Transcriptomic profiles of non-embryogenic and embryogenic callus cells in a highly regenerative upland cotton line (Gossypium hirsutum L.)

Genotype independent transformation and whole plant regeneration through somatic embryogenesis relies heavily on the intrinsic ability of a genotype to regenerate. The critical genetic architecture of non-embryogenic callus (NEC) cells and embryogenic callus (EC) cells in a highly regenerable cotton genotype is unknown. In this study, gene expression profiles of a highly regenerable Gossypium hirsutum L. cultivar, Jin668, were analyzed at two critical developmental stages during somatic embryogenesis, non-embryogenic callus (NEC) cells and embryogenic callus (EC) cells. The rate of EC formation in Jin668 is 96%. Differential gene expression analysis revealed a total of 5333 differentially expressed genes (DEG) with 2534 genes upregulated and 2799 genes downregulated in EC. A total of 144 genes were unique to NEC cells and 174 genes were unique to EC. Clustering and enrichment analysis identified genes upregulated in EC that function as transcription factors/DNA binding, phytohormone response, oxidative reduction, and regulators of transcription; while genes categorized in methylation pathways were downregulated. Four key transcription factors were identified based on their sharp upregulation in EC tissue; LEAFY COTYLEDON 1 (LEC1), BABY BOOM (BBM), FUSCA (FUS3) and AGAMOUS-LIKE15 with distinguishable subgenome expression bias. This comparative analysis of NEC and EC transcriptomes gives new insights into the genes involved in somatic embryogenesis in cotton.


Background
The domain of plant transformation has enabled foundational discoveries in plant biology over the last 30 years. Plant genetic engineering has facilitated fundamental knowledge about gene function, trait genetics, and established critical linkages between the genome structure and function. Transgenic approaches have been widely used in the improvement and breeding of many crops, such as soybean, rice, and maize [1][2][3][4]. The rapid pace of genome sequencing has delivered genome maps of major domesticated crop plants and many non-model and wild species that have led to deep understandings of domestication and genetic diversity, gene function, and the development of tools for precise plant breeding [5]. However, the rate of genome characterization has far outpaced our ability to functionally profile genes and biochemical pathways in crops.
Genome editing through engineered nucleases [6] or CRISPR-Cas9 systems (spCas9-NG, base editing, xCas9, Cpf1, Cas13, Cas14) [7] are unprecedented technological breakthroughs that behold disruptive potential to precisely edit the genome of living organisms. These systems offer biologists the ability to ask precise biological questions; which has led to a new capacity in understanding gene function through targeted knockouts. Furthermore, genome editing technology has the potential to drastically reduce plant breeding cycle times and endow tailored trait genetics of elite breeding lines, which ultimately holds the potential to revolutionize commercial agriculture [5,8]. However, major obstacles remain in deploying these frontier technologies for global crop improvement. Main bottlenecks include plant transformation and regeneration through somatic embryogenesis, and a secondary challenge is the delivery of the genome editing reagents to plant cells to produce the intended effects [5]. Agrobacterium-mediated transformation [9] and subsequent whole plant regeneration from a single somatic cell is perhaps the most historic and preferred method to produce plants homoplastic for the transgene [9]. However, this type of whole plant regeneration is generally limited to a narrow range of genotypes within a species, usually with poor agronomic traits and with low efficiency [10][11][12].
Somatic embryogenesis (SE) is a unique process of embryo development that involves cycles of cellular dedifferentiation and reprogramming events that are controlled through signaling networks and gene expression cascades that eventually lead to embryonic cells [13]. During developmental reprograming, somatic cells de-differentiate to generate non-embryogenic cells from various explant sources, such as hypocotyls, young leaves, and immature embryos [9]. Non-embryogenic callus (NEC) cells can further differentiate to generate EC cells which function in somatic embryo development [14], which can be described by four primary stages in dicots: globular embryos, heart-shaped embryos, torpedo embryos, and cotyledon embryos [14]. Embryogenic cells are considered to be developmentally plastic and their programming of a particular developmental pathway is heavily influenced by environmental factors imposed by the tissue culture micro-environment, in addition to gene expression and regulation.
Manipulation of these programs in vitro has been achieved with some success in certain genotypes through large Design of Experiments (DOE) that determine and adjust the concentration and combination of growth regulators, such as the balance of auxin, cytokinin, and abscisic acid for each genotype [15][16][17][18][19]. Somatic embryogenesis is also sensitive to carbohydrate sources, inorganic salts, antibiotics, and amino acids [20][21][22][23], further adding to the complexity of the process and the difficulty of optimizing a protocol. In addition to medium components, genotype also plays a significant role in a species' ability to regenerate, exhibiting vastly different responses across closely related genotypes [24][25][26][27][28].
The molecular mechanisms that orchestrate somatic embryogenesis and endow biological totipotency are becoming better understood [29]. Previous studies have shown that genes expressed during the induction of SE can be divided into three primary categories; stressrelated genes, plant growth regulator (PGR) related genes, and transcription factors [30][31][32]. A biological stress, such as senescence or an acute stress such as abiotic/biotic factors can trigger stem cell formation through altered chromatin conformation and promiscuous expression of transcription factors [33][34][35]. Furthermore, these transcription factors have been shown to function in an important role on the regulation of plant differentiation [36] and development [37]. In monocots, overexpression of various plant transcription factors such as LEAFY COTYLEDON1 [38], WUSCHEL [39], and BABY BOOM [40] have been shown to improve embryo formation and enhanced regeneration [12]. In dicots, overexpression of BABY BOOM to enhance regenerative capacity has been reported in tobacco [41], sweet pepper [42], and Theobroma cacao [43].
In upland cotton (Gossypium hirsutum L.), successful somatic embryogenesis was reported nearly 40 years ago [44], however mature plants were not obtained until 1983 in the genotype Coker 310 [45]. Five years later, a report that describes the optimal media formulation for the induction of somatic embryogenesis from mature and immature tissues from the genotype Coker 312 was published [46]. A follow-up study screened 38 cotton cultivars for embryogenic potential and found only Coker 312 to be responsive and concluded that embryogenic potential is genotype specific in cotton [28]. Another study substantiated that somatic embryogenesis is a heritable trait in cotton and suggested that it is polygenic because of the segregation patterns that were observed F 1 , F 2 , and BC 1 generations of a recalcitrant by non-recalcitrant cross [47]. However, Coker lines are not well-suited for transformation studies because of their general poor agronomic qualities. In addition, Coker lines suffer from slow growth in tissue culture, a decline in vigor, and low regenerative capacity of the cultureswhich includes low potency of embryogenesis, difficulty of embryo germination, a low rate of rooting, low transplant success, and a progressive loss of totipotency in culture [48]. Recently, a line with a much higher embryogenic potential and better agronomic traits was released called Jin668 [31]. This genotype is a Coker relative and was developed through successive regeneration acclimation (SRA) that involves cycling a genotype through somatic embryogenesis multiple times [31]. This continuous cycling resulted in altered DNA methylation patterns that had an effect on gene expression profiles that persisted in the germ line with a greater regeneration efficiency [31].
In this study, we analyzed and compared the transcriptional landscape of NEC and EC cells harvested from Jin668 at these critical developmental stages. We present and discuss the various classes of genes that are active during the transition from NEC to EC cells, and identify several new candidate genes that enhances our knowledge of somatic embryogenesis in upland cotton.

Results
Somatic embryo formation efficiency of Jin668.
The embryogenic potential of Jin668 is high. NEC cells can be distinguished from EC cell by visual discrimination. NEC is characterized as fluffy, non-differentiated or polarized cells that are actively dividing, Fig. 1a. Typical EC is characterized as small cell clusters, with condensed cytoplasm, that are rapidly dividing and proliferating, Fig. 1b. After inducing callus for 45 days, 4% of the explants were found to be in the NEC stage and subsequently did not develop into embryos ( Fig. 1a and Table 1). The remaining 96% of the explants were found to be induced to form characteristic embryonic callus cells (Table 1 and Fig. 1b). In all experimental replications, both NEC and EC cells could be observed in the same piece of callus ( Fig. 1a and  b). Therefore, those calli that consisted of both NEC and EC cells were sampled for gene expression profiling between the NEC and EC (Fig. 1b).

Differential gene expression and GO enrichment of NEC and EC callus
For each callus cell type, an average of 36.8 and 42.7 million read pairs were produced for NEC and EC samples, respectively (BioProject ID PRJNA629328). A total of 39,105 genes were expressed in both NEC and EC callus cells, Fig. 2. Overall, a total of 5333 genes were differentially expressed with 2534 upregulated and 2799 downregulated in EC (Additional file 1: Table S1). Among these genes, a total of 144 were uniquely expressed in NEC cells, whereas, 174 genes were unique to EC (Fig. 2, Additional file 2: Table S2). Genes with unique expression to either EC or NEC cells were mostly genes of unknown function. Genes with annotations that were highly expressed in NEC include ROTUNDIFOLIA-LIKE 21, gibberellin-regulated, WRKY DNA binding, and others. Genes with high expression unique to EC callus with annotations include MYB family transcription factor, cytochrome C biogenesis, ROTUNDIFOLIA-LIKE 10, heat shock, and several others (Additional file 1: Table S1). Gene expression fold changes ranged from − 12 to 12 in EC when compared to NEC, (Additional file 1: Table S1). A slightly larger set of transcripts indicated transcriptional downregulation from NEC to EC, with a total of 6786 genes (logFC ≥1) becoming downregulated in EC; while 6538 (logFC ≥1) were upregulated ( Fig. 3a and b, Additional file 1: Table S1). The top 10 genes with the sharpest fold changes in upregulation EC versus NEC is a gene with a nodulin late domain (Gohir.D11G247300.1), and 9 genes with unknown function ( Table 2). The top 10 genes with the sharpest fold changes in downregulation are a gibberellin-regulate homolog (Gohir.D04G048900.1), 4 genes with unknown function, a gene with an S-adenosyll-methionine decarboxylase (AdoMetDC) leader peptide, a ROTUNDIFOLIA like gene (Gohir.D07G179500.2), and 60S ribosomal protein ( Table 2). Gene ontology   2a and b) discovered genes enriched in biological processes such as trehalose biosynthesis, transmembrane transporter, sugar transport, iron binding, growth factor activity, embryo development, development, auxin efflux, and response to stress among other processes related to cellular reprogramming and differentiation (Fig. 4a, Additional file 3: Table S3). The most pronounced enriched category was DNA binding and transcription factor activity with 309 genes combined, followed by regulation of transcription (155 genes), membrane (228 genes), oxidation reduction (105 genes), and transmembrane transport (80 genes) (Additional file 3: Table S3). Notable categories that were downregulated were categorized by assigned GO terms enriched for oxidationreduction (524 genes), transmembrane transport (233 genes), response to hormones and auxin (48 genes), methyltransferase activity (112 genes), and peroxidase activity (58 genes) (Additional file 4: Table S4).

Hierarchal clustering and GO enrichment of key clusters
Hierarchal clustering of statistically significant gene expression profiles identified 3 distinct sub-clusters. (Fig. 5a-c, Additional file 5: Table S5). Subcluster 1 contained 2186 genes, and the expression trend of these genes ranged from moderate to low in NEC and EC callus, respectively ( Fig. 5a, Additional file 5: Table S5). Processes enriched in this subcluster were assigned to oxidation reduction, methyltransferase activity, iron binding, and defense ( Fig. 5a, Additional file 6: Table  S6). Subcluster 2 contained 675 genes, and showed an upward trend in expression in EC callus and enrichment of processes related to transcription factor activity, membrane, growth factor, and developmental processes (Fig. 5b, Additional file 7: Table S7). Subcluster 3 contains 183 genes, and specifically showed a more discreet trend of higher expression in NEC and lower expression in EC callus, with enrichment in processes such as membrane and transport, peroxidase activity, carbohydrate metabolism, and fatty acid binding (Fig. 5c, Additional file 8: Table S8).

Phytohormone gene expression profiles in NEC and EC callus
Genes involved in phytohormone signaling and biosynthesis are crucial for the transition from NEC to EC. In total, there are 6132 genes annotated as a phytohormone in the Gossypium hirsutum (CV TM1) reference genome assembly [49]. In general, we observed an abundance of phytohormone related genes being downregulated in EC callus relative to NEC (Table 3). Abscisic acid and auxin related genes were the categories with the most abundant differentially expressed genes, including 1508 and 1332 genes, respectively (Table 3). Expression magnitudes ranged from − 10 (logFC) in genes involved in gibberellic acid to 9 (logFC) in genes involved in abscisic acid (Additional file 9: Table S9 and Fig. 6). Genes annotated as cytokinin, gibberellic acid, jasmonic acid, salicyic acid, ethylene, and brassinosteroid were differentially expressed in NEC and EC (Table 3 and Additional file 9: Table S9).

Transcription factors and signaling cascades
Transcription factors and cellular signaling genes that regulate reprogramming and differentiation during somatic embryogenesis were differentially expressed. We found 24 different classes of transcription factors that were significantly differentially expressed (Additional file 10: Table S10). The class with the most abundant genes was the GRAS (11 genes), followed by TGA like (9), MADS-box (9), CBF/NF-Y/archeal histone domain (9), GATA (8), and others with 7 or fewer members (Additional file 10: Table S10). Specific genes with known involvement in embryogenesis were individually analyzed for expression profiles in NEC and EC callus in Jin668. Our data shows that noteworthy embryogenesis transcription factors are lowly expressed in NEC callus with several key genes that are sharply upregulated in EC (Table 4). For example, the gene with the highest upregulation is LEAFY COTYLEDON 1 (LEC1), with the dominant copy residing in the D-subgenome ( Table 4). The next highly expressed embryogenesis related gene was the morphogenic regulator BABY BOOM (BBM) with the dominant copy on the A-subgenome, followed by FUSCA (FUS3) and AGAMOUS-LIKE15 (AGL15) both with the dominant copies on the D-subgenome. Interestingly, WUSHEL (WUS) had very little to no expression with no apparent subgenome bias ( Table 4).

Regulation of DNA methylation in NEC and EC
DNA methylation has a critical role in governing gene expression. In general, we observed a larger downregulation of genes regulating various methylation pathways in EC. We identified 226 genes with at least one-fold change in EC (71 upregulated, 155 downregulated) relative to NEC (Additional file 11: Table  S11). Of the downregulated methylation genes, 15 proteins belong to protein family methyltransferase, 26 belong to plant invertase/pectin methylesterase inhibitors, 14 involved in S-adenosyl-L-methioninedependent methyltransferase, 12 in the pectin methylesterase inhibitor family, and 35 in the O- methyltransferase superfamily (Additional file 11: Table S11). In contrast, very few methylation related protein families were upregulated in EC callus. Those that were include ribosomal protein L11 methyltransferase (3), tRNA methyltransferase (3), SAM dependent carboxyl methyltransferase (3), N6-adenine methyltransferase (3), and several others (Additional file 11: Table S11).

RT-qPCR and validation of RNA-seq
To validate the RNA-seq data, ten critical embryogenesis related genes were selected for RT-qPCR analysis. All   results of the ten genes, confirmed the results of RNAseq. (Fig. 7, Additional file 12: Table S12).

Discussion
Cotton is a major oilseed and fiber crop that could benefit from biotechnological improvement but is currently restricted to a few semi-recalcitrant genotypes that are amenable to low transformation frequencies and subsequent regeneration through somatic embryogenesis.
Genotype specific recalcitrance to regeneration is common in most crop speciesand has proven to be a profound bottleneck in applying the latest advances in biotechnology to elite breeding material. In monocots, several advancements have been made in ectopically expressing key transcription factors (eg, BBM/WUS2) to improve or enhance the development of embryogenic cell formation [40]. However, this system does not directly translate to dicot system improvement. Here, we   [59,60]. We also identified several genes that were upregulated in EC callus that are involved in phytohormone response, such as genes involved in carrier, efflux, response to auxin, cytokinin dehydrogenase, and homeobox domain (Fig. 6). Plant cells use these TFs to transmit signals to respond to environmental stimuli or initiate developmental programs. Furthermore, studies have shown non-cell-autonomous transcription factors are crucial for successful somatic regeneration [61]. Cellular proliferation and elongation are also very important for callus development during somatic regeneration. In Jin668 at the EC stage, we found upregulation of the TFs BREVIS RADIX (BRX) and MYC/MYB when compared to the NEC stage. In a study by Mouchel et al., their research team demonstrated that the BRX gene controlled cell proliferation and elongation in the growth zone of root tips in Arabidopsis [62]. Thereafter, BRX was found to be involved in regulation of brassinosteroid biosynthesis and maintenance to keep brassinosteroid biosynthesis above a critical threshold, which ultimately affects lateral root development in Arabidopsis [63].
We also found upregulation of 12 GATA genes which indicates that they play an important role in embryo initiation in cotton. GATA TFs are widely distributed in fungi, animals, and plants consist of a protein family containing a DNA-binding domain recognizing the DNA consensus sequence (A/T) GATA(A/G) and one or two highly conserved type-IV zinc fingers (C-X2-C-X17-20-C-X2-C) [64]. In plants, GATA TF expression is involved in light-mediated processes such as flowering, maturation, petal differentiation and expansion, and embryo development [65]. In the Gossypium genus, 30 GATA TFs have been identified [64]. Moreover, Nawy et al. reported that a GATA transcription factor, HANABA TARANU (HAN) was required to position the inductive Arabidopsis pro-embryo boundary and revealed that HAN regulated transcription in the basal pro-embryo [66].
Previous studies suggest that GRAS transcription factors act as essential regulators, not only in polar development [67] and reprogramming [68], but also in response to biotic and abiotic stresses [69,70], and auxin response [71,72]. Stuurman et al. reported that HAM (mutant hairy meristem) is a gene encoding a putative transcription factor of the GRAS family and mediates the signal from differentiating cells. This mechanism controls signals from differentiating tissues that extrinsically control stem cell fate in the shoot apex [59]. We found 11 GRAS genes upregulated in EC, suggesting that the upregulation of these genes is associated with embryo initiation and formation during somatic regeneration in cotton.
Previous work has shown that transient activation or ectopic expression of one or several transcription factors can trigger the transition from NEC to EC or increase the frequency of EC formation. These include WOX5 [73], WUSHEL [74], WRKY2 [75], GRD/RKD4 [76], BBM [40], LEC1 [54], FUS3 [77], ABI3 [78], and AGL15 [79]. For example, the AP2/ERF TF BBM in plants induced plentiful cell regeneration and was used to produce a large number of somatic embryos that could developed to seedlings [80]. Furthermore, ectopic expression of BBM derived from Brassica napus was transformed into pepper and obtained efficiently regenerated transgenic plants from recalcitrant sweet pepper (C. annuum) varieties [42]. Our data revealed dramatic upregulation of GRD/RKD4, BBM, LEC1, FUS3, ABI3 and AGL15 genes, both in A-and D-Subgenome in EC during the somatic embryogenesis. We also found four genes annotated as WRKY2 that were upregulated in EC with subtle fold changes (1.18 to 1.95 logFC). Interestingly, only one of the four copies of WUS (Gohir.A10G0998300), was found up-regulated in EC callus, while the other three genes were not expressed, suggesting this homeologous subgenome specific gene is essential for somatic embryogenesis in cotton. Moreover, unlike the gene mentioned above, two WOX5 genes, Gohir.A10G233000 and Gohir.D10G245300, were found down-regulated in EC callus (Table 4). We hypothesize that this WOX5 transcription factor may be critical in triggering the embryo initiation, while the other transcription factors contribute as regulatory or maintenance factors that are critical to embryo formation and development. Furthermore, this homeobox gene was reported to contribute to signaling spiral differentiation of stem cells in Arabidopsis thaliana; while loss of WOX5 function in the root meristem stem cell niche caused terminal differentiation in distal stem cells [81]. Our data also indicates that a series of related transcription regulators are involved in signaling of embryo initiation and formation in cotton.
Phytohormone sensing is critical to cell fate reprogramming during somatic embryogenesis. Our results showed that during conversion from NEC to EC, a series of genes involved in the biosynthesis and transportation of auxin and cytokinin are upregulated, which indicates that these two phytohormones play important roles during the initiation of the embryos.
Auxins are key components to regeneration, and their exogenous application has shown to recuperate the embryogenic potential of mitotically quiescent somatic cells [82]. In Triticum aestivum, regeneration efficiency was associated with auxin exposure time and catalase metabolism during somatic embryogenesis. Induction of pluripotent cells, termed, "callus," by auxin represents a typical cell fate change required for plant in vitro regeneration [83]. Studies have shown that relative signaling through MONOPTEROS (MP)/ARF 5 was required for shoot formation in Arabidopsis callus. Moreover, variants of MP expression revealed that this gene can promote de novo shoot formation in tissues that are normally recalcitrant [84]. Cytokinin also plays a fundamental role in cell fate reprogramming and is implicated in the process of shoot organogenesis [85]. Cytokinin promotes shoot regeneration through the upregulation of WUS, and recent studies demonstrated that key cytokinin signaling components, namely type-B ARABIDOPSIS RESPONSE REGULATORSs (ARRs), directly bind the WUS promoter and regulate transcription of this gene [86]. These studies also suggest that the transcriptional regulation of early embryo patterning to hormonal control of plant callus cell initiation is comparable to similar strategies for stem cell formation in the animal kingdom [86]. Furthermore, it was found that the WUS expression is controlled by the ratio of cytokinin with auxin [87]. Our results show approximately 3 times more auxin related genes being upregulated in EC than cytokinin genes (129 vs. 42), suggesting a critical role for these genes in somatic embryogenesis in cotton.

Conclusions
Understanding endogenous changes in developmental biology at the molecular level is paramount to developing strategies toward genotype independent transformation and subsequent whole plant regeneration. Here we describe the transcriptional landscape of two critical developmental stages of somatic embryogenesis, NEC cells and EC cells in a highly regeneratable cotton genotype, Jin668. Our study identified genes in this non-recalcitrant genotype that are differentially expressed and categoriezed them according to various biological processes that are important in somatic embryogenesis, such as phytohormone signaling, transcription factors and signaling cascads, and regulation of DNA methylation. We identified a list of candidate genes that will be important to follow up functional studies that characterize their functional role in enabling embryo formation in cotton. Optimizing transformation and regeneration systems is a large stride toward realizing the promises of synthetic biology and genome editing. We anticipate that this work will lead to new understandings of somatic embryogenesis in upland cotton and other dicot species.

Genotype and plant material source, callus, and sampling
The formal identification of the plant material used in this study was performed by Dr. Shuangxia Jin of Huazhong Agricultural University. The nonrecalcitrant plant genotype used in this study, Jin668 (Gossypium hirsutum L.), was developed through successive regeneration acclimation (SRA) and is described in detail in [88]. Jin668 seeds (voucher specimens) are available through Material Transfer Agreement (MTA) with Dr. Shuangxia Jin of Huazhong Agricultural University. Seeds were geminated in April of 2018 at the Clemson University Systems Genetics lab using the following procedure: Seeds were sacrificed by sulfuric acid for 1 min, then rinsed under running deionized water for 2 h. The seeds were washed with 70% ethanol for 1 min and then washed three times with sterile deionized water. The seeds then were sterilized by shaking them in a flask containing 100 mL of 10% commercial bleach (+ 2 drops of Tween-20) for 5 min and rinsed three times with sterile DW. Finally, they were cultured in Magenta boxes on germination media (4.33 g/L Murashige and Skoog (MS) salts (MS basal salts mixture; PhytoTechnology Laboratories, cat. no. M524), 2% glucose (PhytoTechnology Laboratories, USA), pH 6.0, 0.26% Phytagel (PhytoTechnology Laboratories, USA) at 28°C, under kept in dark conditions for 7 days. The 7-day hypocotyls were cut to 0.5-1.0 cm segments and placed on callus induction media [89], with media changes every 4 weeks. The plates were cultivated in a growth chamber with controlled conditions of 28 ± 1°C, 16 h (day)/8 h (night) photoperiod, with light provided by cool-white fluorescent lamps at an irradiation of 135 μmolm − 2 s − 1 , and 50% relative humidity. "The regeneration rates were calculated by the number of explants producing embryogenic callus of 300-330 explants (hypocotyl segments) within the 1.5-month tissue culture process, with 3 replicates.
Two developmental stages were selected for transcriptome profiling based on callus morphology, nonembryogenic cells and embryogenic cells (Fig. 1). The non-embryogenic cells were cells that lacked visual queues of organization or signs of polarization. Embryogenic cells showed visual signs of differentiation and were considered to be at globular-stage. In an effort to minimize gene expression background signals that may originate from different explants from different petri dishes, samples for each developmental stage were prepare as follows: Either non-embryogenic callus (callus that were completely NEC or had both NEC and EC) or embryogenic callus cells were harvested with a sterile scalpel from 20 explants separately and combined into a single tube. This process was repeated three times to collect 3 biological replicates.

RNA isolation and RNA sequencing
Total RNA was isolated from Jin668 callus material at different stages using a modified guanidine thiocyanate method [90]. RNA-Seq libraries were constructed using the Illumina TruSeq Stranded RNA kit (San Diego CA, USA) following the manufacturer's recommendations with three biological replicates per condition. Paired-end sequences were collected on an Illumina NovaSeq through a third-party vendor. Raw sequence files were trimmed of low quality bases and adapter sequences with the trimmomatic software package [91].

Identification of differentially expressed genes and enrichment analysis
Preprocessed reads were aligned to the Gossypium hirsutum v2.0 reference assembly (Chen et al., 2020) with the Bowtie2 short read aligner [92], and alignment files were coordinate sorted and indexed with samtools v1.3.1 [93]. Raw counts, FPKM and TMM values were generated with RSEM [94]. Counts for each replicate were collected in tabular format and normalized according to biological replicates to remove sample-specific effects with EdgeR [95]. Relative pairwise changes in gene level expression were made using a generalized linear model (glm) and classic pairwise comparison methods with the EdgeR package [95]. Genes abounding a p-value of ≤0.01 were corrected by the false discovery rate (FDR) of 0.05 [96]. Gene level fold-change values were output in tabular format and candidate genes abounding assigned thresholds were clustered into functional pathways with the GOSeq software tool [97] and filtered by error corrected P-values ≤0.05. Hierarchal clustering was performed by manually subsetting genes whose logFC was ≥ 2 and FDR corrected p-values were ≤.001. A matrix of log transformed gene expression values were input into the fastcluster R software package and clusters identified by cutting the tree at 50% max height.

Gene validation through RT-qPCR
Quantitative real-time PCR was performed to verify the RNA-Seq data [55]. Three biological replicates of total RNA of cotton callus for RNA-seq were used for RT-qPCR. Briefly, 0.5-1 μg of total RNAs were reversetranscribed to first strand cDNAs using M-MuLV reverse transcriptase (New England Biolabs, GA, USA) and primed by d(T)25-VN following manufacturer's instructions. RT-qPCR was conducted on an iCycler iQ system (Bio-Rad, Hercules, CA, USA) in 20 μl of PCR reaction solution (SsoAdvanced™ Universal SYBR® Green Supermix, Bio-Rad, USA) with 100 nM of each primer; the input amount of the first strand cDNA per reaction is around equivalent to 0.6-1.2 ng of total RNA. Four technical repeats were used for each of the three biological replicates. PCR was conducted with the following program: an initial denature at 95°C for 60 s, followed by 40 cycles of 95°C for 20 s, 62°C for 20 s, and 72°C for 20 s. Finally, a melting curve was performed from 55.0°C to 95.0°C at 0.5°C increments. The ΔΔCt method was used for real-time PCR analysis. Two reference genes, GhPP2A1 [98] and GhUBQ7 [88] were used as endogenous controls. Relative expression level was calculated using the 2-ΔΔCt formula. All primer pairs except for the reference genes used for examining the expression levels of cotton endogenous genes were designed based on the Jin668 RNAseq sequences and listed in Additional file 12: Table S12.

Authors' information
All information is provided on the title page.

Funding
This work was supported by NIFA/USDA hatch funds under project number SC-1700530 and Technical Contribution No. 6867 of the Clemson University Experiment Station. This work was also supported by the National Science Foundation award number 1739092 and Cotton Incorporated project number 2013686. The funding bodies had no role in designing the study, collection, analysis, and interpretation of the data, or writing of the manuscript.

Availability of data and materials
The transcriptomic datasets generated during the current study are available undert the BioProject accession number PRJNA629328 in NCBI (https:// submit.ncbi.nlm.nih.gov/subs/sra/SUB7346009/overview), and in the supplementary information files; Table S1. Differential gene expression profile of NEC vs. EC in Jin668. The logFC value is the observed result in EC; Table S2: Genes uniquely expressed in either NEC or EC cells; Table S3. Gene Ontology enrichment of genes upregulated in EC (> = 1 fold change); Table S4. Gene Ontology enrichment of genes downregulated EC (> = 1 fold change); Table S5. Centered log2(fpkm+ 1) gene expression clusters of NEC and EC callus; Table S6. Subcluster 1 GO enrichment; Table S7. Subcluster 2 GO enriochment; Table S8. Subcluster 3 GO enrichment; Table S9. Phytohormone genes expressed in NEC and EC cells in Jin668; Table S10. Transcription factors with differential expression profiles in NEC and EC callus; Table S11. Genes involved in DNA methylation regulation in NEC and EC callus; Table S12. Selected genes and primer sequences for qPCR expression validation.
Ethics approval and consent to participate Not applicable.

Consent for publication
Not applicable.