Introduction

Cotton (Gossypium hirsutum L.) fibers are single-cell unbranched trichomes developed from epidermal cells of a seed. Approximately one in four epidermal cells develop into fiber (Steward 1975). There are four stages observed during cotton fiber development, including initiation, elongation, secondary cell wall synthesis, and maturation (Kim and Triplett 2001). Most commercially grown cotton cultivars produce two types of fibers: lint fiber that is easily detached during the ginning process and fuzz fiber that strongly adheres to the seed coat. Lint fiber cells grow approximately 30 mm long and are relatively fine, spinnable, and used for textile production. The fuzz fibers are short (about 5 mm length) and are considered an undesirable trait since they hinder seed germination and increases energy consumption during the ginning process (Bechere et al. 2011). There are differences in the time when lint and fuzz fiber cells emerge from the epidermal layer: lint fibers initiate around 3 days before to the day of anthesis (DOA), whereas fuzz fibers delay initiation until 3–5 days post-anthesis (DPA) (Steward 1975; Zhang et al. 2007).

Cotton fiber mutants are valuable resources for studying functions of altered genes and their roles in fiber development. A number of spontaneous fiber mutants, as well as artificially mutagenized lines, are known to cotton scientists. The single dominant N1 mutation is responsible for the ‘fuzzless seed’ phenotype that has no fuzz and different degrees of lint on the seeds (Kearney and Harrison 1927). The recessive n2 mutant produces a fuzzless seed phenotype with normal lint fibers (Ware et al. 1947). Recently a new partially dominant N5 mutant has been reported that produces a tufted-fuzzless seed phenotype (Zhu et al. 2020). A recessive tufted-fuzzless seed n4t mutant was developed through chemical mutagenesis (Bechere et al. 2012). The n4t seeds are partially fuzzless (tufted) with fuzz grown around the micropyle and chalazal ends of the ovule. The n4t locus is not allelic to N1 or n2 (Bechere et al. 2012). Phenotypic evaluation of fiber characteristics determined that the lint yield and percentage of the n4t were not significantly different than that of the wild-type, whereas the short fiber content and amount of coat neps were 15% and 45% lower in the mutant’s fiber; the n4t mutant line required less net energy for ginning compared with fuzzy cotton cultivars (Bechere et al. 2012; Bechere and Auld 2014).

Using a map-based cloning approach the N1 mutation was linked to a gene on the chromosome (Chr.) A12 that was annotated as a MYB-MIXTA-like (MML) transcription factor GhMML3_A12 (Wan et al. 2016). Small interfering RNAs originating from bidirectional transcripts of GhMML3_A12 reduced its transcript level and therefore caused the fuzzless seed phenotype (Wan et al. 2016). RNA interference (RNAi) suppression of the GhMML3_A12 gene resulted in fiberless seeds but normal trichomes elsewhere which demonstrated a critical role of this gene in early cotton fiber development (Walford et al. 2011). Mutations responsible for the n2 fuzzless phenotype were linked to two loci, on Chr. A12 and D12 (Turley and Kloth 2002; Song et al. 2010). Using a map-based cloning approach the gene from Chr. D12 locus has been isolated. The gene is annotated as another MML transcription factor GhMML4_D12, which was proposed to be responsible for lint fiber development (Wu et al. 2017). The most recent survey study has identified two mutations in the n2 line, including a non-synonymous SNP (nsSNP) in the third exon of the MML3-A12 gene and an nsSNP in the MML4-D12 gene that removes stop codon, leading to an extension of the coding region of this gene (Naoumkina et al. 2020). The genetic locus responsible for the N5 tufted-fuzzless seed phenotype had been identified on Chr. D13 of G. hirsutum (Zhu et al. 2020).

Progress has been made in understanding the genetic control of lint and fuzz fiber development. Also, differences have been found in pathways of fiber development between diploid and tetraploid cotton species. In tetraploid G. hirsutum it has been proposed that both MML3_A12 and MML3_D12 are involved in lint development, whereas fuzz formation is mainly determined by the expression level of MML3_D12 at about 3 DPA (Zhu et al. 2018). In diploid G. arboreum the GaMML3 gene was also associated with fuzz fiber development, however its role was not as significant as GaGIR1, which has been proposed to be the main factor associated with fuzz development in G. arboreum according to investigations of several different fuzzless mutants (Feng et al. 2019; Liu et al. 2020). Besides the known key factors, such as GhMML3 and GaGIR1, there are many more yet unidentified genes that are also important for cotton fiber development. For example, two candidate genes from the N5 locus (Chr. D13) have been proposed to be causative genes, which might trigger transcriptional reprograming of fiber cells leading to a tufted-fuzzless phenotype in G. hirsutum (Zhu et al. 2020).

The aim of our study was to identify the n4t locus and potential candidate genes that cause the tufted-fuzzless seed phenotype. We developed an F2 mapping population consisting of 602 individuals from a cross between the n4t mutant and a wild type (WT). We sequenced the genomes of DNA bulks from F2 progenies showing the mutant phenotype along with their parents. The n4t locus was mapped within a genomic region of about 411 kb on Chr. D04. The n4t locus contains 45 genes according to a G. hirsutum TM-1 genome assembly (Wang et al. 2019). The putative candidate genes that might cause the n4t fiber phenotype are discussed in this study.

Materials and methods

Plant materials and development of the F2 population

The n4t tufted-fuzzless cotton line was developed through EMS chemical mutagenesis of the cultivar SC9023 (PI 590933) (Bechere et al. 2012; Bechere and Auld 2014). This mutant has reduced density of short fibers (fuzz) that remain attached to the seed after spinnable lint fibers are removed and have only a small amount of fuzz attached to the micropyle and chalazal ends of the seed. Crosses were made between SC9023 and homozygous n4t mutant plants in the summer of 2018 in a field in New Orleans, LA, USA. Four F1 plants were grown in New Orleans’s greenhouse during the winter to produce F2 seeds. F2 plants (602 individuals) were grown during the summer of 2019 in a field in New Orleans. Standard field management practices were applied.

Fiber quality testing and statistical analysis

Ten naturally opened bolls were manually harvested from each individual plant for fiber quality testing. The cotton samples were ginned using a laboratory roller gin. Fiber quality traits of F2 populations were measured using a High Volume Instrument (HVI, Uster Technologies Inc.). The fiber testing was conducted by the Cotton Fiber Testing Lab, USDA-ARS-SRRC, New Orleans, LA. A t-test (JMP Genomics 9.0; SAS, Cary, NC, USA) was used to compare fiber characteristics between the two groups of samples, including WT and n4t.

DNA isolation and sequencing of super bulked segregants

Young leaves were collected from each individual F2 plant within the segregating population. DNA was isolated as previously described (Fang et al. 2010). The concentration of DNA samples was measured using a NanoDrop 2000 spectrophotometer (Thermo Fisher Scientific, Waltham, MA).

F2 plants were selected for sequencing by a bulked segregant analysis approach (Michelmore et al. 1991; Takagi et al. 2013; Thyssen et al. 2014). We selected six plants of category-1 for the first bulk and 41 plants of category-2 for the second bulk (description of seed categorization is described in the Results section). DNAs from the two mutant bulks along with the n4t mutant and WT (SC9023) parental DNAs were sent to Novogene Corporation (Chula Vista, CA, USA) for library preparation, and whole genome sequencing using Illumina HiSeq 2500 with paired-end 150 bp reads. Each was sequenced at 25 × coverage (about 65 Gb).

Identification of the genomic locus harboring the n 4 t mutation

Sequence reads of the two bulked DNAs and two parents were aligned to the reference Gossypium hirsutum cv. Texas Marker-1 (TM-1) genome (Wang et al. 2019) using BWA-MEM software (Zhu et al. 2018). Variants were identified with samtools mpileup (with “-Ego” flags) and bcftools (“call–vm”) software (Dettmer et al. 2007; Obata and Fernie 2012). A histogram was generated by counting the number of single nucleotide polymorphisms (SNP) and insertion/deletions (indel) in 1 Mb intervals. Primers were designed to analyze the segregation of the SNP markers in the F2 population according to our previously published protocol (Thyssen et al. 2014). The primer sequences are listed in Table S1.

Association analyses between the n4t phenotype and genetic markers were performed using the composite interval mapping (CIM) model in JMP Genomics 9.0. The seed phenotype trait was numerically annotated for association analysis: fuzzy WT—0; category-3—0.7; category-2—0.8 and category-1—0.9. The numbers reflect the degree of the nakedness of seeds: the less fuzz the higher number. Therefore, the seeds of category-1 with a small amount of fuzz attached to both ends were annotated as 0.9, while fully fuzzy WT was annotated as 0. The SNP data was also numerically annotated: homozygous WT—0; heterozygous—1; and homozygous mutant—2. The genotype probability data set for the 602 F2 individuals was built using the same software. The forward regression method was used for control marker selection with a window size of 10.0 cM and the control marker number was 5. The expectation–maximization (EM) algorithm, at a threshold of 10 (LOD > 10), was used to identify the significant association associated locus.

RNA isolation and RT-qPCR

Cotton bolls were harvested at DOA, and 3 and 5 DPA. Epithelial cells were separated from ovules using a glass bead shearing technique (Taliercio and Boykin 2007). Total RNA was isolated using a Sigma SpectrumTM Plant Total RNA kit (Sigma-Aldrich, St. Louis, MO). The cDNA synthesis reactions were performed using the iScriptTM cDNA Synthesis Kit (Bio-Rad Laboratories, Hercules, CA) according to the manufacturer’s instructions with 1 μg of total RNA per reaction. The qPCR reactions were performed with iTaq™ SYBR® Green Supermix (Bio-Rad Laboratories Inc., Hercules, CA) in a Bio-Rad CFX96 real time PCR detection system in three biological replicates with two technical replicates. Amplification of the endogenous 18S rRNA (Genbank accession U42827) was used as a reference for normalization. The relative expression levels were calculated using \(2^{{ - \Delta \Delta C_{{\text{t}}} }}\) (Pfaffl 2001). Primers are provided in Table S2. A two-way ANOVA statistical test (GraphPad Prism 7) was applied to identify significantly differentially expressed samples.

Results

The n 4 t tufted-fuzzless phenotype and its genetic control

The n4t mutation affected only the seed phenotype (Fig. 1). There were no morphological differences observed between the WT and the mutant in the plant’s vegetative organs including stems, leaves and flowers. It was previously reported that no significant differences in lint yield and lint percentage were observed between the WT and the mutant lines [see also details in the Introduction] (Bechere et al. 2012; Bechere and Auld 2014). The homozygous n4t seeds are partially fuzzless, with tuft grown around the micropyle and chalazal ends of the seed (Fig. 1d). F1 seeds are fuzzy, indicating that the n4t mutant phenotype is recessive (Fig. 2a). Seeds from F2 segregating plants showed different degrees of fuzz (Fig. 2b–d). We grouped the tufted seed fiber phenotypes of F2 segregating plants into three categories: category-1 (six plants) with clearly tufted seeds (Fig. 2b); category-2 (41 plants) with a mixture of clearly tufted seeds and seeds that showed sparsely distributed short fuzz fiber cells around the seed (Fig. 2c); and category-3 (103 plants) where all seeds in the boll showed varying degrees of fuzz fiber cells that were nevertheless clearly less than normal for WT seeds (Fig. 2d).

Fig. 1
figure 1

Bolls and seeds of wild type and tufted-fuzzless n4t mutant. a Wild type boll; b n4t boll; c seeds of wild type; d seeds of n4t mutant

Full size image
Fig. 2
figure 2

Seeds of F1 plants and three categories of seeds from F2 segregating population. a F1 seeds; b seeds of category-1; c seeds of category-2; d seeds of category-3

Full size image

We observed 452 plants with wild-type fuzzy seeds and 150 plants with n4t tufted-fuzzless seeds among 602 individuals of the F2 segregating population. Therefore, the segregation of seed fiber phenotypes in this population fits a three wild type to one mutant ratio, suggesting one recessive Mendelian locus.

Fiber quality analysis

To determine whether or not the n4t phenotype affects the fiber quality traits we measured the fiber properties of every individual plant from the F2 population using an HVI. HVI-measured fiber properties of WT and n4t phenotype-groups from the F2 segregating population were subjected to statistical analysis to identify significant differences between the two groups of samples. Among the fiber traits measured by HVI, only fiber length showed a significant difference between the WT and the n4t samples. Analysis of the upper half mean length (UHML) of fiber showed that samples from fuzzless mutant were significantly (p < 0.0001) longer than wild-type samples (Fig. 3).

Fig. 3
figure 3

Statistical analysis of fiber length trait between WT and n4t groups of samples in segregating F2 population. The box plot was built by using R software (http://shiny.chemgrid.org/boxplotr/). Upper half mean length (UHML) of cotton lint fiber is presented in mm. Center lines show the medians; box limits indicate the 25th and 75th percentiles as determined by R software; crosses represent sample means; bars indicate 95% confidence intervals of the means; width of the boxes is proportional to the square root of the sample size; p value is determined by t test; data points are plotted as grey circles; n = 452, 150 sample points

Full size image

The n 4 t locus is positioned on chromosome D04

Since the n4t tufted-fuzzless trait was observed to be recessive, we did not make a DNA bulk from F2 plants showing the wild type phenotype (because those plants may have a heterozygous genotype). The two DNA pools, including category-1 (six samples) and category-2 (41 samples), from the F2 segregating population that showed mutant phenotypes along with the parental DNAs were used for whole-genome sequencing. The two categories of bulked samples with different degrees of fuzz on their seeds were sequenced separately to enable the detection of differences in genomic sequences. However, we did not detect noticeable differences in sequences at the candidate loci between two bulks. We identified SNPs and indels by aligning the sequence reads to the reference TM-1 genome (Wang et al. 2019). To build the histogram of the distribution of polymorphisms our initial criteria was that the SNPs and indels were homozygous for the reference (TM-1) type in the parent WT (SC9023), homozygous for an alternative allele type in the parent n4t mutant and both homozygous and heterozygous for an alternative allele in F2 bulks. This criteria of polymorphisms selection showed two major peaks of SNPs and indels on Chr. D07 (3381) and Chr. D04 (673) [Figure S1]. When we used more stringent parameters by selecting only homozygous SNPs and indels for the alternative allele in both F2 bulks. This approach allowed us to identify only one major locus, containing 164 SNPs and indels on Chr. D04 (Fig. 4). Both initially identified regions of significant polymorphism on D04 and D07 were considered as possible genomic locations of the n4t locus. Therefore, we developed PCR-based SNP markers (Table S1) from Chr. D04 and Chr. D07 to test the association of markers with phenotype in the segregating F2 population.

Fig. 4
figure 4

Histogram of SNPs and indel distribution between the WT and the n4t tufted-fuzzless mutant across 26 chromosomes of G. hirsutum. The major peak on Chr. D04 represents the SNPs and indels homozygous for an alternative allele type in two F2 bulks showing the n4t phenotype

Full size image

We ran 14 SNP markers from Chr. D04 and six SNP markers from Chr. D07 on the segregating F2 population of 602 individuals. The n4t tufted-fuzzless phenotype was significantly associated with the Chr. D04 locus and the highest LOD score was associated with the D4_549 marker (Fig. 5a). There was no significant association between SNPs from Chr. D07 and the n4t phenotype (Fig. 5b). The flanking region of n4t locus was between the D4_546 and D4_550 markers, which is about 440 kb according to the G. hirsutum TM-1 reference genome assembly (Wang et al. 2019).

Fig. 5
figure 5

Association mapping of n4t locus. Composite interval mapping analysis of association of the n4t tufted phenotype with genetic markers on Chr. D04 (a) and Chr. D07 (b). c Physical map of the n4t locus on Chr. D04 G. hirsutum TM-1 reference genome. Table S1 provides the sequences of SNP markers, polymorphism and positions on chromosome

Full size image

Putative candidate genes at the n 4 t locus

The genomic region flanked by our D4_546 and D4_550 SNP markers contained 45 annotated genes according to the TM-1 genome assembly (Wang et al. 2019). To find possible candidate genes we initially evaluated their expression in different cotton tissues by using the publicly available ccNET gene expression database (You et al. 2017), and then the expressed genes were tested by RT-qPCR analysis in early developing fiber cells (DOA, 3 and 5 DPA) of WT and the n4t mutant. Of the 45 genes, 42 were available in the ccNET database and among those only 22 genes were expressed above the detection level (Fig. 6). Though, out of 22 expressed genes, only seven genes showed significantly (p < 0.05) different expression between WT and the n4t mutant in at least one fiber development time point (Fig. 7).

Fig. 6
figure 6

Heat map of expression of the genes from the n4t locus. Expression data for different cotton tissues were obtained from ccNET database (http://structuralbiology.cau.edu.cn/gossypium/)

Full size image
Fig. 7
figure 7

RT-qPCR analysis of seven selected genes from the n4t locus plus two previously reported fiber-related MML3 genes. Error bars indicate standard deviation from three biological replicates with two technical replicates. Asterisks on x-axis indicate significantly different expressions between the wild type (Sc9024) and the n4t samples. The number of asterisks reflects the level of significance (**p < 0.001 and ***p < 0.0001). Table S2 provides primers sequences and annotation of the genes used in RT-qPCR

Full size image

All seven genes were downregulated in developing n4t fiber cells (Fig. 7). Ghir_D04G019250, annotated as ribosomal protein L3B, had reduced expression at 3 DPA. Ghir_D04G019300, annotated as cytochrome c2, was downregulated at DOA. Ghir_D04G019410, annotated as NHL domain transcription factor, was downregulated at 3 DPA. Ghir_D04G019490, annotated as lumazine synthase, was downregulated at DOA and 3 DPA. The most highly expressed gene Ghir_D04G019570, annotated as ribosomal protein L18e/L15, was downregulated at DOA. Ghir_D04G019600 and Ghir_D04G019640, annotated as o-fucosyltransferase and cellulase were downregulated at 3 DPA.

We compared genomic sequences of differentially expressed genes between WT and the n4t mutant. Ghir_D04G019490 has a SNP in its fifth intron and three SNPs and three indels in its promoter region (Table S3). Ghir_D04G019640 has an indel in its second intron and a missense substitution in its third exon, changing isoleucine to methionine. The other expressed genes have identical genomic sequence between SC9023 and the n4t mutant.

Also, we tested expression levels of the MML3_A12 and MML3_D12 genes, which are known to be key regulators of fiber development (Walford et al. 2011; Wan et al. 2016; Zhu et al. 2018). The MML3_A12 gene was significantly down-regulated in the mutant at DOA, 3 and 5 DPA, whereas the MML3_D12 was significantly downregulated at 3 and 5 DPA (Fig. 7).

Discussion

The ratio of seed fiber phenotypes in the F2 population was three (fuzzy) to one (tufted-fuzzless), indicating that the n4t tufted-fuzzless trait is controlled by a single recessive locus; this finding was consistent with the segregation ratio reported before (Bechere et al. 2012). The previous study did not detect a significant difference in lint yield and percentage between the n4t and WT type cultivar, which with the improved ginning efficiency of fuzzless phenotype makes the n4t trait desirable to use in breeding programs (Bechere et al. 2012). In the current study, we analyzed major fiber properties, as measured by HVI, in the F2 population and found that the fiber length of the fuzzless seeds was significantly longer than in the WT cultivar SC9023 (Fig. 3).

Identification of the n4t locus and the gene responsible for tufted-fuzzless trait will provide the means for genetic manipulations to improve ginning efficiency and lint length of modern cotton cultivars. We have identified the candidate locus through genetic mapping of the F2 population (Fig. 5). The n4t locus has been mapped to Chr. D04 where previously no genes have been associated with fiber development. Seven genes from the n4t locus showed significant downregulation in developing fibers of the fuzzless mutant compared to the WT. Ghir_D04G019570, annotated as ribosomal protein L18e/L15, was the most highly expressed gene at the n4t locus and was significantly downregulated in the tufted-fuzzless mutant at DOA (Fig. 7). The ribosomal L18e/L15 gene has been associated with fiber strength QTL in an independent study and showed high expression in the superior fiber MD52ne cultivar compared to a weaker fiber line MD90ne (Islam et al. 2016).

We have detected polymorphisms in the genomic sequence of the n4t fuzzless mutant in two genes. Ghir_D04G019640, annotated as cellulase (glycosyl hydrolase family 5), has a substitution of isoleucine to methionine that can lead to changes in protein structure or function. Cellulases catalyze the breaking of β-1, 4 glycosidic bonds in the interior of longer β-1, 4 glucose chains; genes from this family are suggested to possess multiple functions, including accelerating cell growth through cell wall loosening, and wall reconstruction during cell expansion (Cosgrove 2005; Yu et al. 2013). Antisense suppression of poplar cellulase (PopCel1) reduced leaf growth (Ohmiya et al. 2003). Higher expression of PopCel1 in Arabidopsis increases the size of the rosettes through promoting an increase in cell size (Park et al. 2003). Research in Arabidopsis and poplar showed the involvement of cellulases in leaf growth; however, the role of the Ghir_D04G019640 in cotton fiber development has not been studied yet. Given the expression pattern of the Ghir_D04G019640 in developing fiber cells (Fig. 7) and the observation of downregulation of this gene at 3 DPA in the n4t mutant we suggest this gene as a candidate for further studies to understand its role in fuzz fiber development.

We detected SNPs and indels in the promoter region of the Ghir_D04G019490 gene, annotated as lumazine synthase. Lumazine synthase catalyzes the penultimate step of riboflavin biosynthesis, which is essential for critical cellular processes such as the citric acid cycle, fatty acid oxidation, photosynthesis, mitochondrial electron transport, and de novo pyrimidine biosynthesis (Jordan et al. 1999). The Ghir_D04G019490 gene is our favored candidate since it is located close to SNP marker D04_549, which showed the highest LOD score association with the fuzzless phenotype (Fig. 5). It is unclear what causes downregulation of this gene in the mutant; though, polymorphism in the promoter region of the Ghir_D04G019490 should not be ruled out as a causative factor of expression change. The function of the Ghir_D04G019490 gene is not characterized yet in cotton. Results of this study indicate this gene might be involved in fuzz fiber development and should be considered for future research.

What causes different degrees of fuzz fiber cell development on the n4t seeds is unclear. The genomic sequences of the two studied categories of seeds with different degrees of fuzz were similar, therefore could not explain why such differences occur. However, factors such as small RNAs, epigenetic modifications or other mechanisms that modulate penetrance should not be ruled out. For example, the N1 mutant has a wide range of fiber densities, even on the bolls of a single plant (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.

The current model of fuzz fiber development in tetraploid Gossypium species proposes that it is mainly determined by the expression level of MML3-D around 3 DPA (Zhu et al. 2018). Since we have detected significant downregulation of both MML3-A and MML3-D genes in the n4t fuzzless mutant at 3 and 5 DPA (Fig. 7), we do not challenge the key role of MML3 family genes. But, we have identified a new n4t locus that has not previously been associated with fuzz fiber formation, which suggests that there are more unknown genes involved in fiber development.