The Hairless Stem Phenotype of Cotton (Gossypium barbadense) Is Linked to a Copia-Like Retrotransposon Insertion in a Homeodomain-Leucine Zipper Gene (HD1)

Abstract

Cotton (Gossypium) stem trichomes are mostly single cells that arise from stem epidermal cells. In this study, a homeodomain-leucine zipper gene (HD1) was found to cosegregate with the dominant trichome locus previously designated as T1 and mapped to chromosome 6. Characterization of HD1 orthologs revealed that the absence of stem trichomes in modern Gossypium barbadense varieties is linked to a large retrotransposon insertion in the ninth exon, 2565 bp downstream from the initial codon in the At subgenome HD1 gene (At-GbHD1). In both the At and Dt subgenomes, reduced transcription of GbHD1 genes is caused by this insertion. The disruption of At-HD1 further affects the expression of downstream GbMYB25 and GbHOX3 genes. Analyses of primitive cultivated accessions identified another retrotransposon insertion event in the sixth exon of At-GbHD1 that might predate the previously identified retrotransposon in modern varieties. Although both retrotransposon insertions results in similar phenotypic changes, the timing of these two retrotransposon insertion events fits well with our current understanding of the history of cotton speciation and dispersal. Taken together, the results of genetics mapping, gene expression and association analyses suggest that GbHD1 is an important component that controls stem trichome development and is a promising candidate gene for the T1 locus. The interspecific phenotypic difference in stem trichome traits also may be attributable to HD1 inactivation associated with retrotransposon insertion.

TRICHOMES are mostly single-celled structures formed by the unidirectional extension of the outer layer of epidermal cells. They are of central importance to the Gossypium (cotton) genus, being its major economic product (lint fiber), an important taxonomic character, and an important determinant of pest resistance and other traits. The development of cotton fibers and Arabidopsis trichomes likely shares a similar regulatory mechanism involving closely related genes as well as a similar development and patterning signaling pathway, making trichomes an excellent system in which to study genetic mechanisms regulating cotton fiber development (Wan et al. 2014).

Trichomes are found on the stems of many Gossypium species and are even reflected in the species name of the most widely cultivated species, Gossypium hirsutum (whose name is derived from the Latin word hirsutus, meaning “hairy”). Stem trichomes exhibit striking phenotypic differences between different species and even among cotton varieties within the same species. Despite the fact that stem trichomes show near-discrete genetic variations and are easier to detect than fibers, relatively few studies have focused on the development of cotton stem trichomes and their underlying regulation. Until now, through limited classical genetic studies and genetic mapping, at least five loci (T1–T5) (Lee 1985) and several additional genes and/or quantitative trait loci (QTLs) that modify the densities and distributions of stem trichomes (Wright et al. 1999; Lacape and Nguyen 2005) have been identified. The T1 locus, which was mapped to chromosome 6 of the A genome, was predicted to contain major genes that regulate trichome initiation (Lacape and Nguyen 2005). This locus also has been reported frequently to cosegregate with multiple QTLs related to fiber fineness, length, elongation, color, and length uniformity (Wan et al. 2007), indicating that the underlying gene(s) in this locus is(are) involved in multiple epidermal trichome development processes.

Some important trichome developmental and patterning-related genes include GLABRA1 (GL1), GLABRA2 (GL2), basic helix-loop-helix protein (GL3), ENHANCER OF GLABRA3 (EGL3), and WD40 protein TRANSPARENT TESTA GLABRA1 (TTG1) in Arabidopsis (Larkin et al. 1994; Payne et al. 1999; Esch et al. 2003; Zhang et al. 2003; Wang et al. 2004; Humphries et al. 2005; Zhao et al. 2008; Cui et al. 2013; Zhou et al. 2014). Their homologous genes in cotton also have been identified as critical for trichome development. For example, MYB109 encodes a product that works like the Arabidopsis GL1 gene and is a key regulator in the development of epidermal hairs. GaMYB2, which was cloned from G. arboreum and is functionally homologous to Arabidopsis GLABRA1 (GL1), can complement the GL1 mutant phenotype involved in trichome formation (Wang et al. 2004). The overexpression of GhRDL1 and GhMYB2 increased Arabidopsis seed and leaf hairs by 8–10% (Guan et al. 2011). GhTTG1 and GHTTG3 could recover the trichome phenotype of the Arabidopsis ttg1 mutant (Zhang et al. 2003). Recently, two cotton homologs of an Arabidopsis positive trichome formation regulator (AtML1), GhHD1 and GbML1, were shown to be associated with cotton stem trichome formation (Zhang et al. 2010; Walford et al. 2012). GhHD1, which belongs to the HD-ZIP IV family, plays a vital role in regulating trichomes and fiber initiation as a member of an important protein complex, together with several key MYBMIXTA-like transcription factors (Zhang et al. 2010; Walford et al. 2012; Bedon et al. 2014).

As suggested by its species name, the vast majority of G. hirsutum accessions show pubescent stems and seeds, while G. barbadense varieties generally have glabrous stems and seeds. To date, the gene(s) responsible for these differences have remained unknown. In this study, we identified an insertion of a Ty1-copia retrotransposon into GbHD1 in most G. barbadense varieties that correlated highly with their stem glabrous trichome phenotypes. We also determined that the retrotransposon insertion in HD1 seems to be specific to the A subgenome locus in tetraploid cotton and almost invariably results in the loss of stem trichomes. To the best of our knowledge, this is the first report of a natural transposable element (TE) insertion causing a selectable phenotypic change in cotton.

Materials and Methods

Plant materials and development of mapping populations

Sixty-eight G. barbadense varieties/accessions used in the association study were obtained from the Cotton Research Institute, Chinese Academy of Agricultural Sciences (CRI-CAAS), in Anyang, China, and the U.S. National Plant Germplasm System, in College Station, Texas (kindly provided by James Frelichowski) (Supporting Information, Table S1). The G. barbadense accessions include modern varieties (54/68) and primitive cultivated (14/68) forms (Table S1, primitive cultivated forms); the former were collected from different countries for broad representation of the gene pool and the latter from their native area (South America). In addition, some special germplasm was used, including T586 (an upland cotton genetic marker line with several mutant phenotypes) and TM1 (a genetic standard line). In addition to the materials listed in Table S1, a G. arboreum accession (A2-117) and G. raimondii (D5) also were used for DNA sequencing analysis.

Two mapping populations were developed from crosses between T586 and Pima S6 to map the trichome gene and between T586 and Tu 75-37 to determine the association of the glabrous stem phenotype and retrotransposon insertion.

All cotton materials were grown on the campus farm of Zhejiang A&F University, Lin’an, China, except the primitive cultivated forms, which were grown in pots and kept in the greenhouse during the winter. Plants of each variety/accession were labeled for DNA and RNA extraction as well as for trichome observation.

Marker development and DNA extraction

Genetic marker candidates that are potentially linked to trichome phenotypes were chosen from both published A genome (He et al. 2013) and tetraploid genome mapping studies (Wright et al. 1999; Rong et al. 2007). Candidate RFLP markers were screened for polymorphisms between the parents of the mapping population in this study (T586 and Pima S6), as reported by He et al. (2013). In addition, the genome sequences of G. raimondii (Paterson et al. 2012), spanning 5624.28 kb ranging from 26,943 to 5,651,218 nt [trichome gene (T1) region] of chromosome 10 determined by synteny between genetic maps (Wright et al. 1999; Rong et al. 2007; Desai et al. 2008) and a physical map (Paterson et al. 2012), were used to develop additional DNA markers. Primer design and PCR and SSCP procedures were reported previously (He et al. 2013). In brief, primer design for SSCP marker development was based on the following criteria: the length of the amplified fragments should be 250–600 bp and cover the introns. PCR products were first checked on 1% agarose gels, and those showing one to two clear bands for both genotypes were used for SSCP employing the protocol described by Lee et al. (1992) with some modifications (He et al. 2013).

DNA from different genotypes including the plants of the mapping populations mentioned earlier was extracted as described by Paterson et al. (1993) and slightly modified as follows: about two to three leaf buds were picked from each plant, placed in 2-ml Eppendorf tubes, and homogenized with either an electric drill in a cotton nucleic lysis buffer or in liquid nitrogen. After incubation in 65° for 25–30 min, the supernatant was purified with 24:1 chloroform: isoamylol once. The supernatant was then precipitated with two volumes of 95% ethanol, and the pellets were dissolved in 300 μl H₂O and used as templates for PCR.

Trichome observation and map construction

To construct a genetic linkage map, 124 F₂ plants derived from a cross between T586 and Pima S6 were used. When the plants started to flower, trichomes on the stems were observed and scored as either present or absent (i.e., as a dominant genetic marker). The linkage group was built using MapMaker 3.0, as described by Rong et al. (2004). SSCP markers were pooled together with trichome phenotype data, and linkage was determined at a level of LOD > 3.0.

RNA extraction and subgenome origin assignments of amplicon

Young stems were collected and homogenized in liquid nitrogen. RNA was extracted following the CTAB protocol and purified using Qiagen RNeasy columns (Qiagen, Hilden, Germany). Reverse transcription (complementary DNA synthesis) was performed using the TAKARA PrimeScript 1^st Strand cDNA Synthesis Kit (TAKARA BIO, Inc., Canton, MA). For RT-PCR assay, a pair of primers (SMF7/SMR7) was designed to amplify a 682-bp fragment including a SNP recognized by XmnI to distinguish the A and D subgenome–specific genes. Digestion of PCR products by XmnI will give rise to two fragments of 204 and 478 bp in size, respectively, when only the A subgenome–specific GbHD1 is expressed. However, when only the D subgenome–specific GbHD1 is expressed, the PCR products will not be cleaved by XmnI, and only a single 682-bp fragment will be detected. When both A and D subgenome–specific GbHD1 genes are expressed, DNA fragments of all three sizes will be present.

Cloning of the HD1 gene, restriction digestion, and sequence analysis

Primer design:

A series of primers was designed using DNA sequences around the HD1 gene ranging over 15,362 bp (from 5,518,596 to 5,532,963 bp) of chromosome 10 of G. raimondii as reference for PCR amplification (Table S2 and Figure S1). The primers can be divided into four groups according to their purpose. Group 1 has only one pair of primers, HDZIP1F/HDZIP2R, that was used to amplify the entire HD1 gene. Group 2 consists of At and Dt subgenome–specific primers, including four forward (AP, AGF, DP, and DGF) and two reverse (AGR and DGR) primers that were used to amplify a fragment from a specific subgenome. Group 3 primers were designed to amplify a fragment including a restriction site and used to distinguish the genomic identity of PCR products. Group 4 primers were used to delimit the site where the DNA change occurred in the At subgenome HD1 gene that resulted in no amplification of PCR product.

PCR reaction:

Each PCR reaction generally contains 50 ng of cotton genomic DNA, 10X buffer (Mg²⁺ Plus), 2.5 mM deoxynucleotide triphosphates (dNTPs), 5 μM of each primer, and 0.75 U of Taq DNA Polymerase (TAKARA BIO) in a total volume of 15 μl. PCR reactions were typically run using the following program: predegenerate at 94° for 5 min, then run the cycle of denature at 94° for 45 sec, annealing at 52–58° for 30 sec, and extension at 72° for 1 min for 30 cycles, followed by a final extension at 72° for 7 min. This program may be modified according to the size of the PCR products. For example, TAKARA LA Taq was used to amplify fragments of over 3 kb in size using the primer pair HDZIP1F/HDZIP2R, for which a 3-min extension time was used. When attempting to amplify a fragment with the TE insertion using the subgenome-specific primer such as AF2 and DF2, TAKARA LA Taq with GC buffer was applied, and a 3-min extension time also was used.

Sequencing of PCR products:

For sequencing, PCR products were first checked on agarose gels, and the targeted bands were excised. The fragments were purified using an AxyPrep DNA Gel Extraction Kit (Axygen Biosciences, Union City, CA) and cloned into a TA cloning vector to transform Escherichia coli. Positive clones were identified by colony PCR and sent out for sequencing by the Beijing Genomics Institute (BGI). To assess the attribution of A and D subgenomes in tetraploid cotton, about 10 clones were first sequenced for T586 (G. hirsutum) and Pima S6 (G. barbadence), together with A2-113 and D5 as progenitor genome controls using the primer pair of HDZIP1F/HDZIP2R (Figure 2 and Table S2). DNA sequences from different clones were aligned using Vector NTI Advance 10.0, and sequencing errors were manually inspected to obtain the correct sequence. The cDNA sequences from G. arboreum accessions A2-113 and T586 were used to determine the gene structure (exon and intron).

Figure 2.

Structure of the At-GbHD1 gene showing the position of some important restriction enzyme digestion sites, primers, and transposon insertions. Bold and thin lines represent exons and introns. Numbers are the positions of restriction enzymes, primers, and transposon insertions, indicated in base pairs, with the initiation codon assigned as zero.

Restriction digestion:

To distinguish the A and D subgenomes in a tetraploid cotton, restriction sites were checked according to the preceding alignment. PCR product was cut by the restriction enzyme and checked on 1% agarose gels. The genome attribution of the genes was determined using diploid genomes as control.

Data availability

Table S1 contains detailed descriptions of all used cotton materials. Table S2 contains all of the primers designed for this research. Table S3 and S5 contains all the HD1 related sequencing results for each cotton variety and individual.

Results

Trichome phenotypes in G. barbadense

Six-hundred and fifty-four individual plants from 68 varieties and accessions of G. barbadense (54 modern cultivars and 14 primitive cultivated forms) were observed for their stem trichome phenotypes. Among these, 12 (100 plants) showed pubescent phenotypes, 48 (444 plants) showed glabrous phenotypes, and 8 (110 plants) contained individuals with or without trichomes. The third group could be further classified into two subgroups based on the frequency of plants having trichomes in the progeny population. Six varieties (NLD31, -33, -35, -40, -42, and -47; Table S1) defined the first subgroup, in which only a low frequency [up to 11.3% (or 9/80)] of plants with stem trichomes was observed. The other two varieties (NLD48 and -51; Table S1) defined the second subgroup, in which over 90.0% (or 27/30) of individual plants showed stem trichomes (Table S1).

Close correlation between HD1 and trichome phenotypes revealed by genetic mapping in cultivated tetraploid cottons

The presence or absence of trichomes is a single-gene trait and follows Mendelian inheritance in G. arboreum and G. hirsutum (Wan et al. 2007; Desai et al. 2008). This trait cosegregates with the loss-of-fiber mutant sma-4 (Desai et al. 2008; He et al. 2013). Recently, map-based cloning of sma-4 was conducted by our research group using a large-scale high-resolution fine-mapping population containing 2500+ individuals, which suggested strongly that the underlying causative gene for the sma-4 phenotype is HD1 on chromosome LG A03.

To explore the linkage between the HD1 locus and trichome phenotypes in tetraploid cotton, we developed a mapping population including 124 F₂ plants from a cross between T586 and Pima S6. Using this population, a local linkage map was constructed comprising 10 simple sequence repeat (SSR), 9 single-strand conformation polymorphism (SSCP), and 1 cleaved amplified polymorphic sequence (CAPS) markers for HD1, spanning 49.9 cM, with an average interval of 2.26 cM between markers (Figure 1). The trichome phenotype was found to cosegregate with HD1, a restriction fragment length polymorphism (RFLP)–derived SSCP marker (G1161), and three other SSCP markers developed based on the G. raimondii genome sequence (Paterson et al. 2012), which were flanked by pAR0602 and S34-24 at 0.2 and 0.7 cM, respectively (Figure 1).

Figure 1.

Genetic map of the chromosome fragment around a dominant trichome gene T1 constructed by using an F₂ mapping population of T586 × Pima S6. HD1 was mapped as a CAPS marker (PCR fragment was amplified using the primers UP2/Down2 and digestion by Bstz171I). The loci with lines below the name are SSCP markers, those with stars were developed from the RFLP map (Rong et al. 2004), and others are from the G. raimondii genome sequence (Paterson et al. 2012).

Sequence characterization of the HD1 gene in cultivated tetraploid cottons

A pair of primers (HDZIP1F/HDZIP2R, group 1) was designed to amplify the entire coding region (from the start codon to the stop codon) of HD1 orthologs from both subgenomes for sequencing to study the allelic variations of HD1 in G. barbadense (Figure 2 and Figure 3, Figure S1, and Table S2). A total of 185 clones from 18 plants of 15 G. barbadense varieties, including six primitive cultivated forms, was sequenced (Table S3). Sequencing results confirmed the identity of the HD1 gene, which consists of 10 exons and 9 introns, by the presence of basic HD domains in the predicted protein (Zhang et al. 2010) (Figure 2 and Figure 3). The A genome gene amplicon is 3105 bp long, while the D genome amplicon is 3104 bp long, showing 97.6% identity with 16 predicted amino acid changes. Sequence alignment revealed that all G. barbadense varieties fall into three groups: the first group of varieties retained HD1 orthologs from both At and Dt subgenomes. L-7009 (numbered as NLD52) is an example of this group. The second group comprises varieties whose At-HD1 is missing, and only Dt-HD1 could be detected (Table S3). Examples of this group are 9078N (NLD75-1), Pima S6, 25686 (NLD45-1), and Gb-Yumian. The third group showed a mixed result among individual plants from the same variety, some resembling group 1 with both homologs intact and others resembling group 2 containing only the Dt homolog. For example, NLD51-1 contains only the Dt homolog, while NLD51-2 contains both copies. Interestingly, no plants containing only the At homolog were identified.

Figure 3.

Sequence alignment of At-GbHD1 and Dt-GbHD1 genes showing the gene structure and positions of important primers and restriction enzyme digestion sites. Sequences highlighted in black and gray are exons and introns, respectively. Lines above and below the sequences indicate the primer sequences, with their names listed above and below these lines. Arrows indicate the positions of transposon insertions.

Association between trichome growth and an intact HD1 gene on the A subgenome

The presence and absence of stem trichomes (pubescent phenotype) perfectly correlated with the presence and absence of At-HD1 in the 15 G. barbadense varieties studied (Table S4). In all plants examined, among both uniform and segregating varieties, all plants having trichomes also had both At-HD1 and Dt-HD1 orthologs, whereas plants having no trichomes lacked a copy of At-HD1.

To further verify this correlation, we extended the study of the retention patterns of At-HD1 genes to a larger population of 116 plants from 52 G. barbadense varieties using a homeo-SNP (SNP between genomes, T/C) at a position 1851 bp downstream of the start codon. This site can be used to distinguish At- and Dt-HD1 in tetraploid cottons by digestion with the restriction enzyme Bstz17I (Figure 2 and Figure 3). As expected, the amplified sequences of At-HD1 (primer: UP2/Down2) (Figure 3 and Figure S2) could be cut into two fragments with expected sizes of 985 and 315 bp, respectively, while the Dt-HD1 amplicons remained intact, with the expected size of 1300 bp (Figure 4).

Figure 4.

Different digestion pattern of fragments amplified from different cotton species using primers UP2/Down2 and digestion with Bstz17I. Ga, G. arboreum (diploid A genome); Gr, G. raimondii (diploid D genome); Gh, G. hirsutum (AD1), lanes 3–7; Gb, G. barbadense (AD2), lanes 9–23.

Exploring the sequencing and enzyme-cutting results indicated that among 51 G. barbadense varieties, including the 12 primitive cultivated forms for which trichomes have been scored, 10 contained both At- and Dt-HD1 orthologs (like group 1), 33 exhibited only the Dt-HD1 allele (like group 2), and 8 showed mixed patterns of loss of At-HD1 among individuals (like group 3) (Tables 1 and 2). Again, there was a perfect correlation between the presence of trichomes and At-HD1, lending further support to the hypothesis that lack of At-HD1 is highly related to lack of stem pubescence in G. barbadense.

Table 1. Relationship of trichome phenotypes to At-HD1 genotypes in G. barbadense.

At-HD1	Varietya	No trichomes	With trichomes
No	33 (351)	33 (351)	0(0)
Yes	10(92)	0(0)	10(92)
No/yesb	8(110)	8(74)	8(36)
Total	51(553)	41(425)	18(128)

^aThe numbers in parentheses are the total plant number studied.

^bVarieties showing segregation of the HD1 gene on the A genome [with and without the HD1 gene on chromosome 6(A) and trichome growth (with and without stem trichomes].

Table 2. Relationship of trichome phenotypes to segregation of the At-HD1 gene in G. barbadense varieties.

At-HD1	Varietya	No trichomes	With trichomes
No	8(74)	8(74)	0(0)
Yes	8(36)	0(0)	8(36)

^aThe numbers in parentheses are the total plant number studied.

Lack of PCR amplification of At-HD1 in glabrous plants is caused by Ty1 LTR-retrotransposon insertions

To further investigate the reason for the lack of amplification of At-HD1 in glabrous G. barbadense varieties, a series of overlapping primer pairs (group 4) was designed from both the upstream and downstream regions of the HD1 gene for PCR (Figure S1 and Table S2). Using five pairs of primers (SM1–5) in the promoter regions, the PCR results suggested that the maximum left border of the missing amplification of At-HD1 was near the junction between the At-HD1 promoter and coding regions. In this experiment, we sequenced 92 clones of the promoter region, with a length of 2006–2036 bp upstream from the start codon of 4 G. hirsutum and 10 G. barbadense varieties using the primers sm-pro-F/R (Table S5). These clones could be divided into two groups based on their sequence similarities to either the diploid At-HD1 or Dt-HD1 gene, respectively, using the sequences of eight clones from A2-47 and G. raimondii (D5) genome sequences as references (Table S3). A deleted 15-bp DNA fragment in the Dt-HD1 promoter region, which could easily distinguish At and Dt copies, was used to design forward primers specific to the At copy (named as AP, groups 2 and 4) compared with the Dt copy (designed based on the sequences flanking the specific deleted region and named DP) (Figure 3 and Figure S1 and Figure S2). A series of reverse primers (group 4) also was designed based on the downstream regions of HD1 (Table S2). When the reverse primers were used with UP2-rvs, a fragment of the same size was amplified in varieties both with and without trichomes (Figure 5A), indicating the presence of HD1 on both chromosome 6(A) and chromosome 25(D). However, when the reverse primer was moved down to Down2 or HDZIP2R, a fragment of ∼8 kb was amplified in the varieties without trichomes (and without amplification of At-HD1 under normal PCR conditions), but in the varieties with trichomes (with amplification of At-HD1 under normal PCR conditions), an expected fragment was amplified (Figure 5B). These results suggested that a DNA fragment was inserted into the region between primers UP2 and Down2 (Figure 2) causing failure to amplify the At copy with primers UP2/Down2 and HDZIP1F/HDZIP2R in the initial PCR reactions. Further sequencing results revealed an insertion of a 4999-bp fragment derived from a Ty1-copia LTR retrotransposon at a position 2565 bp downstream from the HD1 start codon. All components of a typical Ty1-copia element, such as long terminal repeats (LTRs), gag and pol, were readily identifiable in the insertion sequence (accession no. KF740825) (Cao et al. 2015).

Figure 5.

PCR amplification of GbHD1 genomic DNA fragments of different sizes in the G. barbadense varieties with glabrous or pubescent stems. A. All varieties (lanes 1–10), whether they have glabrous or pubescent stems, produced a PCR amplicon with the same size using AP/SMR7 primers. B. Varieties (lanes 1–5) with pubescent stems produced a 3007-bp PCR fragment; varieties (lanes 6–10) with glabrous stems produced a PCR fragment of ∼8 kb using AP/Down2 primers.

To reveal whether all G. barbadense plants lacking At-HD1 amplification have the same Ty1 LTR-retrotransposon insertion (hereafter called the TE insertion), three pairs of primers (AGF/AGR, AGF/LTR-R, and LTR-L/AGR) were used. One was a pair of At-genome-specific primers that flanked the insertion point closely and could amplify a 261-bp fragment in HD1 without TE insertion or a 5260-bp fragment including the 4999-bp insertion and 261 bp of the gene (Figure 6). The other two pairs of primers were a combination of the preceding forward primer and the left LTR primer (LTR-R, designed based on the end sequence of LTR), as well as the preceding reverse primer and right (forward) LTR primer. All modern G. barbadense varieties were found to carry a fragment of 5260 or 261 bp or both (heterozygote) when the AGF/AGR primer pair was used (Figure 6) or a strong band of 5092 bp and a weak band of 471 bp when AGF/LTR-R and LTR-L/AGR were used. This confirms that the modern varieties with glabrous stems have the same LTR retrotransposon at the same location in At-HD1 (Table S1).

Figure 6.

PCR amplification of fragments with (e.g., lanes 1–4) and without DNA insertion (e.g., lanes 5 and 19), as well as both (heterozygous, lanes 8 and 17) in G. barbadense varieties collected from Anyang, Henan, China, using AGF/AGR primers.

To further illustrate the correlation between glabrous stem and TE insertion in At-HD1, the segregation of trichome phenotype and the TE insertion were analyzed in a population of 96 F₂ plants derived from a cross between T586 and Tu 75-37 (a G. barbadense variety with glabrous stem and TE insertion). Twenty-seven plants were found with no stem trichomes, like Tu 75-37, and the remaining 69 plants had varied numbers of trichomes, with the ratio of plants with and without trichomes being not significantly different from the 3:1 segregation of a single gene. The TE insertion was found to perfectly correlate with trichome growth; i.e., all 27 plants with glabrous stems had the TE insertion, and all 69 plants having trichomes on their stems either did not have TE insertions (33 plants) or were heterozygous for TE insertion (36 plants) (Figure S2). This result indicated that trichome initiation is perfectly associated with the At-HD1 gene and that the pubescent stem trait is dominant to the glabrous stem.

Ty1 transposon insertion occurred repeatedly in the At-HD1 gene during history

The timing of the TE insertion in At-HD1 of G. barbadense was investigated by identifying both the trichome phenotype and At-HD1 of 14 primitive cultivated cottons originating from Central and South America (Table S1). All these varieties had abundant vegetative growth with many branches and strong main stems 2–2.5 m high. Only three of them bloomed—but almost 2 months later than day-neutral cultivars in this temperate latitude; they also set few bolls. Three accessions (GB81, GB413, and GB656) had many trichomes during the seedling stage, while two others (GB413 and GB1600) grew smaller numbers of trichomes around 2 months after planting. The other nine accessions had few (GB102) or no trichomes.

The preceding HD1 gene primers were applied to study the At-HD1 gene in these primitive varieties. PCR products from 7 of 14 primitive cultivated forms were sequenced and digested by restriction enzymes. As found in modern varieties, the trichome phenotype in the primitive cultivated forms was closely association with the amplification of At-HD1 (Table S1). Enzyme digestion and sequencing of PCR products with primers of HDZIP1F/HDZIP2R and UP2/Down2 indicated that 6 of 14 (43%) primitive cultivated forms have both the At and Dt copies of HD1, which represented a much higher abundance of At-HD1 than that observed in modern G. barbadense varieties (∼5%). The promoters of the two accessions (GB217 with insertion and GB656 without insertion) also were sequenced, and the sequences of both At and Dt subgenomes were the same as those in modern varieties (Table S3). When primer AP/Down2 was used, a fragment of ∼8 kb also was amplified in nine accessions that did not show normal amplification of At-HD1 (Table S1), indicating that a TE insertion probably also occurred in the primitive cultivated forms of G. barbadense (Figure 7). However, the sizes of the larger fragments were slightly different among different accessions, indicating that they were probably different insertions. Using primers AGF/AGR, which are specific to identify the Ty1 LTR-retrotransposon insertion that is ubiquitous among modern G. barbadense cultivars, only three accessions (GB333, GB372, and GB1596) had the same retrotransposon at the same site as modern varieties. Thus, primitive forms appear to have experienced multiple independent insertions into At-HD1, only a subset of which has been carried over to modern varieties. When AP/CK-fan primers were used, a 6663-bp fragment was amplified from the six accessions lacking At-HD1 but that do not have the Ty1 LTR-retrotransposon insertion that is ubiquitous among modern G. barbadense cultivars (Figure 7). Finally, a TE insertion event was confirmed to have occurred in the sixth exon, 1609 bp downstream from the start codon. Sequencing of this DNA insertion indicated that the transposon is almost the same as that isolated from the other cotton varieties, except that it lacked 516 bp, including the right LTR, and had an additional 34 bp on the right side of the TE (Figure S3). Based on the sequence divergence between the two LTRs, we deduced that this TE insertion occurred earlier than the one at the ninth exon in the modern varieties. Consequently, we named the earlier event TE1 and the later event TE2.

Figure 7.

Detection of two retrotransposon insertions (TE1 and TE2) in primitive cultivated G. barbadense forms by PCR amplification using different primer combinations. Lanes 1, 2, and 3 represent different cotton varieties. 1: GB333; 2: GB217; and 3: GB102.

Defective HD1 gene expression is caused by the retrotransposon insertion

Two G. barbadense accessions with pubescent stems and four with glabrous stems were chosen for RT-PCR analysis. Accessions that have stem trichomes (without TE insertion) showed three bands after digestion, and accessions with no stem trichomes (with TE insertion) showed only one band, which was the same size as the undigested Dt genome, or this band plus two very faint bands (Figure 8). These results suggested that both the At-HD1 and Dt-HD1 genes are expressed at similar levels in plants with pubescent stems. In plants with glabrous stems, only the Dt-HD1 gene was expressed, At-HD1 being faint or absent. These results indicated that TE insertion into At-HD1 caused defective gene expression, which may have resulted in the inhibition of trichome growth.

Figure 8.

Allelic gene expression of At-HD1 and Dt-HD1 genes before (A) and after (B) digestion with XmnI of the fragments amplified from stems of primitive cultivated forms by RT-PCR using SMF7/HDZIP-3R primers. Lines 1–6 indicate RT-PCR results from samples from GB102, GB656, GB825, GB1063, GB217, and GB154. Lane 7 is the control sample from the genomic DNA of NLD52-1. The housekeeping gene UBQ was used as the internal control to monitor RNA quality and PCR efficiency.

In addition, HD1 gene expression in the stem also was analyzed by qualitative PCR (qPCR). The results indicated that in modern domesticated cottons, NLD26 (with the retrotransposon insertions at the At-HD1) displayed significantly lower (fivefold) gene expression than NLD52 (containing the intact At-HD1 gene) (Figure 8). Similar lower expression patterns were detected in the At-HD1-disrupted primitive cottons, such as GB217 and GB1596, compared with the At-HD1-intact varieties, such as GB656. To further evaluate the effects of blocked gene expression of At-HD1 on the function of downstream members of the signal pathway in regulating trichome initiation in these G. barbadense accessions, the expressions of GbMYB25 and GbHOX3 also were analyzed. As seen in Figure 9, GbMYB25 and GbHOX3 showed parallel gene expression trends to GbHD1 in stems of these G. barbadense accessions; i.e., they were expressed at a lower level in TE-inserted varieties/accessions than in those with intact HD1 genes, no matter whether they were modern domesticated or primitive cottons.

Figure 9.

Quantitative qRT-PCR analysis of the gene expression of GbHD1, GbMYB25, and GbHOX3 in five G. barbadense varieties/accessions.

Discussion

HD1 is critical to stem trichome growth in Gossypium

Trichome initiation in Arabidopsis thaliana is an important model for understanding cell fate and patterning (Marks 1997; Larkin et al. 2003; Marks and Esch 2003; Hulskamp 2004). By contrast, very little is known about the molecular mechanism of trichome growth in cotton. Although some cotton homologs of Arabidopsis trichome genes have been cloned and found to be related to cotton fiber and trichome growth (Yang and Ye 2013), there is no direct evidence that confirms which gene(s) is(are) critical to cotton trichome growth and how the genes regulate trichome growth and development collectively. Classic genetic studies have revealed that cotton stem and leaf trichomes are mainly controlled by a dominant gene (T1) located on chromosome 6(A) in tetraploid cotton (Wright et al. 1999) and its homoeologous chromosome LG A03 of diploid A genome species (Rong et al. 2005; Desai et al. 2008; He et al. 2013). The same mapping result was found in this study using two F₂ populations from the crosses between T586 and two other G. barbadense varieties (Pima S6 and Tu 75-37). In Arabidopsis, ATML1, encoding a putative HD-ZIP transcription factor, was confirmed to affect trichome growth together with AtPDF2 (Abe et al. 2003). The cotton HD1 gene, which is the homolog of ATML1, cosegregated with T1 in this study, suggesting that HD1 is a candidate gene for the T1 phenotype. Previously, GhHD1 was postulated to be related to trichome formation and fiber initiation by gene silencing and gene expression (Walford et al. 2012). In the present research, a Ty1-copia retrotransposon was found inserted into At-GbHD1 that blocks its expression. Sequence analysis and trichome phenotyping of a large population of G. barbadense genotypes showed that the trichome phenotype in each G. barbadense genotype was closely correlated with this TE insertion event. Mapping of the glabrous trichome phenotype and the TE insertions in At-HD1 also provided evidence of a close association between TE insertion and trichome-less stems. It is extremely unlikely that different varieties of G. barbadense containing this retrotransposon in the same gene and at the same position also would lack trichomes by chance. Thus, we inferred that in G. barbadense the formation of trichomes largely depended on the presence of an intact GbHD1 gene.

Different LTR retrotransposons in At-HD1 clarify the dispersal phases of G. barbadense

LTR retrotransposons, characterized by the presence of LTRs at the ends of elements such as Ty1-copia and Gypsy groups, are widely distributed in plant genomes. In cotton, copia retrotransposon elements have long been known to be widespread in tetraploid and diploid species (VanderWiel et al. 1993). Research in G. barbadense indicated that Ty1-copia retrotransposons are quite heterogeneous and could be grouped into 11 distinct families (Zaki 2005). The activation of retrotransposons, particularly the Gypsy-like retrotransposons, is suggested to be one of the major forces driving genome size variation of diploid cotton species (Hawkins et al. 2009). Also, it is thought that the activity of retrotransposons could drive cotton speciation by inserting into important genes that would result in dramatic changes in important evolutionary or agronomic traits (This et al. 2007; Tsuchiya and Eulgem 2013). Until now, direct evidence of cotton retrotransposons affecting phenotypic evolution had not been reported.

When we checked the geographic locations where these randomly selected primitive cultivated forms originated, we found that the accessions with different TE insertions were not randomly distributed (Figure 10). First, among 13 accessions whose places of origin could be traced, 5 originated from Peru, including 2 accessions with intact At-HD1 and Dt-HD1 genes (named as AD), 2 with TE2 (2565 bp downstream from ATG), and 1 with TE1 (1609 bp downstream from ATG). Two were from Brazil, and the other 6 were from six different countries (Table S1). Thus, Peru has more genetic diversity than other places and contains all kinds of primitive cultivated forms. This result strongly supports the hypothesis that Peru is the origin of modern G. barbadense, which was derived from early domesticated forms (Hutchinson et al. 1947). Second, the geographic distribution is much more dispersed for TE1, with six accessions scattered in Cuba, Guyana, Brazil, Argentina, and Peru, while accessions with the TE2 insertion originated from a neighboring region of Peru and Colombia. Using allozyme analysis, Percy and Wendel (1990) suggested a dispersal route of G. barbadense from west of the Andes, a primary domestication site, to northern South America (primary dispersal), Central America (secondary-stage dispersal), and a region including Argentina and Paraguay (a post-Columbian dispersal) (Hutchinson 1962; Brubaker et al. 1999; Wendel et al. 2010). Although our results could not further distinguish the three dispersal stages or reveal a detailed dispersal route, as described earlier, the distribution of accessions with different TE insertions suggests that the dispersal processes proposed by Percy and Wendel (1990) happened between the TE1 and TE2 insertion events. In addition, the diffusion of accessions with TE1 was limited to an area including South America, Central America, and northern North America but did not spread worldwide. Meanwhile, the TE2 insertion occurred quite recently and spread along the trans-Andean route in northern South America (through Colombia) into Central America, the Caribbean, and the Pacific (Percy and Wendel 1990) and then was further diffused to Egypt, forming “Egyptian cotton,” and to the southern United States, forming “Pima cotton” (Figure 10). Therefore, the two TE insertions found in this research represent important time points dividing the two dispersal phases in diffusion of G. barbadense.

Figure 10.

Dispersal route of primitive cultivated G. barbadense forms inferred by retrotransposon insertion events. Green arrows indicated the probable dispersal route of G. barbadense accessions during the stage between the TE1 and TE2 insertion events. Red arrows indicate the diffusion pathway of G. barbadense accessions after the TE2 insertion. A blue dot represents the point of origin of G. barbadense in Peru.

One notable difference between the two major cultivated cotton species is their stem trichome phenotypes: G. hirsutum are almost all pubescent, while G. barbadense are almost all glabrous. We now know that this difference is closely associated with the disruption of the At-HD1 gene in G. barbadense. Since HD1 is also involved in fiber development (Walford et al. 2012), we speculated that this insertion could be one of the main reasons why most G. barbadense varieties have little or no fuzz fiber. The association between the TE insertion, the glabrous stem, and fuzz-less seeds is being investigated in our laboratory. The initial results suggest that the growth of fuzz fiber is also affected by the GbHD1 gene. This might explain why more modern varieties have more TE insertions than primitive cultivated forms, because linter fiber is more easily detached from smooth-seeded varieties, making it favored by hand ginning in artificial selection (Hutchinson 1962; Percy and Wendel 1990; Brubaker et al. 1999; Wendel et al. 2010).

G. barbadense varieties carrying a TE insertion in At-HD1 are a valuable resource to study molecular mechanisms related to cotton trichome and fiber development

Forward genetics has been explored extensively to study molecular mechanisms controlling plant phenotypes, growth habits, and development. Some natural retroelement insertions in genes have been identified and used to understand the function of related genes (Hori et al. 2007). One of them, a retroelement insertion in the Medicago FLOWERING LOCUS T (MtFTa1) gene helped to determine its function and decipher possible related pathways that regulate flowering time (Jaudal et al. 2013). In cotton, such progress remains limited. This might be because fewer useful mutant genes have been cloned compared with model plants such as Arabidopsis and rice. The varieties with LTR-retrotransposon insertion in the At-GbHD1 gene identified in this study will be very powerful resources to study molecular mechanisms related to epidermal cell growth, similar to Arabidopsis mutants.

Real-time PCR results revealed that the retrotransposon insertion in At-HD1 caused significant down-regulation in the expression of both At-HD1 and Dt-HD1 genes. In G. barbadense, GbML1 (GbHD1) could bind to its own promoter region, leading to positive-feedback regulation of its own gene expression (Zhang et al. 2010). Our gene expression analysis partially confirmed this observation: the insertion in At-HD1 caused significant reduction in the expression of both alleles of HD1. GhMYB25 belongs to a MIXTA type MYB transcription factor family and controls cotton fiber initiation. GhHOX3, a homologous gene of Arabidopsis HDG11/12, regulates plant epidermal cell differentiation (Nakamura et al. 2006; Machado et al. 2009). Significantly reduced gene expression of GbMYB25 and GbHOX3 in the TE-inserted cottons provided further evidence that GhHD1 is a critical upstream gene of GhMYB25 (Bedon et al. 2014) and GbHOX3. Taken together, these results suggest that At-HD1 is an important molecular switch that can activate GhMYB25 and GhHOX3 or that these three proteins may collaborate to regulate epidermal cell development into trichomes (Bedon et al. 2014).