25-like_At^CR2 suppressed the transcriptional activation activity of GhMYB25-like. Additionally, EMSA revealed that GhMYB25-like formed a stable complex with the promoters of GhHD-1 and GhPDF2 (Fig. 4G; Supplementary Fig. S15E, lines 1 to 4). However, the GhMYB25-like-DNA complex was gradually disrupted with increasing amounts of GhMYB25-like_At^hapT (Fig. 4G; Supplementary Fig. S15E, line 5 to 7) or GhMYB25-like_At^CR2 protein (Fig. 4G; Supplementary Fig. S15E, line 8 to 10). These results further demonstrate that GhMYB25-like_At^hapT and GhMYB25-like_At^CR2 exert dominant negative effects by suppressing the transcriptional activation activity of GhMYB25-like.

Combined effects of dominant negative effect and dosage effect determine the recessive trait for lint and the dominant trait for fuzz

After observing that the dominant negative effect of GhMYB25-like_At^hapT was achieved through its binding to the normal GhMYB25-like protein, thereby inhibiting its function, the reduction of this effect was expected with a sufficient amount of wild-type protein. The expression of GhMYB25-like_At/Dt during lint initiation was at least 2-fold higher compared to that during fuzz initiation, with these findings being widely reported in numerous studies (Walford et al. 2011; Zhu et al. 2018; Naoumkina et al. 2021). As protein and mRNA levels are generally positively correlated, it can be inferred that the protein content of GhMYB25-like at lint initiation should be higher than that during fuzz initiation. Based on this, we drew a model to describe the protein amount of GhMYB25-like, GhMYB25-like_At^hapT and inhibited GhMYB25-like in LF, LM, and LL during lint and fuzz initiation (Fig. 4H). In this model, it is assumed that each copy of GhMYB25-like translates the same amount of protein, and the amount of GhMYB25-like during lint initiation is twice that during fuzz initiation. GhMYB25-like_At^hapT inhibited the function of GhMYB25-like, weakening its function. Both lint and fuzz initiation require the same amount of functional GhMYB25-like. In LM, the amount of functional GhMYB25-like protein is sufficient to drive lint initiation but insufficient to drive fuzz initiation (Fig. 4H). This suggests that the dominant negative effect of GhMYB25-like_At^hapT is mitigated by the abundance of GhMYB25-like protein during lint initiation, thereby functioning as a recessive allele in controlling the less-lint phenotype. Conversely, since the content of GhMYB25-like protein is very low during fuzz initiation, GhMYB25-like_At^hapT operates as a dominant allele for fuzzless.

GhMYB25-like_At, not GhMYB25-like_At^hapT was selected for in G. hirsutum

To gain a better understanding of the functional differences between GhMYB25-like_At and GhMYB25-like_At^hapT in the core germplasm panel of G. hirsutum, a candidate gene association analysis was conducted to investigate the relationship between variations in GhMYB25-like_At and LP. A total of 15 SNPs located within the 2-kb promoter and gene body of GhMYB25-like_At were detected in 417 samples (Ma et al. 2018; Li et al. 2021). Among these SNPs, the A314T variant was found to have a weak linkage disequilibrium (LD) with other mutations (Supplementary Fig. S16). Notably, only the A314T SNP exhibited a significant association with LP (Fig. 5, A and B). Based on the A314T SNP, the genome-wide association studies panel was categorized into two distinct haplotypes: AA and TT (Fig. 5C). It was observed that accessions with the TT haplotype (n = 19) exhibited a lower LP compared to accessions with the AA haplotype (n = 398) (Fig. 5C). Taken together, these findings indicate that GhMYB25-like_At^hapT functions as a negative regulator of LP in the G. hirsutum population.

Figure 5.

The LP negative regulator, GhMYB25-like_At^hapT, was not selected during G. hirsutum cultivar breeding. A) Target gene-associated analysis with LP by Fast-LMM in 417 G. hirsutum core germplasm panel. The red dashed line represents the significance threshold which was determined using Bonferroni correction (P < 3.3 × 10⁻³). B) Target gene associated analysis with LP by seqMeta in 417 G. hirsutum core germplasm panel. The red dashed line represents the significance threshold which was determined using Bonferroni correction (P < 3.3 × 10⁻³). C) Lint percentage of GhMYB25-like_At and GhMYB25-like_At^{hap T} plants in 417 G. hirsutum core germplasm panel. Center line, median; box limits, upper and lower quartiles; whiskers, 1.5 × interquartile range; points, outliers. D) Allele frequencies of the GhMYB25-like_At SNP A314T mutation in 1547 G. hirsutum accessions, Supplementary Data Set 2. E) Nucleotide diversity (π) of GhMYB25-like_At and its flanking regions in three G. hirsutum populations: Ghlandrace, GhImpUSO and GhImpCHN. F) EHH across the GhMYB25-like_At genomic region for 2 G. hirsutum subgroups: GhImpUSO and GhImpCHN. Lines represent the EHH values of accessions containing GhMYB25-like_At and GhMYB25-like_At^hapT.

Given the functional divergence of GhMYB25-like_At proteins, the genetic divergence of GhMYB25-like_At alleles was investigated in G. hirsutum (Li et al. 2021). The frequency of the A314T SNP was analyzed to determine the distribution of the GhMYB25-like_At allele in G. hirsutum subspecies. The GhImpUSO group, consisting of improved cultivars from the USA and other countries, exhibited the highest proportion of GhMYB25-like_At^hapT (6.02%), followed by the GhImpCHN group, consisting of improved cultivars from China, with a frequency of 4.20% (Fig. 5D). In contrast, the Ghlandrace group, comprising of landraces, showed the lowest frequencies of GhMYB25-like_At^hapT (0.78%) (Fig. 5D; Supplementary Data Set 2). Overall, the allele frequency of GhMYB25-like_At^hapT was relatively low in G. hirsutum accessions. Nucleotide diversity analysis revealed that the promoter and coding region of GhMYB25-like_At exhibited a relatively lower π value compared to its flanking regions in different subgroups, indicating that GhMYB25-like_At experienced stronger selection pressure than its flanking regions (Fig. 5E). Additionally, an Extended Haplotype Homozygosity (EHH) analysis was conducted in the GhImpUSO and GhImpCHN groups using the A314T SNP as the core haplotype. In both groups, the EHH of GhMYB25-like_At^hapT showed rapid decay, while the EHH of GhMYB25-like_At decayed more slowly (Fig. 5F). The differential decay rates suggested that GhMYB25-like_At, but not GhMYB25-like_At^hapT was selected during G. hirsutum domestication. This divergence in selection intensity may be attributed to the fact that GhMYB25-like_At^hapT functions as a negative regulator of LP.

Discussion

The initiation of lint and fuzz, which play a crucial role in cotton fiber yield, has long been a perplexing topic in studies of inheritance. In this study, we employed fine-mapping and gene editing techniques to identify a naturally occurring dominant negative gene, known as GhMYB25-like_At^hapT, that exerts controls over both lint and fuzz initiation (Fig. 2). We found that the dominant negative mutant protein GhMYB25-like_At^hapT inhibited the normal functioning of the GhMYB25-like protein, thereby suppressing the activation of downstream genes (Fig. 4). Furthermore, the performance of the lint/fuzz trait in the LL mutant was dependent on both the dosage of GhMYB25-like and the dominant negative effect of GhMYB25-like_At^hapT. This elucidates the molecular mechanism underlying the dominant negative effect of GhMYB25-like_At^hapT. Our results provide valuable insights into the genetic and molecular foundations of lint and fuzz initiation, providing a rational basis for cotton fiber quality breeding.

GhMYB25-like acts as a hub in a regulatory network that controls both lint and fuzz initiation

The initiation of lint and fuzz has long been recognized as an important biological problem. According to our research and relevant literature, lint fiber initiation occurs from −1 DPA, while fuzz initiation arises at 4 to 10 DPA (Fig. 3A; Stewart 1975; Wang et al. 2021; Qin et al. 2022). Due to the temporal asynchrony of the protuberance of lint and fuzz fibers, the initiation of lint and fuzz is commonly believed to be two distinct processes. Moreover, studies on fiber genetics also distinguish the loci controlling fiber initiation into two categories: lintless loci and fuzzless loci. Therefore, the initiation of lint and fuzz has long been attributed to distinct genes or regulatory networks. Although some researchers have realized the lint and fuzz developments are controlled by some common loci or overlapped regulatory pathways (Turley and Kloth 2008), no explicit evidence has been provided.

In this study, we identified an allele of GhMYB25-like, GhMYB25-like^hapT, encoding a protein that controls the initiation of both lint and fuzz fiber simultaneously (Fig. 2). Additionally, the majority of the loci controlling fiber initiation (such as the dominant fuzzless locus N₁, the recessive fuzzless locus n₂ and the recessive lintless locus li₃) were all different GhMYB25-like alleles (Wan et al. 2016; Zhu et al. 2018; Chen et al. 2020). These results indicate that GhMYB25-like could indeed simultaneously control the initiation of both lint and fuzz fibers.

GhMYB25-like was identified as a core node of the fiber initiation regulatory network through scRNA-seq analysis, regulating the expression of key fiber initiation-related genes (such as GhPDF2, GhHD-1, and GhMYB25) (Fig. 3, D and E). Moreover, overexpression of GhHD-1 increased quantities of lint fiber and fuzz fiber compared to wild type quantities (Walford et al. 2012). Knocking out the expression of GhPDF2 resulted in fewer lint fiber initials during lint initiation and decreased fuzz fiber density of mature fibers (Qin et al. 2022). This indicated genes downstream of GhMYB25-like also play important roles in both lint and fuzz initiation. Therefore, we propose that both lint and fuzz initiation are controlled by the same regulatory network, with GhMYB25-like playing a central role.

A working model for GhMYB25-like_At^hapT allele-controlled fiber initiation

By analyzing the scRNA-seq data of LF and LL outer integument of ovules from −1.5 DPA to 0 DPA, we were able to identify notable differences in the development of fiber cells caused by the presence of GhMYB25-like and its mutant form GhMYB25-like_At^hapT in this study (Fig. 3). By reclustering all fiber cells of LF and LL, we found there was no difference in the number and distribution of fiber cell subclusters between LF and LL at −1 DPA (Fig. 3C). However, the number of fiber cell subclusters decreased in LL at −0.5 DPA, and only one fiber cell was identified in LL at 0 DPA (Fig. 3C), suggesting the initiation of fiber cells in LL was arrested compared to that in LF. In conclusion, we infer that GhMYB25-like does not participate in fiber cell differentiation but instead maintains the differentiation state of fiber cells.

The fiber cell initiation process involves 4 sequential identity transformations: precursor fiber cells, early fiber cells, protruding fiber cells, and early elongating fiber cells, based on our previous scRNA-seq analysis of ovule outer integument (Qin et al. 2022). Due to the initiation of lint and fuzz being controlled by the same pathway, we deduce the initiation process of lint and fuzz cells should also be similar, except for a few days’ delay in the initiation of fuzz. Drawing on the findings of this study and previous research (Walford et al. 2011, 2012; Qin et al. 2022), we proposed a model that illustrates the distinct roles played by these two alleles in LF and LL, respectively (Fig. 6).

Figure 6.

A proposed model for LF and LL fiber cell initiation. The epidermal cells of the ovule differentiated into fiber cells after undergoing 4 identity changes: precursor fiber cells, early fiber cells, protruding fiber cells, and early elongating fiber cells. In LF, GhMYB25-like expression begins in early fiber cells and promotes cell protrusion by activating fiber initiation-regulated genes, such as GhHD-1 and GhPDF2, and subsequently lint elongates. A few days later, the signals promoting GhMYB25-like expression gradually weaken, causing the protruding fiber initials fail to elongate and turn into fuzz. However, in LL, a dominant negative mutation GhMYB25-like_At^hapT inhibits the function of the wild-type protein GhMYB25-like, resulting in weaker transcriptional activation to fiber initiation-related genes. As a result, most of lint cells and all fuzz cells in LL fail to maintain their differentiated state and lose their identity of fiber cells.

For lint initiation, both LF and LL feature undifferentiated precursor fiber cells at −1.5 DPA, despite LL having less lint on mature seeds. By −1 DPA, these precursor fiber cells in both LF and LL undergo transformation into early fiber cells (Fig. 3A; Supplementary Fig. S12B). Notably, the expression of GhMYB25-like_At/Dt is predominantly observed in both LF and LL fiber cells (Supplementary Fig. S17). The abundant transcripts of GhMYB25-like result in the activation of downstream genes associated with fiber initiation, such as GhHD-1, specifically in LF fiber cells. Conversely, in LL, the interaction between the dominant negative mutation GhMYB25-like_At^hapT and wild-type protein GhMYB25-like inhibits its transcriptional activation function (Supplementary Fig. S15, D and E). At −0.5 DPA, LF fiber cells gradually protrude, whereas in LL, the down-regulation of fiber-initiation-related genes, including GhPDF2 and GhHD-1, due to the dominant negative effect of GhMYB25-like_At^hapT results in the arrest of fiber cell development (Fig. 4, F and G; Supplementary Fig. S15, D and E). Finally, at 0 DPA, protruding fiber cells in LF initiate tip-biased diffuse growth, while only a few fiber cells protrude in LL.

By the time fuzz initiates, lint is experiencing rapid elongation. After differentiation, fuzz cells of LF protrude under the action of GhMYB25-like, but fail to elongate. In LL, the differentiated fiber cells are unable to protrude due to the dominant negative effect of GhMYB25-like^hapT.

The complexes responsible for the differentiation of Arabidopsis trichomes and cotton fibers are not conserved

Despite the shared origin of cotton fiber and Arabidopsis trichomes from epidermal cells, our research has unveiled a substantial regulatory network centered around GhMYB25-like that coordinates the initiation of lint and fuzz cells in cotton (Fig. 3, D and E). In Arabidopsis, GLABRA1 and AtMYB23, the key MYB TFs in MYB-bHLH-WD40 complex for trichome initiation, belong to MYB subgroup 15 and exhibit analogous function (Kirik et al. 2001). However, GhMYB25-like, distinctively categorized under MYB subgroup 9 (Paterson et al. 2012), lacks functionally redundant genes within the cotton genome. Moreover, our results demonstrate that GhMYB25-like engages in self-interaction and activates the expression of downstream genes through multimer formation (Fig. 4). Furthermore, other investigations unveil GhMYB25-like's interaction with more than 10 other crucial TFs involved in fiber initiation, as evident by Y2H experiments. Comprehensive details of this intricate super complex governing fiber initiation will be presented in our forthcoming research. This finding sets it apart from the MYB-bHLH-WD40 complex involved in the development of Arabidopsis trichomes (Oppenheimer et al. 1991; Walker et al. 1999; Payne et al. 2000; Zhang et al. 2003). By revealing this regulatory network governing cell differentiation, our research provides a deeper understanding of the fundamental mechanisms underlying cell fate determination.

Creating fuzzless cotton varieties with a high lint percentage

In cotton breeding, the importance of fuzzless varieties has been acknowledged for their potential to reduce short fiber content and the occurrence of seed coat neps during ginning (Bechere et al. 2014). However, most fuzzless cotton varieties exhibit lower lint percentages as they carry a mutant allele of GhMYB25-like. For instance, the dominant fuzzless locus N₁ causes an approximate 60% decrease in LP, while the recessive fuzzless locus n₂ leads to an approximate 30% decrease in LP (Turley and Kloth 2002). As a result, these fuzzless loci are not directly applicable in fuzzless cotton breeding. Our findings further illustrate that breeders do not deliberately select for GhMYB25-like_At^hapT during domestication (Fig. 5F). Therefore, creating new alleles of GhMYB25-like is crucial for breeding fuzzless cotton with a high LP.

In this study, we made observations regarding the impact of two dominant negative mutations, GhMYB25-like_At^hapT and GhMYB25-like_At^CR2, on linted and fuzzless seeds in the LL mutant (Fig. 1, C, F, and I). The results showed that GhMYB25-like_At^hapT caused a decrease in the number of lint and fuzzless seeds, while GhMYB25-like_At^CR2 resulted in a slight decrease in lint numbers and a near-total disappearance of fuzz (Fig. 2H). The difference in phenotype between LL and the GhMYB25-like_At_CR2 mutant may be attributed to the different effects of the two dominant negative mutations, GhMYB25-like_At^hapT and GhMYB25-like_At^CR2, or the different genetic backgrounds of LF and Jin668. Hence, introducing dominant negative mutation types of GhMYB25-like or transforming the existing mutation of GhMYB25-like_At into cotton with diverse genetic backgrounds could contribute to the improvement of fuzzless breeding. An alternative approach to achieve this is to generate mutant lines with controlled expression of GhMYB25-like using CRISPR technology, specifically targeting the GhMYB25-like promoter. These approaches hold promising potential for breaking the linkage between lint and fuzz, facilitating the development of fuzzless cotton varieties with a high LP.

Materials and methods

Plant materials

The cotton (Gossypium hirsutum) LM line used in this study was selected from an recombination inbred line population, which was obtained by crossing the Xu142 and Xu142 fl varieties (Hu et al. 2018). Selfed seeds from the LM line were then planted, and the leaves of individual plants were collected for DNA extraction. An F₂ population derived from the LF and LL varieties was used for fine mapping. All plants involved in gene mapping were grown in two experimental fields: Huazhong Agricultural University (114°21′49″, 30°28′54″) in Wuhan, China, and the Cotton Research Institute (109°10′22″, 18°23′41″) of the Chinese Academy of Agricultural Sciences in Sanya, China. All transgenic lines were grown in a greenhouse under long-day conditions (14 h light/10 h dark) with a temperature maintained at 26 to 32 °C. Nicotiana benthamiana plants used for LUC assays were cultivated in a cultivation chamber under long-day conditions (16 h light/8 h dark) with a temperature of 25 °C and relative humidity of 40% to 60%.

Phenotype investigation

For each plant, 3 to 5 mature bolls from the middle and lower regions were collected. These mature bolls were then processed using a cotton gin to remove the lint. The LP was calculated using the formula: LP = (W1/W2) * 100%, where W1 represents the weight of the lint and W2 represents the total weight of the lint fiber and seeds. For the fuzz phenotype, visual inspection was conducted to differentiate between fuzzy seeds and fuzzless seeds. This method was used to analyze the phenotype of both the LM selfing population and the LF × LL F₂ population.

Mapping of mutations in LL

BSA was conducted to map the mutant locus responsible for the fiber phenotype in the LM selfing population. A total of 50 LF plants (with lint and fuzz) and 50 LL plants (with less lint and no fuzz) were selected as tissue samples for BSA. Genomic DNA was extracted from the young leaves of each LF and LL individual, and two DNA pools were created by combining an equal amount of DNA from the LF and LL groups, respectively.

Illumina sequencing and ED association analysis were performed by BioMarker Technologies Company. Firstly, library construction and sequencing were carried out following the protocol provided by Illumina. Raw sequencing data was filtered, and the clean data obtained was then mapped to the TM-1 reference genome HAU v1.0 (Wang et al. 2019a). SNPs and Indels were identified using the Genome Analysis Toolkit software on the mapped data. Low-quality SNPs and Indels were removed while the high-quality ones were used for ED association analysis between the LF pool and LL pool, respectively (Hill et al. 2013). Finally, the results of the ED analysis were utilized to generate plots across the entire cotton genome, highlighting the regions associated with the fiber phenotype.

Based on the association results acquired from BSA, the mutant gene responsible for the fiber phenotype was subsequently mapped in the LF × LL F₂ population. To facilitate the genotyping of individuals in the LF × LL F₂ population, Kompetitive Allele Specific Polymerase Chain Reaction (KASP) markers were developed. These KASP markers were designed using the SNPs and Indels located within the candidate region. The primer sequences for the KASP markers can be found in Supplementary Data Set 3.

To perform KASP genotyping, a Polymerase Chain Reaction (PCR) amplification reaction was prepared. The reaction mixture included 5 μL of DNA (50 to 250 ng), 5 μL of KASP PCR mix (Laboratory of the Government Chemist), and 0.14 μL of primer mix (consisting of a 36 μM forward allele-specific primer and a 90 μM typical reverse primer). The PCR amplification process consisted of the following steps: an initial denaturation at 94 °C for 15 min (1 cycle), followed by denaturation at 94 °C for 10 s, annealing at 61 °C for 60 s (−0.6 °C per cycle) for 10 cycles, and a final amplification at 94 °C for 20 s and 55 °C for 60 s for 26 cycles (with an appropriate increase of 3 to 6 cycles if genotyping clusters were not clearly distinguishable). Finally, the fluorescence intensity of the PCR products was measured using the Applied Biosystems 7500 Real-Time PCR System.

Vector construction and cotton transformation

CRISPR/Cas9 technology was utilized to generate GhMYB25-like mutants in cotton. A specific guide RNA (sgRNA) sequence (ACCTGTTTCCAAGAAGGGCG) targeting GhMYB25-like was designed. The tRNA-sgRNA fusion was amplified through PCR using the pGTR vector as a template and then cloned into the BsaI-digested pRGEB32-GhU6.9 expression vector (Wang et al. 2017). The primers for PCR can be found in Supplementary Data Set 3. The transformed pRGEB32-GhU6.9 expression vector was introduced into A. tumefaciens strain EHA105, and the positive strain was subsequently introduced into G. hirsutum “Jin668' (seeds with normal lint and fuzz) (Jin et al. 2006a, 2006b). The gene editing detection of T1 transgenic lines at the sgRNA target sites followed the workflow of high-throughput tracking of mutations (Hi-TOM) (Liu et al. 2019).

To generate the GhMYB25-like_At-CR mutants, the wild-type Jin668 was crossbred with 2 Cas9-free GhMYB25-like mutant lines (MYB25-like CR#1 and CR#3). Hi-TOM was used to detect the mutations at sgRNA target sites in the F₂ progenies of these 2 crosses. After screening, individuals that were specifically edited in GhMYB25-like_At were selected.

Morphological observation

SEM and paraffin sections were used to observe morphological changes during fiber initiation. For SEM, bolls from LF, LM, and LL at 0 DPA were collected to examine fiber initials on the surface of ovules. Additionally, bolls from LF and LL at −1.5, −1, −0.5, and 0 DPA were collected to observe the successive process of fiber cell protuberance. All bolls used for SEM were collected from the middle and lower part of the plants. The ovules were dissected from the bolls and fixed with 2.5% (v/v) glutaraldehyde. Following dehydration and drying, the ovules were examined using SEM (JSM-6390/LV, Jeol, Japan) according to established methods (Hu et al. 2018). For paraffin section analysis, bolls from the middle and lower parts of the plants were collected from LF, LM, and LL at 4 DPA. Ovules were fixed with a formaldehyde-ethanol-glacial acetic acid solution, 50% (v/v) for over 24 h. The detailed procedures for how dehydration, embedding, sectioning, and staining were carried out have been previously reported (Li et al. 2016).

scRNA-seq of cotton ovules

The protocol for preparing protoplasts, performing scRNA-seq, and data analysis of the LL line in this study followed our previously published reports (Qin et al. 2022). scRNA-seq data of the LF line were reused in this study.

For protoplast preparation, ovules were collected from the LL line at different developmental stages (−1.5, −1, −0.5, and 0 days postanthesis) and separated from cotton bolls. The ovules were then placed in culture dishes containing a digest solution composed of 1.5% (w/v) cellulase, 1% (w/v) hemicellulose, 0.75% (w/v) macerozyme, 0.4 M mannitol, 20 mM 2-morpholinoethanesulfonic acid (pH 5.7), 20 mM KCl, 10 mM CaCl₂, and 0.1% (w/v) BSA. The culture dishes underwent vacuum treatment (−1 atmosphere) for 5 min and were then rotated at 60 rpm at room temperature for 4 h to release the protoplasts. The protoplasts were filtered using a 40 μm cell strainer and rinsed with 1×phosphate buffered solution containing 0.04% BSA. Finally, a protoplasts suspension was obtained.

The protoplast suspension from the four samples was used for scRNA-seq following the manufacturer's instructions for the 10× Genomics platform (Zheng et al. 2017). The suspension was loaded onto Chromium microfluidic chips containing 3′ solution v2 chemistry. Each oil droplet contained a single-cell and a gel bead, forming a Gel Bead in Emulsion (GEM). Within each GEM, the RNA was reverse-transcribed into cDNA and tagged with a barcode. The GEMs were then disrupted, and the liquid was collected to construct a sequencing library using the Chromium Single Cell 30 v2 reagent kit. Finally, the library was sequenced on an Illumina Hiseq PE150 sequencer by Novogene Corporation.

The sequencing data from the LL line was integrated with that from the LF line, and the resulting integrated data was used for cell clustering identification. To obtain the cell clusters, the raw data was first filtered, followed by alignment to the TM-1 reference genome (Wang et al. 2019a) using Cell Ranger software. Furthermore, the Seurat package v4.0.1 in R v4.0.0 was loaded to filter out low-quality cells (such as doublets, no-load cells, and dead cells). Subsequently, the top 3,000 highly variable genes in high-quality cells were subjected to principal component analysis (PCA) to reduce data dimension. UMAP and t-SNE techniques were used to visualize the PCA results. By examining the UMAP or t-SNE visualizations, the cell clusters were identified and their identities were annotated using marker genes obtained from a previous study.

RNA-seq assay

The −0.5 DPA flower buds of LF, LM, LL, Jin668, GhMYB25-like_At_CR1, and GhMYB25-like_At_CR2 were collected for RNA isolation. One to two flower buds collected from one plant were used as a biological replicate. The ovules were removed from the flower buds, soaked in RNAlater (Sigma-Aldrich, R0901), and then the outer integuments were stripped from the ovules (Hu et al. 2018). The total RNA of outer integuments from 3 biological replicates was extracted using an RNAprep Pure Plant Plus Kit (TIANGEN, DP441). The cDNA library was sequenced on DNBSEQ sequencing platform by BGI-Tech. The detailed process for RNA-seq data analysis (clean reads mapping, low-quality mapping reads filtering, and gene expression levels calculating) is in our previous study (Qin et al. 2022).

Y1H assays and luciferase reporter system

The Matchmaker Gold Yeast 1-Hybrid System was utilized to assess the transcriptional activation activity of GhMYB25-like on its target genes (GhHD-1 and GhPDF). The construction of vectors involved cloning the promoter fragments of the downstream target genes into the pBait-AbAi vector, and the full-length coding sequence (CDS) sequences of GhMYB25-like into the pGADT7-AD vector. The specific primers used for cloning can be found in Supplementary Data Set 3. Subsequently, the pBait-AbAi vectors were transformed into the Y1HGold strain, serving as the bait yeast strain. The prey plasmid pGADT7, containing the cloned GhMYB25-like, was also transformed into the bait yeast strain to create the test group. As a negative control, the empty pGADT7-AD vector was transformed into the bait yeast strain. Finally, the negative control strain Y1HGold[pGADT7/pro-AbAi] and the test group strain Y1HGold[GhMYB25-like-pGADT7/pro-AbAi] were plated on SD/-Leu/Ura media supplemented with different concentrations of Aureobasidin A (AbA) to confirm positive interactions. All strains were cultured at 30 °C for 2 to 3 days.

The LUC was used to validate the DNA-protein interactions between GhMYB25-like and its target genes. Initially, expression vectors were constructed and transferred to Agrobacterium. The promoter fragments of GhHD-1 and GhPDF2 were cloned into the pGreenII 0800-LUC vector, while the full-length CDS sequences of GhMYB25-like were cloned into the pGreenII 62-SK vector. The primer sequences used for cloning can be found in Supplementary Data Set 3. These recombinant vectors, along with the empty pGreenII 0800-LUC and pGreenII 62-SK vectors were transferred to the Agrobacterium tumefaciens strain GV3101. Subsequently, the transferred GV3101 strains were transiently expressed in leaves of Nicotiana benthamiana to observe the expression of firefly luciferase. The GV3101 strains containing the pGreenII 0800-LUC vector and the pGreenII 62-SK vector were activated and mixed at a ratio of 1:9. The concentration of the mixed strains was adjusted to OD600 = 0.6 with a suspension containing 10 mM MgCl₂, 10 mM MES (pH 5.6), and 20 μM acetosyringone (AS). The mixed strains were then infiltrated into N. benthamiana leaves to induce transient expression of the vectors. After infiltration, the N. benthamiana leaves were cultivated for 48 to 72 h. Luciferase activity was recorded using two methods. For live imaging, the infiltrated N. benthamiana leaves were harvested and treated with a luciferin reaction mixture (1 mM Luciferin, 0.2% Triton X-100) in the dark for 5 min. The treated leaves were subsequently observed using a whole-body fluorescent imaging system (Berthold, LB985 NightSHADE). For quantitation, the luciferase activity was measured using a dual luciferase reporter gene assay kit (Yeasen, 11402).

EMSA

The GhMYB25-like, GhMYB25-like_At^hapT, and GhMYB25-like_At^CR2 proteins, which were used for EMSA, were expressed and purified. The codon-optimized GhMYB25-like and GhMYB25-like_At^hapT CDS fragments were cloned into the pMLink vector for protein expression. The GhMYB25-like/GhMYB25-like_At^hapT-2×strep-sfGFP fusion proteins were induced in mammalian expression system. After cultivation for 60 h at 37 °C, all mammalian cells were harvested and lysed using a high-pressure disruptor. The protein was then purified with Streptactin Agarose Resin, the eluted protein was desalinized to remove the biotin in the buffer. Finally, tobacco etch virus protease was employed to remove the 2×strep-sfGFP tags, resulting in the isolation of GhMYB25-like/GhMYB25-like_At^hapT proteins. The truncated codon-optimized GhMYB25-like_At^CR2 CDS fragments were cloned into the pET28a vector for protein expression. The GhMYB25-like_At^CR2-6×His-small ubiquitin-like modifier (SUMO) fusion proteins were induced in Escherichia coli BL21 (DE3) cells with 0.2 mM isopropyl-1-thio-D-galactopyranoside at 16 °C for 14 h. Following induction, all E. coli cells were harvested and lysed using a high-pressure disruptor. The protein was then purified with a Ni-NTA column, followed by the removal of the 6×His-SUMO tag with ULP1 protease. Finally, GhMYB25-like_At^CR2 proteins were purified through ion-exchange chromatography, resulting in the isolation of GhMYB25-like_At^CR2 proteins.

The purified GhMYB25-like, GhMYB25-like_At^hapT, and GhMYB25-like_At^CR2 protein were incubated with EMSA probes to assess their affinity. First, the cis-elements on the promoters of GhHD-1 and GhPDF2, were predicted using the PlantRegMap website, and 20 to 30 bp EMSA probes were designed accordingly. The sequences of these probes can be found in Supplementary Data Set 3. Next, the EMSA probes were labeled with 5′-biotin in Sangon Biotech Company. Following this, the GhMYB25-like, GhMYB25-like_At^hapT, and GhMYB25-like_At^CR2 proteins (ranging from 0 to 0.3 μM) were then incubated with the labeled EMSA probe (final concentration 0.2 nM) in a binding buffer (20 mM HEPES, pH 7.6, 1 mM ethylenediaminetetraacetic acid (EDTA), 10 mM (NH₄)₂SO₄, 30 mM KCl, 1 mM dithiothreitol, Tween 20, 0.2% (w/v), poly (dI-dC)). The incubation was performed on ice for 1 h. After incubation, the mixture was loaded onto 6% native polyacrylamide gels and subjected to electrophoresis in 0.5× Tris-borate-EDTA buffer for 1 h. After transferring DNA from gel to membrane, the biotin-labeled DNA was then detected by LightShift™ Chemiluminescent EMSA Kit (Thermo Fisher Scientific, 20148).

LCI assay and BiFC assay

The LCI method was used to identify the physical interaction between GhMYB25-like and its mutations. Firstly, expression vectors were constructed and transferred to Agrobacterium. The CDS of GhMYB25-like, GhMYB25-like_At^hapT and GhMYB25-like_At^CR2 were cloned into JW771, and the CDS of GhMYB25-like was cloned into JW772. The primer sequences used for cloning can be found in Supplementary Data Set 3. These positive plasmids were then transformed into A. tumefaciens strain GV3101. Subsequently, the activated bacterial suspension was transiently expressed in N. benthamiana leaves to observe the luciferase signal. The control group consisted of the combination harboring empty JW771 and GhMYB25-like-JW772, while the experimental group consisted of the combination harboring GhMYB25-like/GhMYB25-like_At^hapT/GhMYB25-like_At^CR2-JW771 and GhMYB25-like-JW772. The infected leaves were observed using a whole-body fluorescent imaging system.

The 2in1 cloning system was used in BiFC assay to validate the interaction between GhMYB25-like and its mutations (Grefen and Blatt 2012). First, the CDS of GhMYB25-like was cloned into the pNONOR221-P1P4 vector, while the CDS of GhMYB25-like/GhMYB25-like_At^hapT/GhMYB25-like_At^CR2 was cloned into the pNONOR221-P12P33 vector via Gateway BP recombination to obtain the intermediate vectors. Then, the GhMYB25-like-pNONOR221-P1P4 vector and an empty pNONOR221-P2P3 vector were cloned into pBiFCt-2in1-NN vectors, serving as the control group. Additionally, theGhMYB25-like-pNONOR221-P1P4 vector and GhMYB25-like/GhMYB25-like_At^hapT/GhMYB25-like_At^CR2-pNONOR221-P2P3 vector were cloned into the pBiFCt-2in1-NN vector, forming the experimental group. Following this, the pBiFCt-2in1 vectors were transformed into A. tumefaciens strain GV3101 and then transiently expressed in N. benthamiana leaves. Finally, the yellow fluorescence (YFP channel imaging: 515 nm laser at 29% intensity, 100 nm collection bandwidth, and 3× gain) was observed using confocal microscopy (Olympus FV1200) after 48 h cultivation.

Candidate gene association analysis with lint percentage

The resequenced data and multi-environment LP phenotype data used for association analysis were previously published (Ma et al. 2018; Li et al. 2021). For the resequenced data, SNPs with a minor allele frequency (MAF) of ≥0.05 and a maximum missing rate of ≤0.95 genome-wide were filtered for the purpose of facilitating PCA and kinship construction. Furthermore, SNPs and Indels (MAF ≥ 0.01) in the CDS region and within 2 Kb upstream of the start codon of candidate genes were imputed using Beagle v4.1.

For the phenotype data, the multi-environment LP data was merged using Best linear unbiased predictor (BLUP) (Poland et al. 2011). Two different association analysis methods were employed. First, the mixed linear model (P + G + Q + K) of TASSEL5.0 was utilized for candidate gene association analysis (Bradbury et al. 2007). Second, the rare allele association analysis was conducted using the R packages seqMeta (Wu et al. 2011). This method associated gene variations and LP BLUP values, with the first 3 principal components serving as covariates (Lee et al. 2014).

Population genetic analysis

Population genetic analysis was performed on SNP and Indel variations in the upstream and downstream regions of GhMYB25-like_At using a dataset of 1547 resequenced G. hirsutum accessions (Li et al. 2021). This analysis aimed to investigate the selection of GhMYB25-like_At in these accessions.

To assess the patterns of LD, an LD heatmap was constructed using Haploview 4.2 software (Barrett et al. 2005). This analysis included SNPs and Indels located in the CDS region and 2Kb upstream of the start codon of GhMYB25-like_At. Nucleotide diversity (π) in the vicinity of GhMYB25-like_At was explored using 2,000 bp windows and 100 bp steps in different subpopulations, with a focus on the 10 kb upstream and downstream regions of the gene. Plink 1.9 software was used for this analysis (Purcell et al. 2007). EHH analysis was conducted to examine SNPs and Indels in the 800 kb flanking region of GhMYB25-like_At and the analysis was performed using hapbin software (Maclean et al. 2015).

Statistical analysis

Statistical analyses were performed as described in the text and figure legends. Statistical data can be found in Supplementary Data Set 4.

Accession numbers

Sequence information for the cotton genes in this study can be found in the Cotton Genome Database (https://cottonfgd.org/, Gossypium hirsutum AD1, upland cotton, HAU assembly) according to the accession numbers as shown in Supplementary Data Set 1.

Author contributions

L.T., Z.L., and X.Z. conceived the project and designed the experiments. G.Z. performed the map-based cloning. Y.Q. performed protoplast isolating. G.Z., Y.Q., J.Y., and J.L. performed bioinformatics analysis. G.Z., Y.L., Z.Y., and H.T. contributed to the vector construction. Y.L. and M.S. performed the transgenic experiments. G.Z., J.X., and S.M. expressed the fusion proteins. H. H. constructed the mapping population. G.Z. and L.T. wrote the manuscript, Z.L., S. J., M.W., and X.Z. revised the manuscript with feedback from all other authors.

Supplementary data

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

Supplementary Figure S1. Lint and fuzz phenotypes of LM self-crossed progeny.

Supplementary Figure S2. Frequency distributions of lint and fuzz traits in the LM selfing population.

Supplementary Figure S3. Smooth ED⁵ plot of BSA.

Supplementary Figure S4. Fragments per Kilobase of exon model per million mapped fragments of GhMYB25-like_At and GhMML4_At in 0-DPA and 5-DPA ovule epidermis in LF, LM, and LL.

Supplementary Figure S5. Amino acid sequence alignment of GhMYB25-like_At between LF and LL.

Supplementary Figure S6. K105 is a conserved site in the MYB binding domain.

Supplementary Figure S7. GhMYB25-like gene editing induced by CRISPR/Cas9 system.

Supplementary Figure S8. GhMYB25-like_At gene editing induced by CRISPR/Cas9 system.

Supplementary Figure S9. t-SNE projection plot showing the expression characteristics of known marker genes in individual cells at −1.5, −1, −0.5, and 0 DPA.

Supplementary Figure S10. Comparisons of different cluster types between LF and LL.

Supplementary Figure S11. UMAP projection plots showing the expression characteristics of known marker genes in −1.5 and −1 DPA major clusters.

Supplementary Figure S12. The overlapped distribution of precursor fiber cells between LF and LL.

Supplementary Figure S13. UMAP projection plots showing the expression characteristics of known marker genes in −1.5 and −1 DPA reclustering fiber clusters.

Supplementary Figure S14. Expression heatmap of fiber initiation-related genes.

Supplementary Figure S15. GhMYB25-like^hapT inhibits the function of GhMYB25-like.

Supplementary Figure S16. DNA polymorphism and LD block analysis of GhMYB25-like_At in 1547 G. hirsutum accessions.

Supplementary Figure S17. Expression patterns of Ghir_A12G017450 and GhD12G017660 shown by t-SNE projection plots.

Supplementary Table S1. Segregation ratio in the LM selfing population and LF × LL F2 population.

Supplementary Table S2. The candidate region is determined by BSA between LF pool and LL pool.

Supplementary Data Set 1. Summary of different expression genes related to fiber protrusion.

Supplementary Data Set 2. The A314T mutation of GhMYB25-like_A12 in 1547 G. hirsutum accessions.

Supplementary Data Set 3. Primers used in this study.

Supplementary Data Set 4. Summary of statistical tests.

Funding

This work was supported by grants from the National Natural Science Foundation of China (projects No. 31830062 to X.-L.Z.), the National Key Research and Development Program of China (projects No. 2022YFF1001400 to Lili Tu), and the funds from the Interdisciplinary Sciences Research Institute of Huazhong Agricultural University (2662021JC005).

Data availability

The scRNA-seq data, the next-generation sequencing data, and RNA-seq data have been deposited in the NCBI SRA database (https://www.ncbi.nlm.nih.gov/bioproject/) with BioProject number PRJNA600131, PRJNA848976 and PRJNA1033545, respectively. For any other data inquiries, interested parties can contact the corresponding authors and request access.

Dive Curated Terms

The following phenotypic, genotypic, and functional terms are of significance to the work described in this paper:

• LTPG1 Gramene: AT1G27950

• LTPG1 Araport: AT1G27950

• WRKY40 Gramene: AT1G80840

• WRKY40 Araport: AT1G80840