A majority of cotton genes are expressed in single-celled fiber
- First Online:
- Cite this article as:
- Hovav, R., Udall, J.A., Hovav, E. et al. Planta (2008) 227: 319. doi:10.1007/s00425-007-0619-7
- 302 Views
Multicellular eukaryotes contain a diversity of cell types, presumably differing from one another in the suite of genes expressed during development. At present, little is known about the proportion of the genome transcribed in most cell types, nor the degree to which global patterns of expression change during cellular differentiation. To address these questions in a model plant system, we studied the unique and highly exaggerated single-celled, epidermal seed trichomes (“cotton”) of cultivated cotton (Gossypium hirsutum). By taking advantage of advances in expression profiling and microarray technology, we evaluated the transcriptome of cotton fibers across a developmental time-course, from a few days post-anthesis through primary and secondary wall synthesis stages. Comparisons of gene expression in populations of developing cotton fiber cells to genetically complex reference samples derived from 6 different cotton organs demonstrated that a remarkably high proportion of the cotton genome is transcribed, with 75–94% of the total genome transcribed at each stage. Compared to the reference samples, more than half of all genes were up-regulated during at least one stage of fiber development. These genes were clustered into seven groups of expression profiles that provided new insight into biological processes governing fiber development. Genes implicated in vesicle coating and trafficking were found to be overexpressed throughout all stages of fiber development studied, indicating their important role in maintaining rapid growth of this unique plant cell.
KeywordsCottonGossypium hirsutumFiberSingle-cellMicroarrayVesicle coating proteins
Days Post Anthesis
False Discovery Rates
Soluble NSF attachment protein receptor
Higher eukaryotes contain a multitude of cell types at maturity, each initiating its developmental program from undifferentiated progenitor cells (Honys and Twell 2004; Galbraith and Birnbaum 2006). In plants, it has been estimated that there are approximately 40 different cell types, each with their own function, structure, and location (Demura et al. 2002). Cellular differentiation is accompanied by myriad changes in transcription and translation, metabolism, and synthesis of intracellular structures. Insights into these complex processes require understanding the dynamics of transcription during cellular differentiation, growth, and maturation. At present, relatively little is known about the portion and proportion of the total transcript pool that is transcribed in most cell types or the degree to which global patterns of expression change during growth and maturation. In prokaryotes, whose entire lifecycle is completed in a single cell, the entire transcribed component of the genome is expressed at one or more stages, with the exception of genes induced by specific environmental stresses or challenges. In multicellular eukaryotes, however, this need not be the a priori prediction, because the number of genes is larger and the multiplicity of differentiated cell types.
Advances in technologies for harvesting specific cell types and in amplifying messenger RNA pools for expression profiling have stimulated studies of the transcriptome at the cellular level in plants (Galbraith and Birnbaum 2006). Explorations of changes during differentiation under natural or in vitro systems have been facilitated by isolating cells using sorting procedures or by laser microdissection and capture (Birnbaum et al. 2003; Leonhardt et al. 2004; Casson et al. 2005). Transcription profiling of single cells using microarrays has made it possible to evaluate the involvement of thousands of genes in biological processes, providing a powerful tool for analyzing cell differentiation and development. Recent applications include wood-forming cambial meristem tissue (Schrader et al. 2004), the analysis of the Arabidopsis thaliana pollen transcriptome (Pina et al. 2005), the quiescent-center cells of developing roots (Nawy et al. 2005), epidermis cells or vascular tissue of maize (Nakazono et al. 2003), and rapidly expanding cotton fiber initial cells (Wu et al. 2007). These experiments have a common obstacle of sample preparation and cell-type isolation that could impact interpretations of gene expression. Measuring gene expression in a single, abundant cell type will not have as much experimental induced error.
Here, we studied global patterns of gene expression during development of the single-celled epidermal trichomes of cotton seeds (Gossypium hirsutum). Seed trichomes, colloquially termed “cotton fiber”, comprise the world’s most important textile fiber and a vital component of the agricultural economy of over 50 nations. They also represent one of the most distinct single cell types in the plant kingdom. In some domesticated varieties, cotton fibers may attain a final length of 6 cm, or about one-third the height of an entire Arabidopsis plant (Kim and Triplett 2001). A single cotton ovary contains ∼500,000 elongating cells representing a single cell type. On the day of anthesis (flower opening), approximately one in four epidermal ovular cells has already been destined to become a cotton fiber, initially appearing as a spherical protrusion and subsequently elongating through stages of primary wall synthesis, secondary wall synthesis, maturation, and cell death. Here, we used cotton oligonucleotide microarrays containing 13,178 probes (Udall et al. 2007) to evaluate the transcriptome of cotton fibers across a developmental time-course, from two days post-anthesis (DPA) through the stages of primary and secondary wall synthesis. We show that the transcriptome of the single cell type cotton fiber is extremely complex, with most genes expressed at one or more stages during development, and that more than half of all genes are “up-regulated” when compared to a reference sample. We also present new insight into the biological processes leading to cotton fiber cell elongation and development.
Materials and methods
Plant material and RNA preparation
Gossypium hirsutum cv. TM1 plants were grown in four separate replicates of four to eight plants in the Horticulture Greenhouse at Iowa State University under supplemental lighting (16 h days). For each replicate, ovules were excised, immediately frozen in liquid nitrogen, and stored in −80°C. A heterogeneous mix of tissues was constructed by combining RNAs from, leaves, stems, petals, anthers, calyx, and bracts into a single reference sample. Roots were excluded from the reference tissue due to potentially high homology between roots hairs and cotton fiber cells. Other tissues were selected to ensure a diverse and complex transcriptomic representation, including various developmental stages and different plant organs. At each developmental time-point, fibers were isolated from ovules using a liquid nitrogen/glass bead shearing approach, as described (Taliercio and Boykin 2007). Initially, ovules were visually inspected for cell damage and the fibers were visually monitored under microscope to avoid contaminating epidermis cells. Subsequent RNA extractions were performed using a hot borate method (Wilkins and Smart 1996). RNA quality was confirmed on a BioAnalyzer (Agilent, Palo Alto, CA). Equimolar amounts of RNA (A260) from each separate replicate were pooled into a single sample for the six nonfiber samples destined to become part of the transcriptionally complex reference sample.
RNA amplification and labeling
For microarray analyses, an indirect labeling procedure of amplified aminoallyl a-RNA was used. Amplified-RNA (aRNA) was created for the reference RNA sample and for each fiber growth stage using the TargetAmpTM 1-Round Aminoallyl-aRNA Amplification kit (Epicentre Biotechnologies, Madison, WI, USA). About 0.5 μg of total RNA was used as starting material for one round of aRNA amplification resulting in 20–60 μg of aRNA.
Two dyes, Cy3 and Cy5, were coupled to 8 μg aliquots of aRNA using the Post-Labeling Aminoallyl-aRNA CyDye reactive dyes (Amersham Biosciences, Pittsburgh, PA, USA). The Cy3- and Cy5-labeled aRNA probes were purified using the Qiagen RNA easy Mini kit (Qiagen, Germantown MD, USA). Labeled aRNA products were analyzed for purity and yield (260 nm) using the NanoDrop spectrophotometer and for incorporation of Cy3 (550 nm) and Cy5 (650 nm) dyes.
Microarray hybridization and image analysis
For microarray hybridization, 300 ng of Cy3- and Cy5-labeled aRNA was used per slide. Slides were pre-hybridized using the Pronto!TM Plus system protocol (Promega Corporation, Madison WI, USA) with minor modifications as described below. Slides from each replicate were immersed in 200 ml of Pronto Universal Pre-Soak solution containing 2 ml of liquid sodium borohydride for 20 min at 42°C. Slides were transferred to fresh containers with Wash Solution 2 at room temperature for 2 min and were then immersed in 200 ml of hybridization buffer (5× SSC; 0.1× SDS; BSA 0.1 mg/ml). Slides were incubated with fresh Wash Solution 2 at room temperature for 2 min, and washed two additional times with Wash Solution 3 at room temperature for 2 min each. Finally slides were immersed in nuclease-free water and dried by centrifugation at 1,600g for 3 min. All hybridizations and post-hybridization washes were performed as described in the Pronto!TM Plus system protocol.
Microarray images were captured using an arrayWoRx® Biochip Reader (Applied Precision, Issaquah, WA, USA) using a light exposure period of 0.5 s for each channel (Cy5 and Cy3) at ∼10 μm resolution. GenPix® Pro (v 5.1, Molecular Devices, Sunnyvale, CA, USA) was used to manually align each block of feature positions to the microarray hybridization images. The signal intensity values were quantified as the background-adjusted median of pixels within the area of each feature’s spot circle.
Experimental design and statistical analysis
For each biological replication of each of the five fiber developmental time-points, hybridizations were performed using each fiber developmental stage paired against the same reference sample. With four biological replications, five time-points, and two dye swaps, we generated a total of 20 microarrays. Statistical analyses were performed using R and SAS statistical software (code available upon request).
Background-corrected signal intensity data were log (base 10) transformed and median-centered. For purposes of estimating whether a gene was expressed (see below), we transformed each normalized median value by adding the lowest value in each channel to the actual log/normalized intensity values, thereby forcing all intensity values to be positive. Two approaches were utilized to diagnose “presence” or “absence” of gene expression: (1) t-tests, to determine if mean intensity values were significantly different from zero, and (2) comparisons to ten noncotton control genes (SpotReport® Oligo Array Validation System, Stratagene, CA, USA). For the latter, log-intensities formed a bimodal distribution that showed two genes with high spot intensities values (data not shown); these two genes were omitted from subsequent analysis.
Cluster analysis of fiber gene expression was performed using a k-medoids clustering (Kaufman and Rousseeuw 1990) on the 5,430 genes inferred to be up-regulated in one or more stages of fiber development relative to the mixed sample. For these analyses intensity values were standardized so that each profile had a mean of 0 and a standard deviation of 1. The number of clusters was determined using the gap statistic approach (Tibshirani et al. 2001).
We used Blast2GO (http://www.blast2go.de/) to identify biochemical pathways involved in cotton fiber development and to calculate the statistical significance of each pathway. Blast2GO includes the Gossip package (Bluthgen et al. 2005) for statistical assessment of annotation differences between two sets of sequences, using Fisher’s exact test for each GO term. False discovery rate (FDR) controlled p-values were corrected for all differentially significant metabolic pathways.
Results and discussion
Cotton fiber transcriptome analysis
Gene expression in cotton fiber and a mixed reference sample
Dynamic changes across cotton fiber development and differentiation
Number of differentially expressed, up- and down-regulated genes in cotton fiber in comparison to reference samples
q < 0.1
q < 0.05
q < 0.02
q < 0.01
Differentially expressed genes
Few other studies offer comparable analyses of the transcriptome of single cell types. In a study of the transcriptome of human oocytes (Kocabas et al. 2006), amplified human metaphase II oocyte mRNA was compared to a reference sample consisting of a mixture of total RNA from 10 different normal human tissues. Compared with reference samples, there were 5,331 transcripts significantly up-regulated and 7,074 transcripts significantly down-regulated. These results mirror our own, showing that a single cell in both animal and plant systems may express many thousands of genes that are differentially expressed relative to more complex admixtures. Other studies have evaluated the transcriptome of developing haploid pollen cells in Arabidopsis, in which the pollen tube transcriptome was reduced compared to vegetative tissues, with a high proportion of enriched or selectively expressed genes families important for pollen tube growth (Honys and Twell 2004; Pina et al. 2005). In one of these studies (Honys and Twell 2004), when the pollen transcriptome was compared to seven sporophytic tissues, it was shown that 61.9% of the genome is transcribed in at least one developmental stage. Here we studied gene expression during development of one cell type in the sporophyte, including the most metabolically active stages of rapid fiber cell elongation and primary and secondary wall synthesis. Interestingly, our results, describing another asymmetrically growing cell, yield comparable values.
The foregoing results are compatible with another microarray analysis showing energy and cell components as important processes at the elongation phase of fibers (Arpat et al. 2004). In that experiment, two time points (10 and 24 DPA) were studied, representing primary and secondary wall synthesis stages. Our results show that energy and metabolism processes are more reflective of the initial stage rather than the elongation stage. Cluster 3 contains a quite large group of around 600 genes exclusively up-regulated at 7 DPA; genes overrepresented in this cluster encode components of the ubiquitin/proteasome protein catabolism complex involved in signaling pathways such as hormone-regulated processes, cell cycle control, photomorphogenesis, and senescence (Schaller 2004). This is the first time this group of the ubiquitin/proteasome complex has been observed in developing cotton at this stage. Genes included in clusters 4 and 5 are overexpressed during the time period of most rapid cotton fiber cell elongation (between 10 DPA and 20 DPA) (Applequist et al. 2001). These clusters are enriched for genes involved in water channel activity, microtubule and cytoskeleton formation, and regulation of the cell cycle, processes previously described as participating in cotton fiber elongation (Smart et al. 1998; Whittaker and Triplett 1999; Taliercio et al. 2005; Liu et al. 2006). Other gene families expressed in these clusters include components of cell wall biosynthesis, lipid metabolism, and cuticle biosynthesis, indicating the essential role of these cellular components during rapid elongation. Cluster 6 contains genes that gradually increase in expression during fiber development, with an over-representation of genes characteristic of primary and secondary cell wall biosynthesis, chitinase activity, and cellulose synthase activity, processes known to be enhanced during the later stages of fiber development (Arpat et al. 2004; Zhang et al. 2004). Genes belonging to cluster 7 show increased expression in two different stages, 2 and 25 DPA, and as a class include an over-representation of signal transduction pathway genes that may play a role in initiation and maturation processes, respectively. An example of expression patterns of eight well-characterized fiber development related genes are presented in Supplemental Fig. 1.
Gene expression analyses between adjacent time-points during fiber development
Differential GO annotations between consecutive developmental stages are presented in Supplemental Table 3 and Fig. 2. The results of these analyses parallel to those of the cluster analyses, but provide additional perspective on important biological processes. Genes up-regulated at 2 DPA compared to 7 DPA, for example, are biased toward cellular biosynthesis, ribosome structure, and transcription. Up-regulated genes at 7 DPA (compared to 2 DPA) include those implicated in hormone responses, such as gibberellic acid and ethylene-mediated signaling. By using microarray analysis on wild-type cotton ovules and fiberless mutant, Shi et al. (2006) has previously described the importance of hormone response in cotton fiber elongation. Another gene family overexpressed at 7 DPA is involved with cell membrane fusion reactions (SNARE binding, diacylglycerol O-acyltransferase activity, golgi to vacuole transport), suggesting that processes involved in increasing membrane production are required for the rapid unidirectional cell elongation that characterizes the cotton fiber. Additional biological processes are highlighted during fiber elongation, such as phenylpropanoid biosynthesis (which stops before 20 DPA) and expansin metabolism (Fig. 2). The expansin gene family is known as a key player in cell loosening and is considered to be important during cotton fiber elongation (Ruan et al. 2001).
The foregoing analyses demonstrate that global patterns of gene expression are highly dynamic during development of this single type of epidermal plant cell, with most of the genome being transcribed at one or more stages of cell growth. As we did not sample earlier stages of fiber initiation nor the later stages of maturation and cell death, both the conclusion of high transcriptome representation and that of clustered expression likely underestimate the true levels of both phenomena.
Biological process during fiber development
Classes of statistically overrepresented and under-represented biological processes throughout fiber development, relative to the reference sample. N—number of genes in each group
N in test
N in reference
Organic anion transporter activity
Response to hormone stimulus
Response to salt stress
Response to ethylene stimulus
Response to auxin stimulus
Response to biotic stimulus
Response to abscisic acid stimulus
Response to salicylic acid stimulus
The most remarkable group of genes detected in this fashion are those encoding vesicle coating proteins, in which 17 of 19 genes on the microarray were up-regulated in fiber. The importance of vesicle coating and trafficking to cell polarization and unilateral growth has been widely reviewed (Aroeti et al. 1998; Neumann et al. 2003; Macara and Spang 2006; Campanoni and Blatt 2007). The basic mechanism for vesicle transport membrane trafficking is similar between animals and plants, involving similar regulatory and structural proteins (Sanderfoot and Raikhel 1999; Pratelli et al. 2004). In animals, it has been shown that vesicle transport and localization play a key role in the distal end of the single-celled neuron elongation by fusion of intracellular, specific and dynamic vesicles specialized for plasmalemmal expansion in the growth cone (Pfenninger and Friedman 1993; Hirling et al. 2000; Steiner et al. 2004). In plants, most studies of the role of vesicle coating and transport regulation on cell tip growth have been on root hairs and pollen tubes (reviewed by Campanoni and Blatt 2007). It is known that membrane trafficking is linked to ion gradients and is fundamental to tip growth, particularly in supplying lipid and protein to the new plasma membrane and cell wall. It also has a complementary role for endocytosis in retrieving excess membrane and in recycling various protein fractions. Recent studies even suggest that proteins in the coated vesicle, like SNARE super-family (vesicle soluble NSF attachment protein receptor), are essential not only as housekeeping or “greasing” factors but for cell signaling as well. In this regard, coated vesicle trafficking is considered to be an active regulator of ion channel turnover and activity through its localization in the membrane during polar cell growth (Leyman et al. 1999; Pratelli et al. 2004; Sutter et al. 2006).
List of Oligo Identification Numbers (ID), cotton gene target and gene product names of 17 genes belonging to GO: 0030135 (coated vesicle) (Table 3)
Oligo gene target
Blast results from Arabidopsis nr database
(Q93ZN7) T25N20_16 (COPII domain)
(Q6ZGX8) Putative clathrin coat assembly protein
Nonclathrin coat protein zeta1-COP
(Q5JMS0) Putative syntaxin 6
Expressed protein (transporter activity)
(Q7X9R1) Clathrin coat assembly protein
(Q6ZDG9) Putative SEC23
(Q9LW87) Coatomer protein complex
(Q39834) Clathrin heavy chain
(Q6JJ39) Putative adapitin protein
(Q9FXB1) Putative clathrin-associated adaptor
(Q9SB50) Clathrin coat assembly like protein
(Q8S0N4) Vesicle transport SNARE protein-like
Coatomer delta subunit (Delta-coat protein)
(Q93ZN7) At1g05520 (COPII domain)
R.H., J.A.U., and J.F.W. conceptualized the experiment. R.H. managed the bench experiments and the data analysis. EH conducted part of the bench experiments. J.A.U., L.F., and R.R. contributed to data analysis. All authors assisted in drafting the manuscript. The authors thank Ryan Percifield for technical assistance, Alan Gingle for database management, Prof. Candace Haigler for help in secondary cell microscopic observations and the US National Science Foundation Plant Genome Program for financial support.