Applications and evaluation of DGE-based analysis with the reference database
Cotton is a major crop for fibre and oil production, and has been subject to the application of biotechnology for crop improvement. Cell culture and plant regeneration are the bases for cotton biotechnology through genetic transformation, and so understanding the molecular control of dedifferentiation and redifferentiation is key to manipulating the SE process. However, the large unsequenced genome size (approximately 2.5 gb), polyploid nature and lack of adequate gene model annotations have limited large-scale transcriptome analyses during cotton SE
. Previous studies on the molecular aspects used SSH and microarray
[11, 26] but provided limited information on the complex transcriptome dynamics during cotton SE. However, next-generation technologies, which can generate tens of thousands to tens of millions of sequence reads with exceptional reproducibility, provide new strategies to quantitatively analyse the functional complexity of transcriptomes, despite uncharacterized genome sequences
[27, 29, 40].
Using RNA-Seq technology developed by Illumina and elite high efficient regeneration lines YZ1, we designed a protocol for analysing the transcriptome complexity of cotton SE. Although SE is usually divided into two stages, induction and expression
, our morphological and histological observations indicate that it could usefully be divided into three different processes: dedifferentiation of somatic cells, transition from NECs to ECs and development of somatic embryos. Protoplasts undergo cellular dedifferentiation and initiate cell division within 48 to 72 h in tobacco and Arabidopsis
[41, 42]. Histological observations have shown that cotton somatic cells activate cellular dedifferentiation and division within 72 h, and often within 48 h
. We chose three different time points (6 h, 24 h and 48 h) for initial dedifferentiation sampling and typical NECs after 40 d of induction for late dedifferentiation. Different stages of somatic embryos were selected by distinct morphology observed after synchronization (Figure
For genome sequence references that were unavailable, clean tags were mapped to two different EST reference databases after preprocessing the raw data. One reference database (Reference database 1) was cotton unigenes from NCBI that contains 20,671 unigene sequences, and the other (Reference database 2) was contigs assembly from multiple cotton genes from different databases which contain 65,386 sequences. The two reference databases were compared for efficiency based on several criteria. So many tags were missing using Reference database 2 for critical selection, that tags mapping to unique sequences were used for transcript identification ( Additional file
11 Table S9). However, validation by qRT-PCR of the expression profile for 26 differentially expressed genes derived by using RNA-Seq technology showed that genes mapped based on the two reference databases exhibited a similar correlation (R
2 = 0.7077 for Reference database 1, and R
2 = 0.7073 for Reference database 2) ( Additional file
12 Figure S3). As a result, we selected the cotton unigenes from NCBI (Reference database 1) as our reference database for further analysis.
Up to 50.7% (15,339) of the sequences in our reference database could be unambiguously identified by a unique tag. However, a relatively low number of the tags (43.18%) could be assigned to genes and used for gene expression profiling. This might be partly explained by the fact that most of the sequences in the database were not generated from embryogenesis development. More sequences and annotation for dedifferentiation and redifferentiation in cotton have to be explored to illuminate the large amount of unknown tags that remain. The extremely low abundance transcripts (TPM ≤ 20) were also filtered because of the possible of sequencing error. Among these, 5,076 differentially expressed genes were filtered with a cut-off of TPM ≥ 20, P ≤ 0.001 and the absolute value of log2Ratio ≥1 based on the FDR < 0.05 ( Additional file
Transcription regulation of somatic cells dedifferentiation and redifferentiation in cotton
Somatic cells within the plant contain all the genetic information necessary to create a complete and functional plant (with the exception of anuclear vascular cells). The induction of SE comprises the termination of one gene expression pattern in the explant tissue and replacement with an embryogenic gene expression programme
. The initiation of the embryogenic pathway, which is preceded by cellular dedifferentiation, is restricted only to certain responsive cells in the primary explant because the existing developmental information of somatic cells must be switched off or altered to make the somatic cells responsive for new signals
[1, 5]. Though we described the cotton SE as consisting of three different processes, it is very difficult to dissect the specific cellular events related to the overlapping phases of dedifferentiation, cell cycle reactivation and the acquisition of embryogenic competence.
The embryogenic processes are becoming better understood because of the identification of several genes such as transcription factors that play regulatory roles either in specific embryogenesis phases
 or throughout the whole process
. In the present study, 466 TF mRNAs were differentially expressed over a wide range of abundances during SE. Among these, a subset of TF families were associated with functions in cell differentiation, embryogenic patterning and embryo maturation processes (Zinc finger, b-ZIP, bHLH, B3 and MYB), meristem maintenance or identity (NAC, YABBY, GRAS), while others had roles in hormone-mediated signalling by auxin (Aux/IAA, ARF) or ethylene (AP2/ERF). Zinc finger family proteins have been proven to be involved in cell differentiation and development processes in animals and plants
[45, 46]. PEI1, encoding a protein containing a Cys3-His zinc finger domain, is an embryo-specific transcription factor that plays an important role during Arabidopsis embryogenesis, functioning primarily in the apical domain of the embryo, which is required for the globular to heart-stage transition
. In the present study, genes encoding zinc finger family protein showed complex expression profiles ( Additional file
8Table S7), indicating that they have multiple functions during SE in cotton. In Arabidopsis, two bHLH proteins were required in embryogenic patterning for root formation in the embryo
[47, 48]. The diversity of expression profiles displayed by 78 bHLH homologues in the present study might suggest the complex regulation of SE by bHLH proteins in cotton ( Additional file
B3 domain transcription factors in Arabidopisis (LEC2
FUS3 and ABI3) encode regulatory proteins involved in embryogenesis and induction of somatic embryo development
[49, 50]. Six B3 family transcription factor homologues were present with complex expression profiles: two (zhu1_Ghi#S33821461, zhu1_Ghi#S42277219) were down-regulated and one (zhu1_Ghi#S42340389) was up-regulated. Ectopic expression of BABY BOOM (BBM), a member of the AP2/ERF family in Arabidopsis primarily induces spontaneous somatic embryo formation from seedlings, although ectopic shoots and callus also develop at a lower frequency
. In our study, 26 AP2/ERF genes were differentially expressed during SE in cotton ( Additional file
8 Table S7). MYB and WRKY were transcription factors not only involved in response to biotic and/or abiotic stresses, but also regulated embryogenesis pathways
[52, 53]. As revealed in this study, most MYB genes were up-regulated during embryogenic initiation and showed a relatively low expression level during somatic embryo development, while most MYB-related genes were down-regulated during SE, indicating different functions of these genes during SE ( Additional file
8 Table S7). Further experiments are required to verify the physiological function and interaction between these factors and other genes during SE.
Complex auxin signalling pathway during dedifferentiation and redifferentiation of cotton cells
The importance of PGRs during SE has been widely documented
[4, 54]. To understand better the hormonal regulation of SE, PGRs are added to a culture medium to induce somatic embryogenic process, and endogenous hormone concentrations of plant tissues are measured during morphogenesis or various developmental stages
. Auxin is considered to be a critical PGR in cell division and differentiation, as well as in the induction of SE. This regulation probably occurs by establishing auxin gradients during the induction phase of SE, essential for initiating dedifferentiation and cell division of already differentiated cells before they can express embryogenic competence
. Despite the absolute requirement for exogenous auxins to sustain growth in plant cells cultured in vitro, cultured plant cells produce substantial amounts of the native auxin, IAA. In carrot cells, exogenous auxin stimulates the accumulation of large amounts of endogenous IAA. Thus the application of exogenous auxin and subsequent endogenous auxin content are both determining factors during the induction phase
[4, 56]. For this gradient to be established, relatively high levels of IAA in the competent tissues may be necessary.
Most studies on hormone contents during induction of SE only evaluate NEC and EC cultures
[32, 57]. In the present study, the endogenous IAA concentrations were determined in the original somatic cells, phytohormone-induced dedifferentiation cells and embryogenic cells (Figure
2). De novo synthesis of IAA in these cells occurred under all the examined conditions. The endogenous levels of IAA declined to a half within 6 h and dropped to a quarter of the original values within 24–48 h following excision from seedlings (Figure
2). The kinetics of this decline in IAA levels was similar to the decline of IAA levels in wounded tobacco leaves by activation of the proteinase inhibitor gene system
. However, in IBA-treated soybean hypocotyls, IAA levels increased dramatically after wounding and reached a maximum after 24 h, with a decrease of the cationic peroxidase activity
The mechanism responsible for the decline in IAA levels is not yet understood. The activation of some proteinase inhibitor genes in this study might be one possibility. The endogenous IAA could be influenced at one of several points, including its biosynthesis or degradation or the formation of amide or ester storage forms. Indeed, the decrease in IAA pools could even be influenced through IAA transport. Our data shed new light on these questions.
Our analysis revealed the dynamics of auxin levels during cotton SE (Figure
2). Previous studies have also shown that sharp changes in endogenous auxin levels may be one of the first steps leading to SE
. Redifferentiation was clearly correlated with a sharp increase in auxin responses in cotton cells, which provides direct evidence for the significance of an endogenous auxin pulse in the expression of cellular totipotency. It has also been noted that transition of the globular embryo to the heart-stage embryo and its further development requires either a low level of auxins or their complete absence
. Surprisingly, most RNA-Seq based auxin synthesis, transport, metabolism and signalling pathway genes were down-regulated during redifferentiation and somatic embryo development processes and showed a relatively low expression level in EC cultures ( Additional file
9 Table S8-1). However, transcript levels of genes related to auxin were changed during this process, indicating a possible role of this hormone in cotton SE.
The increase of IAA might be due to the increased synthesis and turnover of putative host auxin precursor in tissues. Although there are several IAA biosynthesis pathways in higher plants, tryptophan has long been regarded as the important one and its active metabolism and biosynthesis during embryogenesis have been highlighted
. A TRP1 homologous transcript was differentially up-regulated during initial dedifferentiation at culture times of 6 h and 24 h (Figure
7A), with a consistent result in wheat
, while NIT4A showed an opposite expression model (Figure
7A). Nitrilases can contribute to IAA homeostasis by hydrolyzing IAN to IAA in higher plants
. Conversion of Trp to IAA by enzymatic complex with nitrilase immunoreactivity in vitro was applied to plants
. Expression of maize nitrilase ZmNIT2 is elevated in embryonic tissue
. In this study, NIT4A homologues were identified and were down-regulated during the whole process (Figure
7A), while qRT-PCR analysis of three additional nitrilase genes derived from cotton database gave a similar expression profile ( Additional file
13 Figure S4). In addition, the up-regulation of a GH3.17 gene was observed in early dedifferentiation at 6 h of induction (Figure
7A). Several members of this family, including the GH3 genes, were up-regulated at about 2–4 h in soybean hypocotyls exposed to auxin
. These enzymes encoded by members of the GH3 family are able to synthesize IAA amino acid conjugates. Two members of the Arabidopsis GH3 gene family have been revealed to be overexpressed in dwarf mutants with reduced apical dominance combined with decreased free auxin levels
[65, 66]. Further, disruption of certain GH3 genes confers hypersensitivity to specific forms of auxin conjugated by the encoded GH3
. The characterized GH3 enzymes in this process might indicate that not only the level of free IAA but also the conjugated IAA is important during SE (Figure
Likewise, chemical and genetic studies have revealed that transport of auxin is complex and highly regulated for embryonic development
. Several Arabidopsis mutants are defective in proteins mediating polar auxin transport. AUX1, which mediates influx of IAA into cells, was localized asymmetrically in the plasma membrane of certain cell types, facilitating directional auxin transport
. Once IAA has entered a cell via AUX1, several factors regulate efflux. Three AUX1 homologous transcripts (zhu1_Ghi#S33799879, zhu1_Ghi#S42308191 and zhu1_Ghi#S29994266) were down-regulated during cotton SE, while only AUXI-LIKE displayed high levels ( Additional file
9 Table S8-1). Like AUX1
PIN is another gene family implicated in polar auxin transport, which is asymmetrically localized in the cell
. A PIN3 transcript showed a high expression level during dedifferentiation but an extremely low level during the embryo development stage ( Additional file
9 Table S8-1). These results indicated the complex auxin flux during SE. More evidence is required, however, to prove a relationship between auxin transporters and auxin distribution during cotton SE.
Most of the auxin-inducible Aux/IAA transcripts, with the exception of one member (IAA19), had decreased expression levels during dedifferentiation and displayed extremely low levels at the EC stage but then increased during SE development (Figure
7C), showing a different pattern from endogenous auxin dynamics. Transcription changes of Aux/IAA genes after an auxin stimulus is likely to be mediated by ARF proteins via AuxREs in Aux/IAA promoter regions. ARFs can bind tandem repeat AuxRE sequences as homodimers, dimers with other ARFs or dimers with repressive Aux/IAA proteins
. Six auxin response factor–related genes were differentially expressed, with two showing the high expression levels (zhu1_Ghi#S42325122, zhu1_Ghi#S42278444 and zhu1_Ghi#S42310727) during the dedifferentiation process ( Additional file
9 Table S8-1). The up-regulation of some ARF transcripts might demonstrate an intimate connection between auxin responses and auxin levels during cotton SE (Figure
Genes connected to degradation of Aux/IAA proteins, such as the putative intracellular auxin receptor TIR1, was down-regulated during the process (Figure
7B), which was shown by two TIR1 homologues (zhu1_Ghi#S37590130 and zhu1_Ghi#S33811942). AXR1, which is part of the auxin-induced Aux/IAA degradation machinery via the 26 S proteasome
, was also differentially expressed ( Additional file
9 Table S8-1). Likewise, ASK1 and ASK2 are necessary for a proper auxin response, through interaction with the TIR1 F-box
. Two of four ASK homologous transcripts (zhu1_Ghi#S42334433 and zhu1_Ghi#S33808515) displayed relatively high expression levels, with ASK2 down-regulated during the whole process and ASK1 down-regulated during the initial dedifferentiation and then up-regulated during somatic embryo development ( Additional file
9 Table S8-1). These findings indicated the integration of the auxin signal pathways during cotton SE.
However, the expression profile of auxin-related genes revealed that the complex and redundant regulation of IAA abundance, transport and response allows an intricate system of auxin utilization that achieves a variety of purposes in SE. As a result, further study of these genes, from auxin biosynthesis to auxin metabolism, from regulated protein degradation to signal transduction cascades, from IAA abundance to auxin transport, is needed in cotton SE.