Up-regulation of a homeodomain-leucine zipper gene HD-1 contributes to trichome initiation and development in cotton

NIU Er-li, CAI Cai-ping, BAO Jiang-hao, WU Shuang, ZHAO Liang, GUO Wang-zhen

State Key Laboratory of Crop Genetics & Germplasm Enhancement, Ministry of Science and Technology/Hybrid Cotton R&D Engineering Research Center, Ministry of Education/College of Agriculture, Nanjing Agricultural University, Nanjing 210095, P.R.China

Abstract Plant trichomes originate from epidermal cells. In this work, we demonstrated that a homeodomain-leucine zipper (HD-Zip)gene, Gh_A06G1283 (GhHD-1A), was related to the leaf trichome trait in allotetraploid cotton and could be a candidate gene for the T1 locus. The ortholog of GhHD-1A in the hairless accession Gossypium barbadense cv. Hai7124 was interrupted by a long terminal repeat (LTR) retrotransposon, while GhHD-1A worked well in the hairy accession Gossypium hirsutum acc.T586. Sequence and phylogenetic analysis showed that GhHD-1A belonged to the HD-Zip IV gene family, which mainly regulated epidermis hair development in plants. Silencing of GhHD-1A and its homoeologs GhHD-1D in allotetraploid T586 and Hai7124 could significantly reduce the density of leaf hairs and affect the expression levels of other genes related to leaf trichome formation. Further analysis found that GhHD-1A mainly regulated trichome initiation on the upper epidermal hairs of leaves in cotton, while the up-regulated expression of GhHD-1A in different organs/tissues also altered epidermal trichome development. This study not only helps to unravel the important roles of GhHD-1A in regulating trichome initiation in cotton, but also provides a reference for exploring the different forms of trichome development in plants.

Keywords: leaf trichome, map-based cloning, a homeodomain-leucine zipper gene HD-1, virus-induced gene silencing(VIGS), functional differentiation

1. Introduction

Plant trichomes originate from epidermal cells and vary widely in their length, density, and distribution. Hairless stems/leaves are potential defenses against oviposition by insects such as white flies, aphids, and boll weevils.However, high densities of trichomes are bene ficial against insect attack (Thomsonet al. 1987; Meredithet al. 1996).Trichomes on the epidermis of ovules in cotton are the main source of fibers for the textile industry and have important economic value (Basara and Malik 1984; Kohel and Leuis 1984). Studying trichome formation on the epidermis of different tissues could help us to elucidate the mechanism of trichome formation and improve trichome traits. The density of trichomes in cotton (Gossypiumspp.) varies from densely pubescent to entirely glabrous (Lee 1968, 1985).Lee (1985) carried out a detailed genetic study and identi fied five major loci related to trichome formation, T1–T5. Of these, the T1gene located on Chr. 06 of the A-subgenome in allotetraploid cotton, was a major locus/gene determining leaf pubescence (Wrightet al. 1999; Lacape and Nguyen 2005). Multiple candidate quantitative trait loci (QTLs)affecting lint percentage, fiber length, fiber length uniformity,fiber strength, and spiny bollworm resistance have also been mapped in the T1region (Guoet al. 2006; Wanet al. 2007).Recently, Dinget al. (2015) identi fied that the insertion of a retrotransposon into the ninth exon ofHD1Awas related to the glabrous stem trait inGossypium barbadensePima S6. By studying the sequences ofHD1Ain differentG.barbadenseaccessions, Dinget al. (2015) also found another retrotransposon insertion in the sixth exon ofHD1Ain several hairlessG.barbadenseaccessions, which might have occurred earlier than the one at the ninth exon in the modern varieties. However, there still exist some hairless and hairy accessions with the normal A- and D-subgenome copies ofHD-1, therefore, the functional role of this gene requires further exploration (Dinget al. 2015).

The epidermal hair inArabidopsisis a single cell structure and can be used as a good model system to study cell cycle regulation, cell polarity, and cell proliferation (Hulskampet al. 1994). Several key transcription factors and regulators involved in trichome formation have been clearly identi fied(Balkundeet al. 2010). The Glabrous1-Glabra3-Transparent Testa Glabra 1 complex(GL1-GL3-TTG1) promotes the expression of Glabrous2(GL2) and the negative factors,Triptychon/Caprice/Enhancer of TRY and CPC 1/Enhancer of TRY and CPC 2 (TRY/CPC/ETC1/ETC2), and then the TRY/CPC/ETC1/ETC2 protein moves into the neighboring cells and competes with GL1 for GL3 protein binding sites(Balkundeet al. 2010). Moreover, the ectopic expression of cotton genes can also promote the occurrence of trichomes inArabidopsis.GaMYB2(anArabidopsisGL1homolog)rescued trichome formation in theArabidopsisgl1mutant(Wanget al. 2004). Two homologs ofArabidopsis TTG1in cotton,GhTTG1andGhTTG3, were also able to restore trichome formation inArabidopsis ttg1mutant plants(Humphrieset al. 2005).GaHOX1(anArabidopsisGL2homolog) rescued trichome development in a glabrousgl2-2 Arabidopsismutant through control of theGL2promoter(Guanet al. 2008). BothAtGL2andGaHOX1belong to the homeodomain-leucine zipper (HD-Zip) IV gene family,and are expressed in the outer cells of plant organs and regulate the differentiation of epithelial cells, the formation of trichomes, and the accumulation of anthocyanins(Nakamuraet al. 2006).

Recently, genome sequences of the diploid cottonsGossypium raimondii(Patersonet al. 2012) andGossypium arboreum(Liet al. 2014), and the tetraploid cottons,G.hirsutum(Liet al. 2015; Zhanget al. 2015) andG.barbadense(Liuet al. 2015; Yuanet al. 2015) have been published, making map-based cloning of target genes of interest faster and more ef ficient. Map-based cloning and the discovery of variation in multiple sites inGhOKRAhas revealed that a homeodomain leucine-zipper class I protein caused the okra leaf phenotype (Changet al. 2016; Zhuet al.2016). A premature stop codon inGoPGF, a basic helixloop-helix domain-containing transcription factor, was found in a duplicate recessive glandless mutant (Maet al. 2016).Here, utilizing the BC1genetic population generated from the crossing of two cultivated allotetraploid cotton accessions,hairlessG.barbadensecv. Hai7124 and hairyG.hirsutumacc. T586, and map-based cloning, we demonstrated that a HD-Zip geneGhHD-1A, belonging to the HD-Zip IV gene family, was a candidate gene affecting the leaf trichome trait. We veri fied that a retrotransposon insertion ofGhHD-1Aortholog in Hai7124 led to a hairless phenotype on the upper epidermis of the leaf, while the upregulated expression ofGhHD-1Ain different organs/tissues was associated with epidermal trichome initiation in cotton. The study not only reveals thatGhHD-1Ais an important component in the regulation of leaf trichome initiation in cotton, but also provides valuable information for elucidating the mechanism of trichome formation in other plant tissues.

2. Materials and methods

2.1. Plant materials

Two cultivated allotetraploid cotton accessions, hairlessG.barbadensecv. Hai7124 and hairyG. hirsutumacc. T586(Kohel and Leuis 1984) were used as parental lines and generated (Hai7124×T586)×Hai7124 BC1generations that comprised 1 065 (545 hairy plants, 520 hairless plants)plants. In addition, three hairyG.hirsutummaterials,72t2-4, 81t3-4, and 121t2-4, were selected for functional con firmation of candidate leaf trichome genes.

Fresh leaf buds were collected for genetic analysis and total DNA was extracted using the cetyl-trimethylammonium-bromide (CTAB) method as described by Patersonet al. (1993). All the DNA samples were quanti fied using “one drop spectrophotometer OD-1000+” (OneDrop,Nanjing, China) and adjusted to a concentration of 20–60 ng μL–1. The second leaves from the top in two-mon-old Hai7124 and T586 plants were sampled for the observation of leaf trichomes and differential expression analysis at the transcriptional level.

All samples were quick-frozen in liquid nitrogen and stored at –70°C before use, and all the materials were planted in the Jiangpu experimental field, Nanjing Agricultural University,Nanjing, Jiangsu Province, China.

2.2. Scanning electron microscopy

To observe the form and density of the cotton epidermal hairs, leaves at different developmental stages including leaf buds, young leaves (the second leaf from the top), and mature leaves (the bottom leaf), were collected from two-mon-old plants and examined by scanning electron microscopy. The leaves were fixed in 4% (v/v) glutaraldehyde and dehydrated in an ethanol series. The samples were then transferred into isoamyl acetate and dried at the critical point. Finally, the morphology of epidermal trichomes was viewed with a Hitachi S-3000N scanning electron microscope (Hitachi, Japan).

2.3. Map-based cloning of leaf trichome genes

Based on the previously published cotton linkage map(Guoet al. 2007) and the physical map of tetraploid cotton TM-1 (Zhanget al. 2015), simple sequence repeat (SSR)and single nucleotide polymorphism (SNP) markers on Chr. A06 were selected to identify candidate leaf trichome genes (Appendix A). PCR reactions were performed in a total volume of 20 μL containing 1.0 μL of 10× buffer,1.0 μL of 2.5 mmol L–1MgCl2, 1.0 μL of 10 mmol L–1dNTPs, 1.0 μL of 10 μmol L–1primers, 1.0 μL of 2.5 U μL–1Taqase(TaKaRa Biotechnology Co., Ltd., Japan), 1.0 μL of 20–60 ng μL–1DNA, and 4.7 μL of ddH2O. The PCR amplification program was as follows: pre-denaturation at 95°C for 5 min, 28 cycles of denaturation at 94°C for 45 s, annealing at 58°C for 45 s, extension at 72°C for 1 min,and a final cycle of 10 min at 72°C. The PCR products were separated using denaturing gel electrophoresis. Linkage analyses and integration of the genetic linkage map and the physical map were then conducted using Join-Map 3.0(Van Ooijen and Voorrips 2001) and MapChart 2.3 (Voorrips 2002), respectively. Finally, the leaf trichome T1locus was further mapped to a 1.3-Mb region between the two markers,S1874 and S1922. By comparing the gene annotation and expression analysis, the DNA sequences of the candidate gene in the two parents Hai7124 and T586 were ampli fied.The SNP marker for the candidate gene was designed to detect whether the genotype and the phenotypes of the BC1population were co-segregated.

2.4. RNA isolation and expression analysis

Total RNAs were isolated with a Biospin Plant Total RNA Extraction Kit (Bioer Technology Co., Ltd., China). The TransScript II QRT SuperMix for qPCR (Vazyme Biotech Co.,Ltd., China) was used to digest gDNA and reverse transcribe RNA samples into cDNA using AMV reverse transcriptase.The cDNAs were stored at –30°C before use.

Reverse transcription-PCR (RT-PCR) and quantitative real-time PCR (qRT-PCR) were conducted for expression analysis. The primers for RT-PCR (Appendix A) were designed using Primer Premier 5 (http://www.Premierbiosoft.com). Each 25 μL reaction mixture contained 0.2 μL ofTaqase, 0.5 μL of dNTPs, 1.0 μL of cDNA (～100 ng μL–1),1.5 μL of MgCl2, 2.5 μL of 10×PCR buffer, 5 μmol L–1of primers, and 17.3 μL of ddH2O. The amplification program was: pre-denaturation at 95°C for 5 min, 35 cycles of denaturation at 94°C for 30 s, annealing at 56°C for 30 s,extension at 72°C for 30 s, and final extension at 72°C for 10 min. The expression level of EF-1α was used as the endogenous control (Jianget al. 2012).

The primers for qRT-PCR (Appendix A) were designed using Beacon Designer 7.91 (http://www.Premierbiosoft.com) and were based on CDS sequences close to 3´UTR(Appendix A). Each 20 μL reaction mixture contained 10 μL of SYBR Green Master (Rox, USA), 1.5 μL of cDNA(～100 ng μL–1), 5 μmol L–1of primers, and 7.5 μL of ddH2O.The amplification program was: pre-denaturation at 95°C for 10 min, 40 cycles of denaturation at 95°C for 15 s, annealing at 60°C for 15 s, and extension at 72°C for 15 s. The relative expression level was calculated using the 2–ΔCTmethod (Livak and Schmittgen 2001) and the expression level of theTubulin 1gene (AF487511.1) was used as the endogenous control.

G.hirsutumacc. TM-1 RNA-sequencing data from vegetative tissues (root, stem, and leaf), ovule tissues on different days post anthesis (DPA) (0, 5, 10, and 20 DPA),and fiber tissues (10 and 20 DPA), were downloaded from http://mascotton.njau.edu.cn/ (Zhanget al. 2015). The fragments per kilobase of exon model per million mapped reads (FPKM) method with Cuf flinks (http://cuf flinks.cbcb.umd.edu/) was used to calculate genes’ expression levels.

2.5. Gene cloning and sequence analysis of the leaf trichome gene

gDNA amplification primers were designed using Primer Premier 5 (http://www.Premierbiosoft.com) and could be found in Appendix A. Vazyme Lamp DNA Polymerase(Vazyme Biotech Co., Ltd., China) was employed in standard PCR analysis following the manufacturer’s instructions.PMD19-T vectors (TaKaRa Biotechnology Co., Ltd., Japan)and DH5α competent cells (Vazyme Biotech Co., Ltd.,China) were also utilized in the DNA sequencing.

The Gene Structure Display Server 2.0 and LTR_FINDER were used for the prediction of the exon/intron structures (Huet al. 2015) and the full-length LTR retrotransposon elements(Xu and Wang 2007). The SMART tool (http://smart.emblheidelberg.de/) was used to predict the conserved motifs of theGhHD-1Agene.

2.6. ldentification of the HD-Zip gene family

The genomic database of the tetraploid cotton TM-1 was downloaded from http://mascotton.njau.edu.cn (Zhanget al. 2015). Data on HD-Zip domains were obtained from Pfam (http://pfam.xfam.org/; Finnet al. 2016) including HD(PF00046), homeobox associated leucine zipper HALZ(PF02183), MEKHLA (PF08670) and START (PF01852).By taking the Pfam data as the query, the HD-Zip members were identi fied using HMMER Software, version 3.0 (Finnet al. 2011). The phylogenetic tree of the HD-Zip gene family was built using MEGA 5.0 Software (www.megasoftware.net) with the maximum likelihood method (Tamuraet al.2011), using the parameters referred to Niuet al. (2015).

2.7. Virus-induced gene silencing (VIGS) assays

To explore the function of the candidate geneGhHD-1A,we suppressed its expression by virus-induced gene silencing (VIGS) assay. A 261-bp fragment ofGhHD-1Afrom T586 was inserted into the pTRV2 VIGS vector with the enzymesXbaI andBamHI (Appendix A).CLA1(KJ123647,Cloroplastos alterados 1), which encoded 1-deoxy-D-xylulose-5-phosphate synthase and when silenced leads to a photo-bleached phenotype in plants, was constructed as pTRV:CLA1and acted as a control to verify the ef ficiency of VIGS. All vectors including the empty vector (TRV::00),the positive controlCLA1-silenced vector TRV::CLA1, and theGhHD-1A-silenced vector (TRV::HD-1), were introduced into theAgrobacteriumstrain GV3101 by electroporation(Bio-Rad, Hercules, CA, USA).

TheAgrobacteriumcolonies were grown at 28°C for 16 h in an antibiotic selection medium containing rifampicin(50 μg mL–1) and kanamycin (100 μg mL–1). When the OD600ofAgrobacteriumcells reached 0.5, the cells were collected and suspended in in filtration medium (10 mmol L–1MgCl2,10 mmol L–12-(N-morpholino)ethanesulfonic acid (MES)and 200 μmol L–1acetosyringone), to produce a final OD of 2.0.Agrobacteriumstrains containing pTRV1 and pTRV2 vectors were mixed at a ratio of 1:1. After being incubated for 3 h at room temperature, the suspension was in filtrated into the mature cotyledons. The plants were then grown in pots at 23°C with 16 h light and at 21°C with 8 h dark per cycle. About 14 days post-agro-in filtration, when the top leaf (the first leaf from the top) of the positive control,TRV::CLA1, displayed chlorophyll degradation, the trichome distribution, and expression ofGhHD-1Aand its orthologs was observed in the corresponding leaves of CK, TRV::00,and TRV::HD-1plants.

3. Results

3.1. Genetic analysis and fine mapping of the leaf trichome trait in cotton

The morphology of the cotton hairs was quite diverse.Utilizing scanning electron microscopy, we observed different forms of epidermal leaf hair in cotton (Fig. 1).As Werker (2000) described, there were two types of epidermal trichome: glandular trichomes (GTs) (Fig. 1-A)and non-glandular trichomes (no-GTs) (Fig. 1-B–I). There were a variety of types of no-GTs, such as the no branch(Fig. 1-B), the simple branch (Fig. 1-C–E), and the multibranch (Fig. 1-F–I), indicating the complexity of trichome development in cotton.

Hai7124 plants generally displayed a smooth surface but displayed hairy leaf veins, leaf margins, and leaf buds,while T586 plants had dense hair over the whole plant(Fig. 2-A). As the leaves expanded, the density of trichomes decreased both in Hai7124 and T586 (Fig. 2-B). We also investigated the phenotypes of leaf trichomes in the BC1population, which consisted of 1 065 plants obtained from crossing Hai7124 and T586. Of these, 545 plants displayed the hairy phenotype and 520 plants displayed the hairless phenotype, which followed Mendelian 1:1 inheritance(χ2=0.5408＜χ20.05,1=3.84). These results con firmed that the leaf trichome phenotype was controlled by a single dominant locus referred to as T1(Lacape and Nguyen 2005). By screening the 1 065 BC1plants with the SSR and SNP markers (Appendix A), the leaf epidermal trichome gene was anchored to Chr. A06 between the two markers S1874 and S1922, with a distance of only 0.4 and 0.2 cM flanking the target locus. With the genome sequence of tetraploid cotton TM-1 (Zhanget al. 2015) as a reference,the molecular marker loci were mapped to the scaffold of A06. As a result, the loci of the markers on the genetic map and the physical map displayed good collinearity, except for NAU5433 (Fig. 2-C). The 1.3 Mb interval between the two markers S1874 and S1922 was selected for further investigation of the leaf trichome gene.

Fig. 1 Scanning electron microscopy of the leaf trichome morphology in cotton. A–I showed the different morphology of epidermal trichomes in hairy T586 plants. The second leaves from the top of plants about 2-mon-old were sampled and viewed with a Hitachi S-3000N scanning electron microscope(Hitachi, Japan).

Fig. 2 Map-based cloning of leaf trichome genes. A, trichome distribution in different tissues of Hai7124 and T586. B, trichome distribution in leaves of different developmental stages in Hai7124 and T586. Young leaves (the second leaf from the top to down)and mature leaves (the first leaf from the down to top) were sampled from two-mon-old plants for scanning electron microscopy analysis. C, fine mapping of the leaf trichome gene. The genetic linkage map (left) and the physical map (right) of cotton Chr. A06 were integrated for mining the candidate leaf trichome gene (T1 locus).

3.2. ldentification and con firmation of the candidate leaf trichome gene

Within the mapping region, 34 putative open reading frames were predicted and assigned to different metabolic pathways(Appendix B). Of these,Gh_A06G1283andGh_A06G1310encoding HD-Zip and MYB proteins respectively, whose homologs had been proven to play an important role in regulating trichome formation inArabidopsis(Balkundeet al.2010), were further investigated. We developed the genespecific, not the subgenome-specific primers for expression detection of candidate genes (Appendix A). The transcripts ofGh_A06G1283homologs were detected both in Hai7124 and T586 by RT-PCR, but the transcripts ofGh_A06G1310homologs were not (Appendix C). Further, qRT-PCR was conducted to compare the expression difference ofGh_A06G1283homologs in these two cotton accessions, which showed a significant reduction in the hairless Hai7124, than that in the hairy T586 (Fig. 3-A). Therefore, we temporally consideredGh_A06G1283to be the candidate gene for regulating the leaf trichome trait.

Subsequently, we ampli fied the genomic DNA ofGh_A06G1283from the diploid cottonsG.raimondiiandG.arboreumand the tetraploid cottonsTM-1, Hai7124, and T586. One of the polymorphic loci in the A-subgenome(T586 At: A; Hai7124 At: C) was selected for SNP marker design to screen the two hairless/hairy phenotypes(Appendix C). As shown in Appendix C, genotyping of the SNP marker was consistent with the trichome phenotypes and no cross-overs were detected between the genotypes and phenotypes in the BC1segregation population. Taken together, these results suggest that the homologous gene ofGh_A06G1283was the gene that caused the differences in leaf trichomes between Hai7124 and T586.

Fig. 3 Con firmation of the candidate leaf trichome gene. A, expression analysis of Gh_A06G1283 in Hai7124 and T586 by qRTPCR. The Tubulin 1 gene (AF487511.1) was used as the endogenous control. The relative expression level was calculated using the 2–ΔCT method (Livak and Schmittgen 2001) and error bars indicated the standard deviation (SD) of three biological replicates.＊＊ denoted the significant difference at the 1% level. B, gene structure of the candidate leaf trichome gene in the hairless plant Hai7124. The exon/intron organization and the main elements of long terminal repeats (LTRs) were identi fied using the Gene Structure Display Server (Hu et al. 2015) and LTR_FINDER (Xu and Wang 2007). TSR, target site repeats; PPT, polypurine tract; RNaseH, ribonuclease H; RT, reverse transcriptase; INT, integrase; Gag, group-specific antigen; UBN2, ubinuclein 2; PBS,primer binding site. C, the motif analysis of leaf trichome gene. The motif analysis was conducted with SMART tool (http://smart.embl-heidelberg.de/). HOX, homeodomain; START, StAR-related lipid-transfer.

3.3. Sequence and phylogenetic analysis of the leaf trichome gene

To clarify the source of the different leaf trichome phenotypes in Hai7124 and T586, we compared the sequences ofGh_A06G1283in the two materials. The sequence from Hai7124 had an extra 4 997 bp-fragment insertion in the ninth exon compared to the sequence from T586 (Fig. 3-B;Appendix D). The insertion fragment displayed elements such as a long terminal repeat (LTR), group-specific antigen(Gag), integrase (INT), reverse transcriptase (RT), and ribonuclease H (RNaseH), and was a typicalTy1-copiaLTR retrotransposon (KF740825). This suggested that aTy1-copiaLTR retrotransposon inserted intoGh_A06G1283and caused the hairless leaf phenotype in Hai7124.

Gh_A06G1283was predicted to encode a protein of 725 amino acids with a theoretical isoelectric point(pI) of 5.77 and a molecular weight (MW) of 79.62 kDa.DNAMAN (http://www.lynnon.com/) analysis showed thatGh_A06G1283was the homologous gene ofGhHD-1A(AFO11041.1), with an amino acid similarity of 99.03%.TheGhHD-1protein displayed the typical homeodomain(HOX) and StAR-related lipid-transfer (START) domain belonging to the HD-Zip IV subfamily (Fig. 3-C; http://smart.embl-heidelberg.de/). We obtained orthologs of all HD-Zip IV inArabidopsis(Henrikssonet al. 2005) to predict their functions in trichome formation. As shown in Appendix E,Gh_A06G1283,Arabidopsis thaliana Meristem Layer1(ATML1; CAB36819.1), andProtodermal Factor 2(PDF2; OAP00951.1) were grouped into a single cluster.BothATML1andPDF2were shown to potentially regulate epidermal cell differentiation inArabidopsis(Luet al. 1996;Zapataet al. 2016), which implied thatGh_A06G1283had a similar role in trichome development in cotton.

3.4. Silencing of GhHD-1A and its homologs GhHD-1D in different cotton accessions

To further characterize theGhHD-1Agene and to distinguish it from the other HD-Zip IV genes, we conducted genomewide identification and phylogenetic analysis of the HD-Zip gene family in the tetraploid cottonTM-1. All HD-Zip members shared the HD and LZ domains, but the domains downstream of the LZ domain differed. The members of the HD-Zip II subgroup had a conserved CPSCE motif while the members of HD-Zip I did not (Ciarbelliet al. 2008). The members of both HD-Zip III and HD-Zip IV shared a common START domain followed by an HD-START-associated domain (HD-SAD) (Henrikssonet al. 2005). However, the HD-Zip III members contained an additional C-terminal MEKHLA domain compared to the members of HD-Zip IV.Using Pfam data (Homeodomain HD: PF00046; Homeobox associated leucine zipper HALZ: PF02183; MEKHLA:PF08670 and START: PF01852), we identi fied 137 HD-Zip genes in TM-1,with 25 of these in the HD-Zip IV subgroup(Appendix F). By aligning the sequences with 25 HD-Zip IV genes, a specific cDNA fragment (99 bp before the start codon “ATG” and 159 bp after the start codon “ATG”) of theGhHD-1Agene was identi fied and introduced into the TRV array (TRV::HD-1) to suppress the expression of theGhHD-1Ain the hairy plants T586 (Appendix G). Due to the 97.8% nucleotide sequence identity ofGhHD-1homoeologs inG.hirsutumacc. T586 (designatedGhHD-1AandGhHD-1D), silencing of the TRV::HD-1was specific toGhHD-1but not to the A or D sub-genome. As shown in Fig. 4-A, when the leaves of the positive control, TRV::CLA1,displayed the photo-bleached phenotype about two weeks post-agro-in filtration, theHD-1-silenced plants (TRV::HD-1)exhibited less hair or a hairless phenotype on the epidermis of leaves and buds, compared to untreated control (CK)plants and empty vector-silenced (TRV::00) plants. In addition, the abundance ofGhHD-1transcripts in the TRV::HD-1leaves was also significantly lower than that in the CK and TRV::00 leaves (Fig. 4-B), indicating that theGhHD-1homeologs were effectively silenced in the TRV assay.

To con firm the reliability of the results, we also suppressed the expression ofGhHD-1in three hairyG.hirsutummaterials (72t2-4, 81t3-4, and 121t2-4). When the expression levels ofGhHD-1were reduced, the density of hairs on the leaves also decreased significantly, until the leaf epidermis was almost smooth (Fig. 4-C and D). These data further support a role forGhHD-1in regulating trichome formation in cotton.

Since Hai7124 was allotetraploid cotton, only A subgenome ofGhHD-1ortholog (GbHD-1A) was interrupted by a LTR retrotransposon. Therefore, we also suppressed the transcription of A and D sub-genomes ofGbHD-1in Hai7124 to explore the potential role of D sub-genome ofGbHD-1(Fig. 5). After silencingGbHD-1in Hai7124, the new leaves displayed a relatively smooth phenotype on the lower epidermis and buds. These results implied that the D sub-genome ofGbHD-1could also regulate trichome initiation mainly on the lower epidermis of leaves and new buds.

4. Discussion

4.1. Mutation of GhHD-1A caused the hairless leaf epidermis in Hai7124

Various candidate genes affecting trichome development have been revealed inArabidopsisand generally fall into two classes: positive and negative regulators (Balkundeet al.2010). The protein complex comprisingTTG1(a WD40 protein),GL1/MYB23(R2R3 MYB-related transcription factor), andGL3/EGL3(a bHLH gene) can regulateGL2(a HD-Zip IV protein) and further promote trichome initiation.The negative regulatorTRY/CPC(a R3 MYB gene), moves into neighboring cells and competes withGL1for binding toGL3, repressing trichome initiation immediately next to the trichome cell (Szymanskiet al. 1988; Ishidaet al. 2007).The candidate leaf trichome genein allotetraploid cottonGhHD-1A,was identi fied as the orthologous gene ofGL2and a mutation in theGhHD-1Aortholog in Hai7124 led to hairless leaves (Fig. 2). To characterize the other genes related to trichome formation, we retrieved their orthologous genes in cotton using the BLASTp method and detected their expression in Hai7124 and T586 (Appendix H).Despite the fact that the expression ofGL1had no significant changes, the positive factorsGL3andTTG1displayed higher expression in the hairy plant T586 than in Hai7124, while the negative factorsTRYandCPChad lower expression in T586 than in Hai7124. It seemed that mutations inGhHD-1Aalso affected the expression patterns of the other genes and these created a feedback process in the molecular regulation of trichome formation.

4.2. Sub-genomes of GhHD-1 in allotetraploid cotton might regulate trichome initiation on the leaf epidermis

Fig. 4 Silencing of HD-1 gene in hairy cotton accessions via virus-induced gene silencing (VIGS) assay. A and C, silencing of the GhHD-1 gene in hairy Gossypium hirsutum materials from T586 (A) and 72t2-4, 81t3-4, 121t2-4 (C) via TRV-VIGS assay. The labels CK, TRV::00, TRV::CLA1, and TRV::HD-1 indicated the untreated, the empty vector-silenced, positive control CLA1-silenced(KJ123647), and the HD-1-silenced plants, respectively. Two weeks post-agro-in filtration, the top leaf (the first leaf from the top of the plant) was sampled to observe trichome initiation. B and D, comparison of expression differences after silencing of the GhHD-1 gene. The Tubulin 1 gene (AF487511.1) was used as the endogenous control. The relative expression level was calculated using the 2–ΔCT method (Livak and Schmittgen 2001) and error bars indicated the standard deviation (SD) of three biological replicates.＊＊, the significant difference at the 1% level.

In allotetraploid cotton,HD-1gene had four copies. Hairy accession T586 with four copies ofHD-1gene displayed dense hair over the whole plant hairless accession, while hairlessHai7124 with two copies ofHD-1gene displayed a relatively smooth epidermis on its leaves. However,there still existed some trichomes on the lower epidermis of leaves and new buds (Fig. 2-A). After silencingGbHD-1in Hai7124, the trichomes on lower leaf epidermis and buds almost disappeared (Fig. 5). It could be speculated that the inactivation of A sub-genome ofGbHD-1mainly affected the trichome initiation on the upper epidermis of leaves, while D sub-genome ofGbHD-1mainly on the lower epidermis of leaves and new buds, which indicated functional differentiation ofHD-1in the A and D sub-genomes in cotton. Moreover, although a smooth epidermis was obtained in theGhHD-1Asilenced hairy accession (Fig. 4-A and C), there were still trichomes could be seen on the leaves, which might be due to the some transcription ofGhHD-1left inGhHD-1Asilenced plants(Fig. 4-B and D).

4.3. The important roles of GhHD-1 in trichome initiation in different tissues of cotton

Roles ofGhHD-1in trichome formation have been reported previously (Walfordet al. 2012; Shanet al. 2014; Dinget al.2015). The cotton fiber is a highly elongated and thickened single-cell trichome, andGhHD-1was first proved to regulate fiber initiation and formation (Walfordet al. 2012).Silencing ofGhHD-1could delay the timing of fiber initiation,and overexpression ofGhHD-1increased the number of fibers initiating on the seed. Moreover,GhHD-1interacted withGhHOX(Homeobox 3), another HD-Zip IV gene, and activated downstream gene transcription and promoted fiber elongation (Shanet al. 2014). Based on genetic mapping,gene expression, and association analyses, the hairless stem phenotype in someG.barbadenseaccessions was con firmed to be related to a copia-like retrotransposon insertion and the reduced transcription ofGhHD-1A(Dinget al. 2015). Here, we found thatGhHD-1Awas involved in trichome initiation, mainly on the leaf epidermis of cotton and the retrotransposon-interrupted allele ofGhHD-1Awe described was indeed identical to the H7124 TE2 allele identi fied earlier (Dinget al. 2015). Taken together,these findings suggest thatGhHD-1and its orthologs inG.barbadenseplay important roles in multiple epidermal trichome development processes in cotton.

Fig. 5 Silencing of HD-1 in hairless Hai7124 accession via virus-induced gene silencing (VIGS) assay. The trichome phenotypes(A) and transcription levels of HD-1 (B) after HD-1 silencing in the untreated control plants (CK), the empty vector-silenced plants(TRV::00), the positive control CLA1-silenced plants (TRV::CLA1) and the HD-1-silenced plants (TRV::HD-1). The expression levels of the Tubulin 1 gene (AF487511.1) was used as the endogenous control. The relative expression levels were calculated using the 2–ΔCT method (Livak and Schmittgen 2001) and error bars indicated the standard deviation (SD) of three biological replicates.＊＊, the significant difference at the 1% level.

Similar functions in the regulation of epidermis hair development have also been reported for other HD-Zip IV members (Henrikssonet al. 2005; Nakamuraet al. 2006).In this study, we detected differences in the expression of HD-Zip IV genes in different tissues of TM-1 (Appendix I).For example, mRNA transcripts of the HD-Zip IV genes were abundant in the ovules on different days post anthesis.However, some members displayed preferential expression patterns in other tissues. Compared to other HD-Zip IV genes,Gh_A05G3845andGh_D05G1246showed more abundant expression in roots, suggesting a role in the regulation of root hair development. We also found that some members were expressed preferentially in stems,leaves, and fibers of different stages. Although there has been some progresses in elucidating the regulation of the leaf and root hairs ofArabidopsis, the development of hair is quite complex. Further studies are required to clarify how trichomes form in different tissues. Hopefully,GhHD-1and its orthologs will be a valuable start point for the study of molecular mechanisms related to stem, leaf trichome, and fiber development in cotton.

5. Conclusion

The morphology of the cotton hairs was quite diverse.G.barbadensecv. Hai7124 plants generally displayed a smooth surface but displayed hairy leaf veins, leaf margins,and leaf buds; WhileG.hirsutumacc. T586 plants had dense hair over the whole plant. Here, we generated a mapping population consisting of 1 065 BC1individuals from the crossing Hai7124 and T586 and presented the fine genetic mapping of the T1locus, which involved in the leaf epidermal trichome initiation in allotetraploid cotton asGh_A06G1283(GhHD-1A) gene.GhHD-1Agene encoded a protein of 725 amino acids and belonged to HD-Zip IV gene family. However, in hairless accession Hai7124, there was aTy1-copiaLTR retrotransposon inserted in the ninth exonofGhHD-1Agene. To con firm the functions ofGhHD-1Aand the homologsGhHD-1Din trichome formation, we silenced the transcription ofGhHD-1(bothGhHD-1AandGhHD-1D) in T586 and Hai7124, and other three hairy cotton accessions 72t2-4, 81t3-4, and 121t2-4 by VIGS assay. The results implied bothGhHD-1AandGhHD-1Dgenes regulated the trichome initiation, but might mainly on different tissues. Meanwhile, cotton fiber was a highly elongated and thickened single-cell trichome on ovules,with important economic value. Further analysis found that except for the leaf trichome, the expression level ofGhHD-1gene was also positively consistent with the density of fibers on ovules. Therefore, we deduced that up-regulation ofHD-1might contribute to trichome initiation in cotton.

Acknowledgements

This work was supported by the National Natural Science Foundation of China (31471539) and the Jiangsu Collaborative Innovation Center for Modern Crop Production Project, China (No.10).

Appendicesassociated with this paper can be available on http://www.ChinaAgriSci.com/V2/En/appendix.htm