Research article

Cotton BLH1 and KNOX6 antagonistically modulate fiber elongation via regulation of linolenic acid biosynthesis

Fatty acid desaturase 3 (Δ¹⁵FAD) is the key enzyme involved in C18:3 biosynthesis (Liu et al., 2015b). Two fatty acid desaturase 3 genes, GhFAD7A-1 and GhFAD3-1, responsible for linolenic acid biosynthesis, were identified among the DEGs (Supplemental Data 6). qRT–PCR analysis demonstrated that GhFAD7A-1 and GhFAD3-1 were preferentially expressed in rapidly elongating fibers (Supplemental Figure 2G and 2H). We also investigated the relative expression levels of other DEGs involved in the unsaturated fatty acid metabolic pathway in fibers at different developmental stages. Relative expression levels of GhSTAD, GhSDR1, GhKCR2, GhM5XA18, GhTECR, GhKCR1, and GhKDSR were higher at 10 DPA; those of GhFAD12-1 and GhACOX2-2 were higher at 20 DPA; those of GhACOX3-1, GhFAD12-2, GhACOX4, and GhACOX3-2 were higher at 5 DPA; and those of GhACOX2-1 were higher at 15 DPA and 20 DPA (Supplemental Figure 3). Both GhFAD7A-1 and GhFAD3-1 were significantly upregulated in fibers from the GhBLH1-OE lines and downregulated in fibers from the GhBLH1-RNAi lines compared with those from wild-type plants (Figure 2H and 2I). The expression patterns of other DEGs involved in unsaturated fatty acid metabolism were similar to those of GhFAD7A-1 and GhFAD3-1 (Supplemental Figure 4). Taken together, these results suggest that linolenic acid is involved in GhBLH1-mediated fiber cell elongation in cotton.

TGGA cis-elements in the promoters of target genes serve as binding sites for BLH1 (Staneloni et al., 2009). We found that four TGGA cis-elements were located in the GhFAD7A-1 promoter. According to the distribution of TGGA cis-elements, we divided the GhFAD7A-1 promoter into three fragments, designated P1, P2, and P3. A Y1H assay showed that GhBLH1 bound to the P2 fragment, which contained one TGGA cis-element (Figure 3G). Similarly, transient in vivo expression assays showed that co-expression of GhBLH1 significantly activated transcription of the LUC reporter gene driven by the P2 fragment, but not the P1 or P3 fragment, of the GhFAD7A-1 promoter in tobacco leaves (Figure 3H, 3I; Supplemental Figure 5). Moreover, an electrophoretic mobility shift assay (EMSA) using purified GhBLH1 recombinant protein fused to a His-tag showed that GhBLH1 had a significant binding affinity for the P2 fragment of the GhFAD7A-1 promoter in vitro (Figure 3J). A chromatin immunoprecipitation qPCR (ChIP–qPCR) assay using transgenic plants overexpressing GhBLH1 with an HA tag at the C terminus (35S::GhBLH1-HA) revealed that the P2 fragment of the GhFAD7A-1 promoter was specifically enriched (Figure 3K). To further confirm that the TGGA cis-element in the P2 fragment of the GhFAD7A-1 promoter is indispensable for binding of the GhBLH1 protein, we performed a Y1H assay, a transient in vivo expression assay, and an EMSA. The P2m sequence is provided in Supplemental Data 7. GhBLH1 bound specifically to the P2 fragment containing one TGGA cis-element and activated transcription of the LUC reporter gene, but this binding was abolished when the TGGA cis-element in the P2 fragment was mutated (Figure 3L–3N). Moreover, the binding affinity was gradually reduced by addition of increasing amounts of unlabeled native fragment probes (cold probe) but was not affected by addition of unlabeled mutated probes (P2m) (Figure 3O). Taken together, these results strongly suggest that GhBLH1 positively regulates the transcription of GhFAD7A-1 by binding to the fourth TGGA cis-element in its promoter region.

To determine which domains are critical for the interaction of GhKNOX6 and GhBLH1, we divided GhBLH1 into two conserved domains (POX domain and HOX domain) and separated GhKNOX6 into four conserved domains, the KNOX1 domain (K1), KNOX2 domain (K2), ELK domain, and homeobox KN domain (HOX) (Supplemental Figure 10). A Y2H assay showed that only the POX domain of GhBLH1 was able to interact with GhKNOX6 (Figure 5E) and that the K2 and K1K2 domains, but not the ELK domain and HOX domain, of GhKNOX6 were able to interact with GhBLH1 (Figure 5F). To further verify this interaction, we performed a Y2H assay with the POX domain of GhBLH1 and the K2 and K1K2 domains of GhKNOX6. The results showed that the POX domain of GhBLH1 could interact with both the K2 and K1K2 domains of GhKNOX6 (Figure 5G). This interaction was confirmed by BiFC and pull-down assays (Figure 5H–5J). To further confirm the interaction between the POX domain and the K2 and K1K2 domains in planta, we performed a coIP assay in which K2-GFP or K1K2-GFP and POX-FLAG were transiently co-expressed in tobacco leaves. An anti-GFP antibody successfully precipitated POX-FLAG protein (Figure 5K and 5L). Taken together, these results suggest that the POX domain of GhBLH1 and the K2 domain of GhKNOX6 are essential domains for the interaction of GhBLH1 and GhKNOX6.

The results above showed that GhKNOX6 interacts with GhBLH1 to form a heterodimer and thereby interferes with the transcriptional activation of GhFAD7A-1 to negatively regulate fiber cell elongation in cotton. To further test the genetic relationship between GhKNOX6 and GhBLH1, we hybridized GhBLH1 transgenic lines and GhKNOX6 transgenic lines to generate GhKNOX6-OE and GhBLH1-RNAi double transgenic lines, GhKNOX6-Cas9 and GhBLH1-OE double transgenic lines, GhKNOX6-Cas9 and GhBLH1-RNAi double transgenic lines, and GhKNOX6-OE and GhBLH1-OE double transgenic lines. After transgene PCR amplification and analysis of relative expression levels, three lines of each hybrid progeny were selected for subsequent studies (Supplemental Figure 13). Compared with wild-type plants, the GhKNOX6-OE and GhBLH1-RNAi double transgenic lines produced much shorter fibers, whereas the GhKNOX6-OE and GhBLH1-OE double transgenic lines and GhKNOX6-Cas9 and GhBLH1-OE double transgenic lines produced longer fibers (Figure 7F and 7G). Compared with the GhKNOX6-OE plants, the GhKNOX6-OE and GhBLH1-OE double transgenic lines had significantly greater fiber lengths (Figure 7H). Conversely, the GhKNOX6-OE and GhBLH1-OE double transgenic lines had significantly reduced fiber lengths relative to the GhBLH1-OE plants. Similarly, the GhKNOX6-Cas9 and GhBLH1-RNAi double transgenic plants produced much longer cotton fibers than the GhBLH1-RNAi plants. Fibers from the GhKNOX6-Cas9 and GhBLH1-RNAi double transgenic plants were much shorter than those from the GhKNOX6-Cas9 plants. We also measured GhFAD7A-1 transcript abundance in the GhBLH1 and GhKNOX6 double transgenic lines. The results showed that GhFAD7A-1 transcripts greatly accumulated in the GhKNOX6-OE and GhBLH1-OE double transgenic lines as well as the GhKNOX6-Cas9 and GhBLH1-OE double transgenic lines (Figure 7I). However, GhFAD7A-1 expression was slightly decreased in the GhKNOX6-OE and GhBLH1-RNAi double transgenic lines compared with wild-type plants, and GhFAD7A-1 transcription in the GhKNOX6-Cas9 and GhBLH1-RNAi double transgenic lines was identical to that in the wild type. Linolenic acid accumulation was increased in the GhKNOX6-Cas9 and GhBLH1-OE double transgenic lines as well as the GhKNOX6-OE and GhBLH1-OE lines but decreased in the GhKNOX6-OE and GhBLH1-RNAi double transgenic lines (Figure 7J). Fiber length analysis revealed that this trait was stably inherited in multiple generations of GhBLH1 and GhKNOX6 double transgenic cotton (Supplemental Figure 14). Taken together, these results indicate that GhKNOX6 negatively regulates the GhBLH1-mediated promotion of fiber elongation by fatty acids.

The TALE homologous domain transcription factors, which contain a three-amino-acid extension in the loop that connects the first and second helices of the homologous domain, are key regulators involved in plant organ and tissue development, including meristem formation and maintenance, organ morphogenesis, and tissue development (Traas and Vernoux, 2002; Aida and Tasaka, 2006; Rast and Simon, 2008; Bleckmann and Simon, 2009). The plant TALE homology superclass is composed of BLH and KNOTTED1-LIKE HOMEOBOX (KNOX) proteins, which are involved in plant development and stress responses (Modrusan et al., 1994; Bürglin, 1997; Bellaoui et al., 2001; Hay and Tsiantis, 2010).

BLHs are conserved in eukaryotes and participate in various developmental processes (Staneloni et al., 2009; Ung et al., 2011). BLH1, SAWTOOTH 1 (SAW1), SAW2, and POUND-FOOLISH control leaf and shoot development, and ARABIDOPSIS THALIANA HOMEOBOX 1 (ATH1) regulates cell wall matrix formation (Byrne et al., 2003; Smith et al., 2004; Pagnussat et al., 2007; Yu et al., 2009; Etchells et al., 2012). In A. thaliana, BEL1-LIKE HOMEODOMAIN GENE 3 (BLH3) regulates apical meristem development (Cole et al., 2006). Verticillate primary branch 1 (VPB1), a member of the BLH family, is expressed in the apical meristem during early panicle development (Li et al., 2021). The maize blh12blh14 double mutant was not able to produce normal tillers and exhibited defects in axillary meristem growth (Tsuda et al., 2017). In this study, we identified the BLH protein GhBLH1, which regulated GhFAD7A-1 transcription through direct binding to the TGGA cis-element in the GhFAD7A-1 promoter region, thereby increasing linolenic acid accumulation to promote fiber cell elongation in cotton (Figures 1 and 3). Overall, this research expands our understanding of the function of BLHs in plant organ and tissue development.

KNOX proteins have been reported to regulate embryonic and postembryonic development in diverse plants (Tsuda and Hake, 2015). KNOX proteins are divided into two subfamilies, Class I KNOX (KNOXI) and Class II KNOX (KNOXII). KNOXI members are mainly expressed in the shoot apical meristem (SAM) to control SAM maintenance and leaf shape formation, and KNOXII proteins show functional diversity in plant development (Zhou et al., 2014). Despite some advances in clarifying the functions of KNOX proteins in Arabidopsis, information on the involvement of KNOX proteins in cotton fiber cell elongation is still scarce. Previous work showed that GhKNL1 controls fiber elongation and SCW synthesis through repression of downstream genes in cotton (Gong et al., 2014; Wang et al., 2022). Overexpression of GhKNOX2-1 resulted in more and deeper serrations in cotton leaves (He et al., 2021). In addition, KNOX proteins are likely to act as activators and repressors (Tsuda and Hake, 2015). Here we demonstrated that the novel Class I KNOX protein GhKNOX6 acts as a repressor of fiber cell elongation in cotton (Figure 6).

It is well established that members of the BLH family are expressed selectively in plants. The BLH6–KNAT7 physical interaction in Arabidopsis enhances both BLH6 and KNAT7 repression activities (Liu et al., 2014). The BEL1-like homeodomain protein VAAMANA interacts with the KNOX proteins BREVIPEDICELLUS and SHOOT MERISTEMLESS to regulate inflorescence stem growth in Arabidopsis (Bhatt et al., 2004; Kanrar et al., 2006). Moreover, BLH1 and KNAT3 regulate abscisic acid (ABA) responses during germination and early seedling development in Arabidopsis (Kim et al., 2013). The BLH1 and KNAT3 interaction is required for normal embryo sac development and modulates seed germination and early seedling development (Pagnussat et al., 2007; Kim et al., 2013). To date, most functions of the BLH and KNAT interaction have been verified in Arabidopsis, but the function of BLH–KNAT in cotton remains unclear. Our study identified a function of GhBLH1–GhKNOX6 heterodimers in cotton fiber elongation (Figures 5 and 7). GhBLH1 and GhKNOX6 perform opposite functions during fiber elongation, with GhBLH1 playing a positive role and GhKNOX6 a negative role (Figures 1 and 6). Specifically, interaction of GhKNOX6 with GhBLH1 interferes with the ability of GhBLH1 to promote fiber elongation (Figure 7).

Fatty acids are important components of various lipid classes in most organisms (Lv et al., 2021). In higher plants, long-chain fatty acids (LCFAs; fatty acids >C₁₆) and very-long-chain fatty acids (VLCFAs; fatty acids >C₂₀) play important roles in the plant cell elongation process (Dunn et al., 2004; Shi et al., 2006; Qin et al., 2007b; Liu et al., 2015b; Shang et al., 2016; Lv et al., 2021). VLCFAs have been reported to positively regulate the development of cotton fiber cells by activating ethylene biosynthesis (Qin et al., 2007a; Liu et al., 2020). The cotton fiber-specific gene GhCER6 regulates fiber cell elongation by regulating VLCFA synthesis (Qin et al., 2007b), and GhBZR3 suppresses cotton fiber elongation by inhibiting VLCFA biosynthesis (Shi et al., 2022). Previous research has shown that linolenic acid (C18:3) is one of the most abundant fatty acids in developing cotton fibers (Wanjie et al., 2005) and that linolenic acid promotes cotton fiber cell growth (Liu et al., 2015b). However, the molecular mechanism by which linolenic acid regulates fiber growth and how linolenic acid synthesis is precisely regulated remain unknown. In this study, RNA-seq analysis combined with a genetic analysis of GhBLH1 transgenic plants revealed that GhBLH1 regulates the biosynthesis of linolenic acid (Figures 2 and 4). GhBLH1 directly binds to the GhFAD7A-1 promoter to activate transcription of GhFAD7A-1, which mediates linolenic-acid-mediated fiber cell elongation (Figure 7K). However, formation of a functional dimer between GhKNOX6 and GhBLH1 affects the binding of GhBLH1 to the GhFAD7A-1 promoter to further influence linolenic acid synthesis and ultimately affect fiber development. These findings broaden our understanding of the function of unsaturated fatty acids in cotton fiber elongation.

The transgenic and non-transgenic cotton lines were grown in experimental fields in Xi’an, Shaanxi Province, and Sanya, Hainan Province, under standard farming conditions in accordance with relevant national approvals for biotechnology research (China). Two generations of breeding were conducted per year. In each experimental plot, transgenic and control lines were grown in a subplot that contained 30 plants in three rows (10 plants per row with a line spacing of 0.8 m). Fibers from 5, 10, 15, and 20 DPA cotton plants were frozen in liquid nitrogen immediately after harvest and stored at −80°C until use. Fibers at 10, 15, and 20 DPA and mature fibers were collected for measurement of fiber length. Cotton bolls from wild-type and transgenic plants were observed at 15 DPA. Tobacco (N. benthamiana) plants were grown in an incubation house at 22°C with a 16-h light/8-h dark cycle.

The coding sequences of GhBLH1, GhKNOX6, and GhFAD7A-1 were amplified from a cDNA library of G. hirsutum Xu142 with KOD FX Neo (TOYOBO, KFX-201) and inserted into the PC2300S-HF vector (Towin Biotechnology, Wuhan, China). The constructs were transferred into G. hirsutum (Jin668) to obtain overexpressing transgenic lines (Jin et al., 2012; Li et al., 2019). GhKNXO6- and GhFAD7A-1-silenced plants were obtained using the CRISPR/Cas9 system, and GhBLH1-silenced plants were obtained using an RNA-interference system. Single-guide RNAs (sgRNAs) were cloned into the plant expression vector pN7 (Towin Biotechnology, Wuhan, China) containing Cas9, and the constructs were transferred into G. hirsutum (Jin668) to obtain silenced transgenic lines (Wang et al., 2018). The RNAi vector was constructed according to the method of Tian et al. (2015), and genetic transformation was subsequently performed using a previously described method (Jin et al., 2005; Jin et al., 2012; Tian et al., 2015). For overexpression and silenced lines, at least three independent transformants per construct were generated. Pollen from the GhBLH1-OE lines was transferred to emasculated flowers of the GhKNOX6-OE lines, and pollen from the GhBLH1-RNAi lines was transferred to emasculated flowers of the GhKNOX6-cas9 lines. Fiber length was measured using 30 seeds collected randomly from each transgenic line. For hybridization, stamens (including anthers and filaments) were removed from flowers at −1 DPA, and pistils (including stigma, style, and ovary) were retained and covered with sealed bags to prevent interference from other pollen. The next morning, the stamen pollen prepared for hybridization was applied to the stigmas of the pistils treated the previous day, and the stigmas were again covered with a sealed bag. The next day, the sealed bag was removed, and the hybridization process was completed.

Transgenic and control lines were planted on two plots in an experimental field at Shaanxi Normal University, and immature fiber length was measured. Cotton bolls at 10, 15, and 20 DPA were collected and measured as described by Tang et al. (2014). Immature fibers that fell off the same boll location at the same time were placed in boiling water until the seeds attached to the fibers floated freely. The fibers were then held with tweezers, fiber length was measured with a ruler, and the fibers were washed with running water. In general, the fiber content of 10 seeds per boll from three to five bolls, which comprised a biological replicate, was measured.

Sequences of Arabidopsis KNAT family members were obtained from TAIR (https://www.arabidopsis.org/). A phylogenetic tree of AtKNATs and GhKNOX6 was constructed using MEGA 6.0 (Tamura et al., 2013) with the neighbor-joining method and 1000 bootstrap replicates. Multiple sequence alignment and prediction of conserved domains were performed with DNAMAN 4.0 and SnapGene Viewer, respectively.

Total RNA was extracted from fiber samples at 15 DPA using the RNAprep Pure Plant Kit (TIANGEN, Beijing, China). RNA samples (2 μg) were subjected to reverse transcription. First-strand cDNA synthesis was performed using TransScript All-in-One First-Strand cDNA Synthesis SuperMix (Transgen, Beijing, China).

qRT–PCR analysis of candidate gene expression in cotton tissue was performed with the LightCycler 480 system using SYBR Green (Roche, Indianapolis, USA). GhUBQ7 served as an internal control gene, and the primer sequences were (5′–3′) F-CCGCATTAGGGCACTCTTTTC and R-GGCATTCCACCTGACCAACAA. The other primers are listed in Supplemental Data 9. The reactions were performed in 20-μL volumes on a 96-well plate. The relative expression level of each gene was analyzed using LightCycler 480 Gene Scanning software. Gene expression levels were calculated by the 2^−ΔΔCt method with three independent PCR amplifications (Livak and Schmittgen, 2001).

Fibers (100 mg) from wild-type, GhBLH1-OE, and GhBLH1-RNAi plants at 15 DPA were used for total RNA extraction, and 2 μg of RNA was used to construct RNA-seq libraries. Transcriptome sequencing (RNA-seq) on the Illumina NovaSeq 6000 platform and subsequent data analysis were performed by LC Bio Technology Co., Ltd. (Hangzhou, China). Three independent biological replicates were included in the experiment. Genes that met the criteria of false discovery rate–corrected P < 0.05 and |log2(fold change)|  > 1 were considered to be DEGs. KEGG analysis was performed with KOBAS version 3.0 (http://kobas.cbi.pku.edu.cn/). The RNA-seq datasets are available at the NCBI SRA under accession number PRJNA967105.

Unsaturated fatty acids were extracted and measured as described previously (Liu et al., 2015). Fibers (2 g) from wild-type and transgenic plants at 15 DPA were immersed in chloroform/methanol (2:1, v/v) for 1 min to remove surface waxes. Each sample was then ground to powder in liquid nitrogen and extracted using methanol with 2.5% H₂SO₄ (v/v). Heptadecanoic acid (C17:0 [C17:0, Sigma-Aldrich, St. Louis, MO, USA]) was added to the extraction mixture as an internal standard to determine the percentage of fatty acids, and the sample was dried under nitrogen and heated in H₂SO₄ (3 N) at 85°C for 5 h. After the sample was returned to room temperature, fatty acid methyl esters were extracted three times with hexane and concentrated to a final volume of 200 μL. A gas chromatograph–mass spectrometer (GC/MS) system was used for fatty acid detection.

The coding sequences of GhBLH1 and the HOX and POX fragments of GhBLH1 were inserted into the pGBKT7 vector, and the coding sequences of GhKNOX3, GhKNOX6, and GhKNOX7 or the K1, K2, K1K2, ELK, and HOX fragments of GhKNOX6 were inserted into the pGADT7 vector. The primers are listed in Supplemental Data 10. Empty pGADT7 and pGBKT7 vectors were used as controls. The pGBKT7 recombinant plasmid and the pGADT7 recombinant plasmid were co-transformed into yeast strain AH109 by the LiCl–PEG method. The transformed yeast was grown on selective plates (SD/-Trp/-Leu) (Coolaber, Beijing, China) for 3–4 days, and the interaction was detected on SD/-Ade/-His/-Leu/-Trp (Coolaber, Beijing, China) selective plates.

Five milligrams of GST-tag-fused GhKNOX6 protein was incubated with GST-binding resin (GenScript, L0026, Nanjing, China) at 4°C for 8 h and washed with phosphate-buffered saline, and 2 mg of purified His-tag-fused GhBLH1 protein was added. The resulting mixture was incubated at 4°C for another 6 h. The resin was then washed three times with binding buffer, and 100 mL of elution buffer was added. The mixture was incubated at 4°C for 1 h. Western blotting was performed as described previously to analyze the pulled-down proteins using an anti-His antibody (Proteintech, Beijing, China).

The coding region of GhBLH1 was fused to a FLAG tag and inserted into the pCAMBIA1305-FLAG vector, and the coding region of GhKNOX6 was fused to a GFP tag and inserted into the pCAMBIA1305-GFP vector. The recombinant plasmids were transformed into Agrobacterium tumefaciens strain GV3101. Agrobacterium cells were resuspended in infiltration buffer (10 mM MgCl₂, 10 mM MES [2-(N-morpholino) ethanesulfonic acid] [pH 5.7], 150 mM acetosyringone) at an OD₆₀₀ of 0.6–0.8. Two or 3 days later, the leaf proteins were extracted using extraction buffer (pH 7.5) (100 mM Tris–HCl, 5 mM EDTA, 100 mM NaCl, 1.0% Triton-X-100, 0.5 mM PMSF, and protease inhibitor cocktail [Sigma, St. Louis, MO, USA]) and incubated with Pierce Protein A/G Magnetic Beads (Thermo Scientific, Waltham, MA, USA) at 4°C for 5 h. The coIP products were gently washed five times with wash buffer (150 mM NaCl, 1 mM EDTA, 1.0% Triton-X-100, 0.5 mM PMSF, protease inhibitor cocktail, and 20 mM Tris–HCl [pH 7.5]). GhBLH1-Flag and GhKNOX6-GFP fusion proteins were detected by immunoblotting with anti-FLAG antibody (1:1000; Proteintech, Beijing, China) and anti-GFP antibody (1:1000; Proteintech, Beijing, China), respectively. The chemiluminescence signal was detected by autoradiography.

For BiFC, GhBLH1 or its truncated POX fragments were inserted into the cYFP vector, and GhKNOX6 or its truncated K2 and K1K2 fragments were cloned into the nYFP vector. The primers are detailed in Supplemental Data 10. The cYFP recombinant plasmid and the nYFP recombinant plasmid were co-infiltrated into N. benthamiana leaves. After 3 days of culture, the leaves were observed under an Olympus FV1200 (Olympus, Tokyo, Japan) confocal laser scanning microscope at 20 μm. Fluorescence images were collected for detection of binding sites.

The coding sequence of GhBLH1 and the truncated POX or HOX fragments of GhBLH1 were inserted into the JG4-5 vector, and the GhFAD3-1 and GhFAD7A-1 promoters were inserted into the LacZi vector. The primers are detailed in Supplemental Data 10. The JG4-5 and LacZi recombinant vectors were co-transformed into yeast strain EGY48. JG4-5 was co-transfected with the LacZi vector into yeast strain EGY48 as a control. To test the effect of GhKNOX6 on binding of GhBLH1 to the GhFAD7A-1 promoter, JG-GhKNOX6 and JG-GhBLH1 were co-transformed with the indicated GhFAD7A-1 promoter reporter plasmids into the yeast strain EGY48. After 3–4 days of growth on SD/-Trp/-Ura medium (Coolaber, Beijing, China), the yeast was transferred to medium containing raffinose (Coolaber, Beijing, China), 40% galactose (Coolaber, Beijing, China), 10× BU salts (0.25 M Na₂HPO₄·7H₂O and 0.25 M NaH₂PO₄), and 20 mg/mL 5-bromo-4-chloro-3-indolyl-β-D-galactopyranoside (X-gal) (Sangon Biotech, Shanghai, China) and grown at 30°C under dark conditions for blue color development.

Tobacco transient expression and dual-luciferase assays were performed as described previously (Sparkes et al., 2006). In brief, coding sequences of GhBLH1, its truncated POX and HOX fragments, and GhKNOX6 were inserted into the pGreenII 62-SK vector, and the GhFAD7A-1 promoter and P1, P2, and P3 promoter fragments were inserted into the pGreenII 0800-LUC vector. The primers are detailed in Supplemental Data 10. Tobacco leaves were infiltrated as described above. Fluorescence was observed using the NightSHADE LB 985 in vivo plant imaging system (Berthold LB 985, Germany). A Dual-Luciferase Reporter Assay System kit (Promega, USA) and a GloMax 20/20 luminescence NightSHADE LB 985 detector (Promega, USA) were used to measure relative luciferase activity as described in the kit manual.

The coding sequences of GhBLH1 and GhKNOX6 were inserted into the PET-28a vector, and the protein products were purified by Ni-bead purification. 5′-Biotin-labeled primers with or without the putative GhBLH1-binding site were synthesized by Sangon Biotech (Shanghai, China). The primers are detailed in Supplemental Data 10. The EMSA was performed using a Chemiluminescent EMSA Kit (Beyotime Biotechnology, Shanghai, China). A fully automated chemiluminescence instrument (Tanon, Shanghai, China) was used for fluorescence observation.

The ChIP–qPCR assay was performed following a previously described protocol (Saleh et al., 2008). Fibers from GhBLH1-overexpressing cotton plants were collected at 8 DPA, fixed in 1% formaldehyde under vacuum for 10 min and washed three times with glycine buffer (Solarbio, Beijing, China). The immunoprecipitation was performed according to a previous report (Saleh et al., 2008) using anti-FLAG antibody (Proteintech, Beijing, China). Primer pairs were selected to amplify 150- to 200-bp promoter fragments. The primers are detailed in Supplemental Data 10.