Plant Cell Reports

, Volume 26, Issue 8, pp 1309–1320

Genes expression analyses of sea-island cotton (Gossypium barbadense L.) during fiber development


  • Li-Li Tu
    • National Key Laboratory of Crop Genetic ImprovementHuazhong Agricultural University
    • National Key Laboratory of Crop Genetic ImprovementHuazhong Agricultural University
  • Shao-Guang Liang
    • National Key Laboratory of Crop Genetic ImprovementHuazhong Agricultural University
  • Di-Qiu Liu
    • National Key Laboratory of Crop Genetic ImprovementHuazhong Agricultural University
  • Long-Fu Zhu
    • National Key Laboratory of Crop Genetic ImprovementHuazhong Agricultural University
  • Fan-Chang Zeng
    • National Key Laboratory of Crop Genetic ImprovementHuazhong Agricultural University
  • Yi-Chun Nie
    • National Key Laboratory of Crop Genetic ImprovementHuazhong Agricultural University
  • Xiao-Ping Guo
    • National Key Laboratory of Crop Genetic ImprovementHuazhong Agricultural University
  • Feng-Lin Deng
    • National Key Laboratory of Crop Genetic ImprovementHuazhong Agricultural University
  • Jia-Fu Tan
    • National Key Laboratory of Crop Genetic ImprovementHuazhong Agricultural University
  • Li Xu
    • National Key Laboratory of Crop Genetic ImprovementHuazhong Agricultural University
Genetics and Genomics

DOI: 10.1007/s00299-007-0337-4

Cite this article as:
Tu, L., Zhang, X., Liang, S. et al. Plant Cell Rep (2007) 26: 1309. doi:10.1007/s00299-007-0337-4


Sea-island cotton (Gossypium barbadense L.) is one of the most valuable cotton species due to its silkiness, luster, long staples, and high strength, but its fiber development mechanism has not been surveyed comprehensively. We constructed a normalized fiber cDNA library (from −2 to 25 dpa) of G. barbadense cv. Pima 3-79 (the genetic standard line) by saturation hybridization with genomic DNA. We screened Pima 3-79 fiber RNA from five developmental stages using a cDNA array including 9,126 plasmids randomly selected from the library, and we selected and sequenced 929 clones that had different signal intensities between any two stages. The 887 high-quality expressed sequence tags obtained were assembled into 645 consensus sequences (582 singletons and 63 contigs), of which 455 were assigned to functional categories using gene ontology. Almost 50% of binned genes belonged to metabolism functional categories. Based on subarray analysis of the 887 high-quality expressed sequence tags with 0-, 5-, 10-, 15-, and 20-dpa RNA of Pima 3-79 fibers and a mixture of RNA of nonfiber tissues, seven types of expression profiles were elucidated. Furthermore our results showed that phytohormones may play an important role in the fiber development.


cDNA arrayESTsFiber developmentNormalized cDNA librarySea-island cotton


The fiber of cotton is a differentiated single epidermal cell of the ovule. Fiber development progresses through four overlapping stages (Basra and Malik 1984): initiation, elongation (primary cell wall synthesis), thickening (secondary cell wall synthesis), and maturation. Cotton fiber exhibits diffuse cell growth, in which new wall and membrane material are added throughout the surface area of the cell (Tiwari and Wilkins 1995). These biological processes contribute to the quality and productivity of cotton fiber: the initiation stage determines the number of fibers on each ovule, elongation determines the length, and the secondary cell wall synthesis and maturation determine the strength and fineness. Elucidating the genetic developmental mechanisms of cotton fiber is necessary in order to manipulate fiber output traits.

Since the first fiber gene, E6, was isolated (John and Crow 1992) by screening the cDNA library, about 30 fiber genes and several promoters have been isolated and partially characterized using traditional molecular approaches, which provided a snapshot of the growth mechanisms of cotton fiber (Ruan et al. 2003; Li et al. 2005; Zhu et al. 2003; Wang et al. 2004, 2005). Then systematic molecular and genomic surveys of the genes crucial for the fiber development process have shed light on mechanisms responsible for fiber cell differentiation, expansion, and thickening. The first systematic study of the mechanisms involved cloning and analysis of a set of cotton genes related to fiber elongation using Gossypium hirsutum L. cv. Xuzhou-142 as the tester and its fuzzless-lintless mutant as the driver for suppressive subtraction hybridization (Ji et al. 2003). A total of 280 independent cDNAs were obtained and 172 fiber-specific or fiber-preferential cDNAs were identified using cDNA macroarrays. The findings suggested that KCBP and the MAPK pathway may be important for cotton fiber elongation, and vacuolar ATPase, AGPs, and two major types of wall-loosening enzymes, xyloglucan endotransglucosylases (XETs) and expansins, are important for fiber development. In an excellent groundwork study by Arpat et al. (2004), 14,000 unique genes were assembled from 46,603 expressed sequence tags (ESTs) from developmental stage fiber (7–10 dpa) cDNAs of a cultivated diploid species (G. arboreum L); this study provided an in-depth view of the genetic complexity of the transcriptome of the expanding cell. They isolated 2,553 significantly down-regulated fiber genes and 81 significantly up-regulated genes during the developmental switch from primary to secondary cell wall (Arpat et al. 2004). Using Arabidopsis genes, Lee et al. (2006) designed 1,334 70-mer oligo microarrays from a subset of cotton fiber ESTs. The gene expressions of the 0 and 3 dpa ovules of TM-1 and N1N1 mutant (the isogenic line of TM-1), leaves, and petals of TM-1 were analyzed. They concluded that ovular development at 3 dpa is a critical step for rapid cell expansion and cellular growth; the patterning gene GhPDF1, MYB transcription factors, and RDL1 proteins are important for fiber or fiber cell differentiation; and WBC1, FDH, EF1A, and NOD26 are likely involved in the late stages of fiber cell differentiation. In the first systematic survey of the action of phytohormones in the fiber development, Shi et al. (2006) found that ethylene plays a major role in promoting cotton fiber elongation. By sequencing a cotton fiber cDNA library of Xuzhou-142 (G. hirsutum L.) and subsequent microarray analysis, they showed that ethylene biosynthesis was one of the most significantly up-regulated biochemical pathways during fiber elongation. Another greater work is computational and expression analyses of 32,789 high-quality ESTs derived from TM-1 immature ovules, which were assembled into 8,540 unique sequences including 4,036 tentative consensus sequences and 4,504 singletons (Yang et al. 2006). They compared these ESTs with approximately 178,000 existing ESTs derived from elongating fibers and non-fiber tissues and concluded that general roles for genome-specific, phytohormonal and transcriptional gene regulated the early fiber cell development in cotton allopolyploids.

Among the cultivated cotton species, upland cotton (G. hirsutum L. 2n = 4× = 52, AD genome) accounts for about 90% of the world’s cotton production, sea-island cotton (Gossypium barbadense L., 2n = 4× = 52, AD genome) accounts for 5–8%, and Asiatic cotton (G. arboreum L. 2n = 26, A genome) accounts for 2–5%; herbaceous cotton (G. herbaceum L. 2n = 26, A genome) is not currently planted ( Although the cultivation of sea-island cotton is limited, it is one of the most valuable and expensive cotton varieties. Characterized by silkiness, luster, long staples, and high strength, G. barbadense is an ideal candidate for providing new genetic material useful for improving fiber quality in G. hirsutum. And transferring the excellent fiber traits from G. barbadense as the secondary gene pool to the vastly cultivated G. hirsutum via traditional and molecular-aided selection is an attractive aim of breeders. A series of molecular dissections of interspecific variation between G. hirsutum and G. barbadense were performed by a backcross-self approach, and many QTLs for fiber elongation, fiber length, and fiber fineness were detected (Chee et al. 2005a, b; Draye et al. 2005). The availability of DNA markers linked to the fiber quality traits of G. barbadense will assist breeders in transferring and maintaining valuable traits from exotic sources during cultivar development. The developmental mechanism of G. barbadense fiber versus that of G. hirsutum has not been surveyed, however, and only few ESTs involved in the G. barbadense fiber development are in GenBank. In this study we analyzed the genes specifically expressed in fiber of G. barbadense, and we discuss the species’ fiber development process based on sequencing data of preferentially expressed clones.

Materials and methods

Plant materials and RNA isolation

Gossypium barbadense cv. Pima 3-79 plants were cultivated in an experimental field (Wuhan Hubei) using normal farming practices. The bolls were tagged on the day of anthesis. All fibers at different developmental stages were removed carefully from ovules and immediately immersed in liquid nitrogen. Tissues such as leaves, petals, and ovules were harvested from plants in the field and placed in liquid nitrogen. Roots were collected from 15-day-old seedlings cultured in a growth chamber. All the frozen materials were preserved at −70°C before use. For the cDNA library constructions and cDNA arrays, development stage ovules (<5 dpa) and fibers (≥5 dpa) were harvested randomly from plants (Pima 3-79) to create biological pools. Total RNA was isolated from the collected tissues using a modified guanidine thiocyanate method (Zhu et al. 2005).

Normalized cDNA library construction

Purification of polyA mRNA from total RNA was carried out using the Oligotex mRNA kit (cat. no. 70042, Qiagen, Germany) following the manufacturer’s protocol. Five micrograms of polyA mRNA was used for cDNA synthesis using a SuperScript cDNA synthesis and plasmid cloning kit (Invitrogen Life Technologies, USA, cat. no. 18248-013) following the manufacturer’s instructions. After the individual libraries were constructed, the plasmid DNA of the cDNA libraries were extracted by the alkalinelysis method (Sambrook et al. 1989), and the normalized library was constructed by saturation hybridization between genomic DNA and mixed plasmid DNA from individual cDNA libraries (Chu et al. 2003). The Pima 3-79 genomic DNA was isolated by the modified CTAB method (Paterson et al. 1994). One hundred micrograms of genomic DNA was digested completely by either HindIII or EcoRI, and the digested DNA was precipitated by double volume of absolute ethanol and one-tenth volume of 2 M sodium acetate (pH 4.2) at –20°C. The recovered DNA, which had the adhered ends, were filled with C-16 biotin dUTP. Next, the DNA previously digested by HindIII was digested by EcoRI, and vice versa. Then the purified DNAs were mixed, denatured, and bound to M-280 Streptavidin (Dynal®, Biotech), which was hybridized to a mixture of denatured plasmid DNA for 24 h at 68°C. After washing off mismatched plasmids, the well-matched plasmids were recovered following the manufacturer’s instructions and then electroporated to DH10B (Chu et al. 2003).

Preparation of cDNA arrays and hybridization

Amplification of cDNA by PCR with primers complementary to the plasmid vector sequences flanking the insert cDNA (SP6 primer, 5′-ATTTAGGTGACACTATAG-3′; T7 primer, 5′-TAATACGACTCACTATAGGG-3′) was performed in 50-μl reaction volumes containing the following reagents: 1× PCR buffer, 0.2 mM dNTPs, 2 μl bacterial culture, each primer at 0.5 μM, and 2 U Taq polymerase. A PTC-100 thermocycler (Mu cor., USA) was used for the amplification. The DNA template was denatured at 94°C for 5 min, followed by 30 cycles of denaturation at 94°C for 30 s, annealing at 50°C for 30 s, and elongation at 72°C for 1 min 30 s, with a final round of elongation for 5 min at 72°C. The PCR products were precipitated with addition of 100 μl anhydrous ethanol and resuspended in ddH2O for arraying.

The plasmids (the alkalinelysis method) of cDNA clones randomly chosen from the normalized cDNA library or PCR products were transferred into 384-well-plates and dotted onto Hybond-N+ nylon filters (Amersham, UK) with Biomek 2000 Laboratory Automation Workstation (Beckman, USA). Cotton polyubiquitin cDNAs were also printed on each membrane as internal controls. Distilled water and empty vectors (plasmid array) or PCR products without template in the above reaction system (PCR products array) were used as negative controls. After denaturation, the DNAs were permanently attached to the filters in an oven at 80°C for 2 h and then the filters were stored at −20°C. The filter (8 × 12 cm) was arrayed into 384 grids, each containing 6 or 8 dots. The first-strand cDNA labeled with [33P] dATP was used as the hybridization probe for the cDNA array analysis (Superscript II, Invitrogen). Then probe was added to the hybridization solution (Sambrook et al. 1989), which contained the following reagents: 5× SSC, 0.1% SDS, 50 mM Tris–HCl, 10 mM EDTA, 1× Deharts. Hybridization was carried out overnight at 65°C. The washing procedure was as follows: wash 1 (2× SSC, 0.1% SDS), 65°C, 5 min; wash 2 (2× SSC, 0.1% SDS), 65°C, 5 min; wash 3 (2× SSC, 0.1% SDS), 65°C, 30 min.

Image acquisition and analysis

The hybridized membranes were exposed to Imagingplate (Fuji Photo Film Co., Ltd., Japan) for 12 h. The signals of filters were scanned using PhosphorImager SI (Molecular Dynamics, USA) and analyzed with the software Array-Gauge Version 1.0 (Fuji). After the levels of the local background were subtracted, the radioactive intensity of each spot was quantified. The array was normalized by dividing the quantified intensity of each spot by polyubiquitin cDNA quantified radioactive intensity, which was used as the internal control. Differential screening and expression pattern analysis data were averages of two independent experiments, respectively. Therefore, the expression levels for each EST were the average of four values. The expression profile was analyzed by CLUSFAVOR 6.0 (Peterson 2002). The intensity ratios were log10-transformed, and the clustering method was complete linkage (furthest neighbor).

ESTs processing and annotation

The clones of different expressions were subjected to single-pass sequencing reactions from the 5′ end with ABI 3700xl (Auke Biotech Co., Ltd., Beijing, China). Vectors were screened, and sequences shorter than 100 bp were removed. The remaining ESTs were aligned with the GenBank database (nr) using TBLASTX (, with 10−5 as the cutoff E-value. ESTs were named after the homologous sequences in GenBank. ESTs with BLAST scores lower than 45 bits (no homologous stretch longer than 50 bp) were designated as unknown. EST assembly was performed to obtain uniESTs using phrap software ( Functional categories were assigned by GO annotations (Ashburner et al. 2000; All the EST sequences have been submitted to the EST division of GenBank.

Northern analysis

After denaturing at 65°C for 15 min, 20 μg of total RNA in the loading buffer from each sample was loaded on a 1.2% agarose–formaldehyde gel for electrophoresis in the MOPS buffer (Sambrook et al. 1989). The fractionated RNA samples were then transferred to nylon membranes (Amersham), which were immobilized in an oven at 80°C for 2 h and stored at −20°C. Then the filters were hybridized with [α-32P] dCTP-labeled probe at 42°C for at least 20 h; the solution included 50% hybridization solution used in the array hybridization and 50% N,N-dimethylformamide. The washing procedure was as follows: wash 1 (2× SSC, 0.1% SDS), 42°C, 5 min; wash 2 (2× SSC, 0.1% SDS), 42°C, 5 min; wash 3 (2× SSC, 0.1% SDS), 42°C, 20 min. The Prime-a-Gene labeling system (Promega, USA) was used to generate probes. The radioautograph methods were similar to those of the array hybridization.


Individual libraries and the normalized library

We constructed four individual libraries of G. barbadense cv. Pima 3-79 fiber at the developmental stages of −2 to 4, 5 to 10, 11 to 16, and 17 to 25 dpa (Table 1). The length of inserts was identified by amplifying randomly selected clones with M13 forward and reverse primers after the individual library was constructed. The plasmids of the four individual libraries were mixed and hybridized to genomic DNA of Pima 3-79. The well-attached plasmids were washed and transferred to DH10B. The inserts of the normalized library were released by double enzyme digestion of the randomly selected clones with EcoRI and BamHI; most of the inserts were 0.8–1.5 kb, and the average length was 1.14 kb (Table 1). We also randomly sequenced about 100 clones and checked the redundancy of the library (data not shown). E6 and aquaporinPIP2-2, the most redundant transcriptions in cotton fiber (Arpat et al. 2004), were not included in the sequenced clones. The clones also contained rare genes, for example, a clone similar to Arabidopsis thaliana rubber elongation factor (REF) family protein had no hit when we performed a BLAST search of the ESTs in xgi ( Therefore, we concluded that the normalized library fitted the cDNA arrays preparation.
Table 1

Individual and normalized cDNA libraries for Gossypium barbadense cv. Pima 3-79

Fiber developing stage

Average length of inserts (kb)

Recombinant number

−2 to 4 dpa


1.20 × 106

5 to 10 dpa


1.70 × 106

11 to 16 dpa


2.52 × 106

17 to 25 dpa


1.48 × 106

Normalized library


2.12 × 106

aThe inserts were fragments double-enzyme digested with EcoRI and BamHI

Screening the cDNA library with Pima 3-79 fiber

To mine the novel genes involved in fiber development, we screened the 0-, 5-, 10-, 15-, and 20-dpa RNA of Pima 3-79 fiber using a cDNA array including 9,126 plasmids randomly selected from the library. Two-copy hybridizations were performed, and only those clones whose expression was significantly different between any two developmental stages were selected and sequenced (n = 929; Fig. 1). After processing the sequences, we obtained 887 high-quality ESTs (GenBank accession no.: EE592400 to EE593286), which were assembled into 645 consensus sequences (582 singletons, 63 contigs). TBLASTX results returned no significant similarities (E-value >10−5) for 111 ESTs (17.2%), and at least some of them were likely unique to cotton and important for fiber development. Of the 645 consensus sequences, 455 were assigned to functional categories using gene ontology (GO) ( with an E-value cutoff of 10−5.
Fig. 1

Screening of the normalized fiber cDNA library with Gossypium barbadense cv. Pima 3-79 fiber. The filter (8 × 12 cm) was arrayed into 384 grids with each containing four clones, and each clone has two copies (as the inset indicates). Those clones with different signal intensities between any two stages (e.g., 06A03, 06A04, 19N01 were up-regulated at 10 vs. 20 dpa and 24A02 was down-regulated at 20 vs. 10 dpa) were selected to sequence

Of the 455 consensus sequences (Fig. 2), the two largest categories were energy/carbohydrate metabolism (18%) and protein metabolism (18%). Almost 50% of the binned genes belonged to metabolism functional categories: energy/carbohydrate metabolism, protein metabolism, nucleic acid metabolism (7%), and other metabolism including amino acid metabolism and secondary metabolism (7%). These findings suggest that fibers are very metabolically active cell types. The genes involved in signaling and transporting formed the fourth (10%) and fifth (8%) largest groups, respectively. Only 4% of the binned genes were involved in environmental response/defense, although some functions of these genes may have not yet been discovered. For example, chitinase is classified in the environmental response/defense category because this enzyme is commonly considered to be used in resistance to diseases or pests; however, chitinase has been shown to be involved in plant development (Kragh et al. 1996; Mouille et al. 2003).
Fig. 2

Functional groups of the cotton (Gossypium barbadense) fiber ESTs. Consensus sequences were assigned to functional categories using gene ontology

Expression profiles of the sequenced cDNAs

The subarray was dotted with the sequenced cDNAs amplified by PCR. The sequences included in each contig were also amplified and dotted on the array, because they may have been different members of the gene families but not truncate sections of the same sequence. Then subarrays were hybridized with the 0-, 5-, 10-, 15-, and 20-dpa RNA of Pima 3-79 fiber and an equal mixture of RNA from the cultivar’s leaf, root, bud, and flower tissues. After removing the dots for which the cDNAs were not amplified or showed poor repetitions, 781 of 887 high-quality ESTs were analyzed. Based on the hybridization signal intensities of the nonfiber tissue mixture, it was obvious that the intensities of 0- and 20-dpa fiber were lower than those of 5-, 10-, and 15-dpa fiber. When comparing the fiber signal intensities of these five stages to the intensity of the nonfiber tissue mixture, only 118 of 781 ESTs (excluding possible overlapping or redundant sequences) were found to be expressed constitutively with a ratio <2 (data not shown). We then analyzed the expression profiles of the 663 genes specifically or preferentially expressed in fiber. The 596 cDNAs (67 ESTs were unclassified) showed seven expression types (the details of the ESTs involved in each type are shown in Figs. S1 to S7).

Type I (12 ESTs) had the highest levels around 0 dpa and then decreased. For example GbGA20ox1, which is similar to gibberellin 20-oxidase, a class of GA biosynthetic genes, had the highest expression level at 0 dpa (threefold vs. 20 dpa; Table 2, Figs. 3, 4, S1) and then decreased continuously. This expression mode was similar to the one of gibberellin 20-oxidase 1 (Gossypium hirsutum) reported by Shi et al. (2006). The protein sequences and nucleic acid sequences of the CDS of these two genes had only several amino acids/nucleic acids different, but the difference of 3′-UTR was large (data not shown). This maybe suggested that GAs may only be necessary during fiber initiation or the early stages of fiber development.
Table 2

The expression levels of selected ESTs

GenBank accession no.



5:0 dpa

10:0 dpa

15:0 dpa

20:0 dpa
























































































































Fig. 3

The expression profiles of the 596 ESTs in the different development stages of Gossypium barbadense cv. Pima 3-79 fiber (analyzed with CLUSFAVOR 6.0). The y-axis is the normalized signal intensity (log10 scale), and the x-axis represents fibers at 0, 5, 10, 15, or 20 dpa
Fig. 4

Expression patterns of eight ESTs as revealed by RNA gel blot analysis. Twenty micrograms of total RNA of Gossypium barbadense cv. Pima 3-79 tissues were loaded. The 0 dpa represents the ovules, and the 5–25 dpa represent fibers. The probes were labeled with the complete cDNA sequences: GbPDF2 (type III) (A); GbGA20ox1 (type I) (B); GbEX3 (type IV) (C); GbGLP3 (type V) (D); GbTLP1 (type VII) (E); GbCLT1 (type VII) (F); H2A (type IV) (G); E6 (type VI) (H); and rRNA (I), ethidium bromide staining 28S and 18S rRNA showed equal loading of RNA

Type II (22 ESTs) had high levels around 5 dpa and then decreased significantly. They were divided into two clusters. Cluster 1 had a higher expression level than cluster 2 and included ESTs similar to the tubulins, E6 genes, ribosomal protein S28, and two ESTs of unknown function (Fig. S2). In cluster 2, an EST was similar (74% identity, 87% positive) to fiber annexin (U89609), which was isolated from the G. hirsutum fiber cDNA library and the recombinant protein showed a GTPase activity (Shin and Brown 1999) and possibly modulated the activity and/or localization of callose synthase in cotton fibers (Andrawis et al. 1993).

Type III (115 ESTs) generally reached their highest expression from 5 to 10 dpa, and there were four clusters in this type. Cluster 1 (34 ESTs) had the lowest expression level. Cluster 4 (12 ESTs) had the highest expression level, and at 0 dpa cluster 4 had the highest level among the four clusters. Cluster 2 (38 ESTs) and cluster 3 (31 ESTs) had the intermediate expression levels (Fig. S3). Protodermal factor 1 (PDF1) is a protein involved in cell fate determination. In Arabidopsis, PDF1 is exclusively expressed in the L1 layer of vegetative and floral meristems, organ primordia, and protodermal cells during embryogenesis, but its expression is undetectable in the epidermis of mature organs (Abe et al. 1999, 2001). GhPDF1, a G. hirsutum gene similar to PDF1, was highly expressed from immature ovules (−3 dpa) to 5 dpa in the N1N1 mutant although undetectable in 10dpa, but its expression was detectable at 3 dpa and reached its highest level at 5 dpa in TM-1 fiber (Lee et al. 2006). We also found a GbPDF gene whose deduced amino acid sequence was 69% identity and 79% positive to PDF1 in Arobidopsis in type III, which had similar expression mode according to array and Northern hybridization analyses (Table 2, Fig. 4). The exact mode of action of expansins remains uncertain, but the evidence suggests that these proteins are nonhydrolytic and act to loosen the hydrogen bonds between the structural polymers of the cell wall matrix (McQueen-Mason et al. 1992; McQueen-Mason and Cosgrove 1994, 1995). It is likely that expansins are very important and redundant in fiber elongation (Orford and Timmis 1998; Harmer et al. 2002; Ji et al. 2003; Shi et al. 2006). We found six expansins after screening the plasmid array, which were divided into two classes. They had similar deduced amino acid sequences (only several amino acids different) but the different DNA sequences, especially the 3′-UTR. Class 1 (GbEX1, GbEX2) was a truncated version of class 2, which was about 70 amino acids shorter at the C-terminal. The G. hirsutum expansins in GenBank are all “complete version” of class 2. There were three expansins in type III. GbEX1 and GbEX2 were included in cluster 2, and GbEX5 had the highest expression level and was in cluster 4. Another EST, similar to G. hirsutum fiber annexin U89609 (99% identity, 99% positive), was also found in cluster 2 of type III. The signal intensity of this EST was higher than that in type II.

Type IV (87 ESTs) maintained similar RNA levels from 5 to 15 dpa. This type was classified into four clusters, with 34, 10, 28, and 15 ESTs included in clusters 1, 2, 3, and 4, respectively (Fig. S4). ESTs similar to tubulins, E6, LTPs, arabinogalactan proteins, and expansins were included in this type (Fig. S4). Expansin and XET are major wall-loosening enzymes in different plant cells (Carpita and McCann 2000), and the fiber cell elongates rapidly from 5 to 15 dpa. There were three expansins and an XET in type IV. GbEX3 was included in cluster 2, and GbEX4 and GbEX6 were in cluster 3. GbXET (100% identity to XET AAO92743.1) was included in cluster 1. We also found an auxin binding protein homologue in cluster 2. GbGLP2 (Table 2, Figs. 3, S4) showed 68% identity to ABP19 (Ohmiya et al. 1998) and 98% identity to GLP1 and ABP (Kim and Triplett 2004; Kim et al. 2004; Ji et al. 2003).

The expression level of type V (225 ESTs; about 38% of the 596 ESTs) peaked at 10 dpa, and this type was divided into nine clusters (Figs. 3, S5). Type III, type IV, and clusters 1 and 2 of type VI (discussed below) also had the highest expression levels at 10 dpa, although there was not a sharp peak of expression at this point (Fig. 3). These findings suggest that the majority of fiber development genes are expressed at high levels around 10 dpa, parallel to when fiber cells elongate most rapidly. Many genes involved in hormone signal-transduction pathways and the Ca2+-CaM second-messenger pathway (Table 2) were included in type V.

GbGLP1 (79-fold vs. 0 dpa, Table 2) showed 70% protein sequence identity to ABP19 and 100% to GLP1, and GbGLP3 67% identity to ABP19 and 82% to GLP1. The genes had the similar expression modes, but the expression level of GbGLP1 was higher (Table 2, Fig. 3). The expression levels of GLP1 and ABP also peaked at 10 dpa, although the function of GLP1 and ABP in fiber growth and development remains unclear (Kim and Triplett 2004; Kim et al. 2004; Ji et al. 2003). Differential display analysis revealed GLP1 expression in the early stages of cotton fiber development in TM1 and N1N1 (the near isogenic mutant), and the deduced protein sequence was similar to Prunus persica auxin-binding protein ABP19 (Ohmiya et al. 1998). Future studies are needed to clarify whether GbGLP genes bind auxin, but we can infer that these genes are important to fiber elongation.

The other two auxin-related genes, GbBHLH (auxin-induced basic helix-loop-helix transcription) and GbarCaM (auxin-regulated calmodulin), were also included in type V (Table 2). The deduced amino acid sequence of GbBHLH had 56% identity and was 66% positive to GhBHLH which is an auxin-induced cDNA fragment isolated from 6-day-old in vitro cultured cotton ovules (Tahhan and Randy 1999), and had a highly conservative TCP domain, a noncanonical basic-helix-loop-helix structure. This domain is also found in two rice DNA-binding proteins, PCF1 and PCF2, where it has been shown to be involved in DNA-binding and dimerization (Kosugi and Ohashi 1997). GbarCaM was an auxin-regulated calmodulin, and its deduced amino acid sequence had 100% identity to arCaM, which was isolated by differential screening from a mung bean (Vigna radiata) responsible to auxin (IAA, 2,4-D and NAA) (Okamoto et al. 1995). At 10 dpa GbarCaM was up-regulated 2.5-fold compared to 0 dpa, and then decreased.

GbERS1, whose deduced amino acid sequences showed 74% identity and was 88% positive to ERS1 (Ethylene Response Sensor 1), was also included in type V (Table 2). ERS1 was up-regulated by ethylene (Hua and Meyerowitz 1998) and was shown to encode an ethylene binding protein by using a yeast expression system and in vivo ethylene-binding assays (Hall et al. 2000). Shi et al. (2006) found that ethylene biosynthesis was one of the most significantly up-regulated biochemical pathways during fiber elongation and exogenously applied ethylene promoted robust fiber cell expansion, whereas its biosynthetic inhibitor l-(2-aminoethoxyvinyl)-glycine specifically suppressed fiber growth. Using array analysis, we found no up-regulated genes involved in ethylene biosynthesis, although the expression of GbERS1 (ERS type ethylene receptor) peaked at 10 dpa and maintained a similar level from 10 to 20 dpa during fiber elongation (Table 2).

Type VI (55 ESTs) had the higher expression level from 15 to 20 dpa and was classified into three clusters (Figs. 3, S6). ESTs similar to profilin, ADF1 (actin deploymerizing factor), α-tubulin 2, MAPK, UDP-glucose dehydrogenase, and Fb28 were included in this type.

Genes in type VII (80 ESTs) included three clusters (Figs. 3, S7). The expression levels of clusters 1 and 2 increased continuously from 0 to 20 dpa and included Fblate-2, celA1 (John 1996; Pear et al. 1996; Rinehart et al. 1996), GbCTL genes (chitinase-like protein), and GbTLP genes (thaumatin-like protein). There were four GbCTL genes and the DNA sequences of GbCTL1 (99% protein sequence identity to GhCTL1), GbCTL2 (96% protein sequence identity to GhCTL1), GbCTL3 (97% protein sequence identity to GhCTL1), and GbCTL4 (87% protein sequence identity to GhCTL1) were markedly different, which suggest that GbCTLs belong to a multigene family and the various members may have different roles in the development of the fiber, highlighting the genetic complexity of the transcriptome of fiber procession. GbTLP1 and GbTLP2, which displayed continuously increasing expression levels from 0 to 20 dpa (Fig. 4) showed 97% protein sequence identity to each other. The expression level of cluster 3 was low at 0 dpa, spiked around 5 dpa, and then increased to 20 dpa; members of this cluster included keratin-like proteins, FLA9 (a homologue to FLA, Arabidopsis fasciclin-like arabinogalactan-protein), and Fb10.


Reducing redundant genes by construction of normalized fiber cDNA library

A normalized cDNA library is an efficient tool for gene identification because it reduces the frequencies of prevalent mRNAs while enriching the rare ones. There are two normalization approaches: one is to decrease the abundance of cDNA of the prevalent class based on second-order kinetics, and the other is achieved by saturation hybridization with genomic DNA depending on the relatively even copies of most of the genes in a genome (Bonaldo et al. 1996). We used the latter method and successfully constructed the normalized fiber cDNA library (from −2 to 25 dpa) of the Pima 3-79 (genetic standard line of G. barbadense). By randomly sequencing about 100 clones and checking the redundancy of the library (data not shown), we found that most of the redundant genes in fiber, such as E6 and aquaporin PIP2-2, were not found, which meant that the normalized library fit the cDNA array preparation. In total, 63 contigs were assembled based on the sequencing data after screening 9,126-plasmid array (data not shown). The largest contig that was similar to α-tubulin 1 included 20 ESTs. Genes redundant in fiber as reported before, such as E6, Fb28, Fblate-2, α-tubulin 2, α-tubulin 4, metallothionein-like protein, and keratin-like protein (Fb10), were also redundant in the 887 ESTs. These redundancies were understandable, because these ESTs are highly expressed in fiber, these gene families were originally very redundant in the tetraploid species genome, and the screening processes had an accumulation effect on the specific fiber genes. The true level of redundancy will be clarified by more experiments.

Expression profiles of the sequenced ESTs

The cDNA arrays, which included 9,126 plasmids randomly selected from the library, were produced and screened with 0-, 5-, 10-, 15-, and 20-dpa RNA of Pima 3-79 fibers. We mined 887 ESTs that provided the groundwork for studying the fundamental biological processes of G. barbadense fiber development. By analyzing the profiles of the ESTs, we found that the fiber development mechanism is very complicated and many genes are involved in fiber development, including E6, Fblate-2, Fb-B6, celA1, Sus, GhACT1, GhWBC1, GaMYB2, and GhPFN1 (John and Crow 1992; John 1996; Rinehart et al. 1996; Pear et al. 1996; Li et al. 2005; Ruan et al. 2003; Zhu et al. 2003; Wang et al. 2004, 2005). Subarrays analysis of expression patterns indicated that most of the sequenced cDNAs had the high expression levels from 5 to 15 dpa and about 38% of the ESTs peaked around 10 dpa (Figs. 1, 3). Arpat et al. (2004) analyzed 24- versus 10-dpa fibers with oligonucleotide microarrays based on highly stage-specific dbEST (7- to 10-dpa fiber); they found that 2,553 “expansion-associated” fiber genes were significantly down-regulated and 81 genes were significantly up-regulated. The highly stage-specific oligonucleotide microarrays may partly explain why the down-regulated genes were more numerous than up-regulated genes during the developmental switch from primary to secondary cell wall syntheses. Lee et al. (2006) hybridized 3- versus 0-dpa fiber mRNA of TM-1 with 1,334 70-mer oligos microarrays; they found that 8 genes were down-regulated but 83 genes up-regulated. Our results showed that the total signal intensities at 0 and 20 dpa were markedly weaker than the intensities at 5, 10, and 15 dpa. All of the results suggested that the genes were less active before 0 dpa and after 20 dpa.

GbCTLs and GbTLPs may participate in the secondary cell wall synthesis

Plant chitinases are commonly considered to be defensive enzymes through the degradation of chitinous exoskeletons of insect pests or pathogens (Collinge et al. 1993). The understanding of chitinases has changed, however, since a chitinase able to rescue a carrot mutant in somatic embryogenesis (Kragh et al. 1996) was identified, and Mouille et al. (2003) showed that Arabidopsispom1 (encoding a chitinase-like protein) mutants were cellulose deficient. These findings suggest that chitinases are also involved in plant development. GhCTL1 and GhCTL2 isolated from G. hirsutum (Zhang et al. 2004) was preferentially expressed during secondary cell wall deposition in cotton fiber, and a fusion of the GhCTL2 promoter to the β-d-glucuronidase gene showed preferential activities in numerous cells during secondary cell wall deposition. We also found four GbCTL genes that displayed high expression levels from 15 to 20 dpa. All of the results confirmed that the chitinases are not only defensive enzymes.

Since the sweet-tasting thaumatin was isolated from Thaumatococcus danielli (Van der Wel et al. 1975), many TLPs have been mined, such as the bifunctional maize α-amylase/trypsin inhibitor (Richardson et al. 1987), salt-induced protein NP24 from tomato (King et al. 1988), osmotin from tobacco (Singh et al. 1989), PR-protein from tobacco (Cornelissen et al. 1986), NtTLP1 from tobacco (Kim et al. 2005), a drought-inducible but ABA-independent thaumatin-like protein from carrot (Jung et al. 2005), and TLG1 from Lentinula edodes (Sakamoto et al. 2006). All of these findings suggest that thaumatin-like proteins may be induced during abiotic stress. In our study, the thaumatin displayed continuously increasing expression levels from 0 to 20 dpa, the Northern hybridization confirmed this trend. It suggested that the thaumatin may participate in the secondary cell wall synthesis.

The possible mechanism of fiber development

For in vitro culture of unfertilized cotton ovules, addition of IAA and/or GA3 to the medium is required for normal ovule and fiber development. In the absence of phytohormones, unfertilized ovules browned and failed to increase in size or to produce fibers (Beasley and Ting 1974). These results suggest that phytohormones are important to fiber development. In this study we found many genes related phytohormone synthesis and signal transduction pathway had different expression profiles in the process of the fiber development. In vitro cultured ovules of 0 dpa usually needed exposure to GA3 for at least 1 day prior to transfer to IAA for 13 days, in order to produce the same amount of fiber obtained when they were cultured in the presence of IAA and GA3 continuously for 14 days (Beasley and Ting 1974; Beasley et al. 1974); however, the converse treatment (IAA pretreatment prior to ovule transfer to GA3) did not increase fiber development beyond that provided by constant GA3. In our study GbGA20ox1 were up-regulated in the early fiber development stage. This suggested GAs and their pathways were important in fiber initiation and early elongation. They may stimulate the expression of GbPDF genes, which determine the cells’ fate and the development of fibers. Then the accumulation of expansins, RD22-like proteins, and proteins related cell elongation are induced by auxin signal transduction pathways (GbABPs, GbBHLH, and GbarCaM). Vreeburg et al. (2005) found that ethylene regulated fast apoplastic acidification, expansin A was up-regulated during submergence-induced petiole elongation in Rumex palustris, but by blocking ethylene perception with 1-MCP, 1-NAA also up-regulated RpEXPA1 in submerged petioles; they concluded that the effect of auxin was independent of ethylene signaling. In our study, ethylene and its conduction also were involved in fiber elongation. The question of whether ethylene alone regulated the fiber development, as proposed by Vreeburg et al. (2005), or by cross-talking with auxin and other phytohormones needs further investigation.

The Ca2+-CaM second-messenger pathways may also have a role in fiber elongation, because we found two annexins and an auxin-regulated calmodulin (GbarCaM) were involved in fiber development. The annexin family is characterized by several internal repeats of 70 residues in length, each of which contains a highly conserved consensus sequence, the endonexin fold, which is believed to be involved in the binding of both Ca2+ and phospholipids (Burgoyne and Geisow 1989). GbarCaM was an auxin-regulated calmodulin with 100% protein identity to arCaM, whose expression level is increased by IAA (Okamoto et al. 1995).


This work was supported by a grant from the National Basic Research Program of China (2004CB117301) and National Natural Science Foundation of China.

Supplementary material

Copyright information

© Springer-Verlag 2007