Background

Cotton is a major source of natural fibers used in the textile industry. Cotton fibers are single-celled trichomes that emerge from the ovule epidermal cells. About 25–30 % of the seed epidermal cells differentiate into spinnable fibers [1]. Lint fibers of the economically important Gossypium hirsutum generally grow about 35 mm in length. Cotton fiber development consists of four distinct but overlapping stages, including fiber initiation, elongation, secondary cell wall (SCW) biosynthesis, and maturation [2]. Fiber elongation starts on the day of anthesis and continues for about 3 weeks before the cells switch to intensive SCW cellulose synthesis. During peak elongation fiber cells can increase in length at rates of 2 mm per day or more depending on environment and genotype [1–3]. The rate and duration of each developmental stage are important to the quality attributes of the mature fiber. Cell elongation is crucial for fiber length, whereas SCW is important for fiber fineness and strength. Understanding the molecular basis of fiber elongation would provide a means for cotton breeders and researchers to improve the fiber length while maintaining yield and other fiber characteristics.

Genetic mutants are useful tools for studying the molecular mechanisms of fiber development. Our laboratory uses two short fiber mutants, Ligon lintless-1(Li 1 ) and Ligon lintless-2 (Li 2 ) as a model system to study fiber elongation [4–10]. Both Li 1 and Li 2 are monogenic and dominant mutations, resulting in an extreme reduction in the length of lint fiber to approximately 6 mm on mature seeds [11, 12]. Both mutations are located in the DT subgenome of G. hirsutum: the Li 1 gene is on chromosome 22 [6, 13, 14], whereas the Li 2 gene is on chromosome 18 [7, 14–16]. Cytological studies of cotton ovules did not reveal much difference between mutants and their near-isogenic WT lines during initiation and early elongation up to 3 days post anthesis (DPA) [7, 13]. In a fiber developmental study Kohel and co-authors observed that the elongation pattern is similar and restricted in both, Li 1 and Li 2 fibers [17]. However, unlike the normal morphological growth of the Li 2 plants, the Li 1 mutant exhibits pleiotropy in the form of severely stunted and deformed plants in both the homozygous dominant and heterozygous state [6, 11, 12]. The near-isogenic lines (NILs) of Li 1 and Li 2 with the elite Upland cotton variety DP5690 previously used in our research [6, 7] provide an excellent model system to study mechanism of fiber elongation.

Micro RNAs (miRNAs) are a class of non-coding endogenous small RNA that post transcriptionally regulate target genes expression [18]. Plant miRNAs range in size from 20 to 24 nucleotides. They negatively regulate gene expression by either mRNA degradation or translation inhibition [19–22]. miRNAs play important roles in most biological processes such as development, cell proliferation, stress response and metabolism [23–26]. Recently, identification and characterization of miRNA involved in fiber initiation and development have attracted much attention. For example, a recent study shows that miR828 and miR858 might regulate homeologous MYB2 (homologous to Arabidopsis GL1) gene functions in cotton fiber development [27]. Another study revealed that miR156/157 family plays an essential role in fiber elongation; suppressing expression of miR156/157 resulted in the reduction of mature fiber length [28]. However, currently the mechanism underlying the miRNA-mediated regulation of fiber development is largely unknown. Therefore, using short fiber mutants for identification and analysis of new miRNAs may provide a new insight in fiber development process.

In this work small RNA libraries from developing fiber cells of short fiber mutants and wild type were sequenced. Gossypium hirsutum TM-1 genome was used for miRNA structural prediction. We identified 24 conservative and 147 novel families whose targets were confirmed through degradome sequencing. Expression levels of 20 miRNAs families were extensively tested through the fiber development time course in wild type and mutant plants. Correlation analysis between expression levels of miRNAs with fiber length of 11 diverse cotton cultivar revealed 4 miRNAs that were significantly correlated with fiber length.

Results

Deep sequencing of small RNA libraries from developing fibers of short fiber mutants and wild type

We identified and tested expression level of miRNAs in rapidly elongating cotton fiber cells (at 8 DPA) of Li 1 , Li 2 mutants and WT. The time point 8 DPA was selected because our earlier research revealed significant transcript and metabolite changes between the short fiber mutants and their WT NIL during this time of fiber development [7, 8, 10]. Three small RNA libraries were constructed and sequenced from developing fibers of Li 1 , Li 2 and WT. The small RNA sequencing data were deposited into the National Center for Biotechnology Information (NCBI) with accession PRJNA307581. A total of 14,894,333, 18,911,773 and 11,143,698 mappable reads (about 80 % of total raw reads) were obtained from WT, Li 1 and Li 2 fiber cells, respectively. Figure 1 represents a flow chart of data processing to identify miRNAs and their targets. Mappable reads were run through miRPlant software for identification of plant miRNA from RNAseq data. Gossypium hirsutum TM-1 genome was used for mapping reads and prediction of hairpin structure of precursor’s miRNAs. A total of 9,080 miRNA sequences were predicted in three libraries (Additional file 1).

Fig. 1
figure 1

Flow chart of data processing to identify miRNAs and their targets in rapidly elongating cotton fiber cells transcriptome

Full size image

Identification of conserved, previously reported and candidate miRNAs

We used the term “conserved” for miRNAs present in multiple species throughout at least one major ancient clade of land plants. To identify conserved and previously reported miRNAs, the 9,080 predicted miRNA sequences were BLAST searched against the miRNA sequences deposited in the miRBase release 21 [29]. The criteria of the BLAST search required no more than two mismatches with the sequences in miRBase. In total, 405 predicted miRNA sequences were identified as conserved or previously reported in miRBase release 21. The 405 redundant sequences were clustered into 24 known miRNA families (Fig. 2). Eight deeply conserved families were present across all land plants species; 2 families were present across vascular plants; 5 families were present across seed plants; 5 families were present across flowering plants; miR2950 was detected in eudicots and 3 families were previously reported in cotton. Six families, including miR159, miR165/166, miR167, miR168, miR390 and miR482, were the most abundant whose normalized expression levels were more than 30,000 reads per million (rpm) across 3 libraries (Table 1). After removing conserved and known miRNAs, the remaining sequences were clustered into 1726 candidate miRNA families. Additional file 2 provides information on loci of the candidate miRNA sequences in the TM-1 genome of and count of reads in each library.

Fig. 2
figure 2

Deeply conserved and previously reported miRNA families detected in small RNA libraries from developing cotton fibers of WT, Li 1 and Li 2 . miRNA families (columns) are conserved between plants families (rows) for plant species represented in miRBase release 21 [29]. Boxes are highlighted if a miRNA family was identified in at least one species for each plant families listed

Full size image
Table 1 Target identification with degradome of highly expressed miRNAs
Full size table

Target identification with degradome and selection of novel miRNAs

Given the high false-positive rate of computational predicted targets experimental confirmation of these targets is an important step. Degradome sequencing provides a high-throughput strategy for the global experimental identification of targets for miRNAs [30–32]. The degradome library was constructed from pooled RNA samples isolated from developing fiber cells at 8 DPA of WT, Li 1 and Li 2 . The degradome reads were deposited into the NCBI with accession PRJNA307581. After removing low quality reads 6,901,161 mappable reads were obtained. We used the automated plant-compatible pipeline software CleaveLand4 to facilitate the interpretation of degradome data [33]. G. hirsutum TM-1 cDNA sequences were used for mapping degradome data (Fig. 1). Only miRNA targets with p-value ≤ 0.05 were used in further analyses. Candidate miRNA families whose cleaved targets were detected in degradome data were selected as novel miRNAs in this study. There were 147 novel miRNA families that met these requirements (Additional file 3).

In total 157 non-redundant targets, including 36 targets of 24 known miRNAs and 121 targets of 147 novel miRNA families, were identified with degradome (Additional file 3). The target genes of known or predicted function were sorted into the 12 functional categories based on functional catalogues established for Arabidopsis [34]. The distribution of the target genes into functional categories is represented in Fig. 3. The largest set of genes (21 %) was assigned to the transcription category. The most abundant transcription factors (TFs) were auxin responsive, AP2, GRAS, and TCP (Additional file 3). Genes involved in protein biological processes and cell structure functional categories formed the second (17 %) and the third (16 %) largest groups (Fig. 3). Genes encoding structural constituent of cytoskeleton such as tubulin and actin were the most abundant (60 %) members of the cell structure functional category.

Fig. 3
figure 3

A pie chart showing functional categories of the target genes detected in degradome of rapidly elongating fibers of wild type and short fiber mutants

Full size image

Expression profiling of differentially expressed miRNAs in developing fibers of WT and short fiber mutants

Expression levels of 11 conserved, 2 previously reported and 7 novel miRNA were tested by RT-qPCR in cotton fiber cells at 7 developmental time points (0, 3, 5, 8, 12, 16 and 20 DPA) of WT and short fiber mutants. The 20 miRNAs for RT-qPCR analysis were randomly selected from the Table 1. Expression profiles of miRNAs significantly differentially expressed between wild type and both mutants at multiple time points are shown in Fig. 4, whereas Additional file 4: Figures S1 and Additional file 5: Figure S2 provide RT-qPCR data for the rest of tested miRNAs. Overall, the majority of tested miRNAs exhibited higher expression levels during initiation (Day of anthesis, DOA) or transition to SCW deposition 16 – 20 DPA developmental stages. However, two identified miRNAs showed the elongation stage related pattern in short fiber mutants with transcript abundance decreasing at the beginning of the SCW stage from 16 – 20 DPA. The transcript abundance of Novel-7 (N7) miRNA was significantly increased (7.1-fold) in Li 1 and (6.0 – 9.6-fold) in Li 2 fiber cells at 8 – 12 DPA, consequently (Fig. 4). The transcript level of miR160 was significantly increased (3.2 – 4.9-fold) at 8 – 12 DPA, consequently, only in Li 1 fiber cells (Additional file 4: Figure S1). The expression level of miR164 was significantly increased at 3 – 5 DPA in both mutants and at 8 – 16 DPA only in Li 1 developing fibers. N1 and N4 miRNAs were up-regulated in mutants at the elongation (8 – 12 DPA) and the beginning of the SCW stages.

Fig. 4
figure 4

RT-qPCR expression analysis of miRNAs and their potential targets in developing cotton fiber cells of wild type and mutants. Each section represents expression profiles of miRNA and its target. The relative expression level is shown on the left y-axis of each graph. Asterisks indicate significant (p-value < 0.05) difference in gene expression level between mutant and wild type. Asterisks on x-axis represent significant difference in gene expression between wild type and both mutants, while asterisks on top of expression bars represent significant difference in gene expression between only one mutant line (bar with asterisk) and wild type. Error bars indicate standard deviation from 3 biological replicates

Full size image

Expression of target genes

To assess the influence of the miRNAs on their targets expression, we tested RNAseq expression of all target genes at peak of elongation (8 DPA), and evaluated by RT-qPCR the expression patterns of several selected target genes across different fiber developmental stages.

To test the expression level of target genes we used RNAseq data from 9 libraries previously reported [10]. RNAseq libraries were constructed from developing fibers at 8 DPA of Li 1 , Li 2 and WT in three biological replicates. Normalized expression data for 157 target genes, including least squares means, log2 ratios of comparisons mutants vs. wild type and p-values are provided in the Additional file 3. Of the 157 genes, 38 were significantly down regulated in Li 1 and 21 in Li 2 fiber cells and 15 were down regulated in both mutants (Fig. 5). Among the down-regulated genes in both mutant lines were plasma membrane intrinsic protein, 3-ketoacyl-CoA synthase, two eukaryotic aspartyl proteases, and tubulin. Among 23 down-regulated genes in Li 1 were six tubulin genes, two 3-ketoacyl-CoA synthases, galactosyl transferase, and pectin lyase. Among 6 genes down-regulated specifically in Li 2 were GRAS family TF and proline-rich cell wall protein. The number of genes significantly up-regulated in fibers of Li1 and Li2 were 7 and 6, respectively. Two genes of unknown function were up-regulated in fibers of both mutant lines (Fig. 5).

Fig. 5
figure 5

Venn diagrams of significantly down-regulated (left) and up-regulated (right) target genes in developing cotton fibers at 8 DPA of wild type and mutants. The total amount of down-regulated or up-regulated genes indicated in square brackets

Full size image

The transcript patterns of target genes of 6 miRNAs significantly up regulated in mutants fiber cells at multiple time points were evaluated by RT-qPCR (Fig. 4). TFs MYB33 (Gh_A05G3434), NAC (Gh_D11G0347) and ARF6 (Gh_A12G0813) were identified as targets of miR159, miR164 and miR167, respectively. Plasma membrane intrinsic protein 2;8 (PIP2;8, Gh_A01G0019) was identified as target of miRNA N1, whereas Glycine-rich RNA binding protein (Gh_D13G2550) and 3-ketoacyl-CoA synthase (Gh_D01G1810) were identified as targets of miRNAs N4 and N7, respectively. All tested targets revealed a negative relationship with corresponding miRNAs expression patterns (Fig. 4). The most interesting correlation pattern was observed between miRNA N7 and its target 3-ketoacyl-CoA synthase. The highest increase in transcript abundance of miRNA N7 was during peak of elongation (8 – 12 DPA), which corresponded to the lowest transcript abundance of 3-ketoacyl-CoA synthase during the same time points in mutants’ fiber cells.

Correlation analysis between miRNA expression and fiber length

To investigate the affects of miRNAs expression levels on cotton fiber properties, we conducted correlation analyses between miRNAs transcript abundance and fiber lengths of 11 diverse Upland cotton lines. Fiber quality measurements were collected during 3 years (2009–2011) of growing plants in Starkville, MS field. For this study we used only measurements of the fiber length (upper half mean length, UHML). RNA for RT-qPCR expression analysis was collected from plants growing in Stoneville field in summer of 2015. Twenty above-described known and novel miRNAs identified in the study were used for correlation analyses. The correlation results, including Pearson coefficients and p-values, are represented in Fig. 6a. The expression patterns of 4 miRNAs, including N7, N4, N1 and miR167 revealed significant (p-value < 0.05) negative relationships with the fiber lengths of the 11 cotton lines. The transcript levels of targets of those 4 miRNAs were evaluated in the 11 cotton lines. As shown in Fig. 6b the transcript abundance profiles of targets revealed positive correlations with fiber length. The internal cleavage sites in the predicted targeted genes of N7, N4, N1 and miR167 miRNAs were confirmed by RNA ligase-mediated rapid amplification of 5’ and 3’ cDNA ends (RLM-RACE). The sequencing result of ten independent RACE fragments matched the site predicted by degradome analysis (Additional file 6: Figure S3).

Fig. 6
figure 6

Negative correlation between miRNA expression level and fiber length of eleven diverse Upland cotton lines. The length of cotton fiber is shown on the left axis of each graph, whereas relative expression level of miRNAs (red line) and corresponding target (black line) is shown on the right axis of each graph. UHM, upper half mean length, the average length of the longer one half of the fibers sampled. Error bars for fiber length represent standard deviation between mean values from 3 years (2009–2011) of fiber measurements. Error bars for miRNAs or target genes expression represent standard deviation between 3 biological replicates

Full size image

Discussion

The causative mutations of the cotton short fiber mutants Li 1 and Li 2 have yet to be identified. In this study we analyzed the small RNA libraries constructed from developing fibers of the two short fiber mutants and their WT NIL to examine possible regulation of genes by miRNAs during fiber elongation. We identified 24 conservative and 147 novel miRNA families in those small RNA libraries. Different phenotypic changes caused by the Li 1 (dwarf deformed plants and short fiber phenotype) and the Li 2 (short fiber phenotype only) mutations suggested that mutated loci are different types of genes. Expression profiles of a number of tested miRNAs were different in short fiber mutants than in WT during fiber development (Fig. 4, Additional file 4: Figures S1 and Additional file 5: Figure S2). These results suggested that the mutations changed the regulation of miRNAs expression during fiber development. Further investigations of differentially expressed miRNAs in the Li 1 – Li 2 mutants will contribute to better understanding of the regulatory mechanisms of cotton fiber development.

miRNAs negatively regulate gene expression by either mRNA degradation or translation inhibition. Previously, plant miRNA targets have been studied via computational prediction, which is based on sequence complementarity between miRNA and the target mRNA. Usually such predictions give high false-positive rates. Recently, a new method called degradome sequencing has been successfully established to screen for miRNA targets in plants. Degradome sequencing provides a high-throughput strategy for the global identification of small RNA-directed target cleavage by sequencing the 5’ ends of uncapped RNAs [30–33, 35]. In this study we have described only targets detected by degradome sequencing.

A large proportion of detected targets were TFs (21 %, Fig. 3), which was consistent with previous reports in cotton [36–38]. An essential role in fiber elongation has been illustrated for miRNA156/157 (targets SBP domain TF) in Gossypium barbadense [28]; however, our results demonstrated no significant difference in transcript abundance of miR156/157 in developing fibers of the short fiber mutants compared to WT fibers (Additional file 5: Figure S2). Therefore, miR156/157 most likely is not involved in the truncated fiber elongation caused by Li 1 or Li 2 mutations. In Arabidopsis, miR159 mediates cleavage of GAMYB-like genes that encode R2R3 MYB domain TFs that have been implicated in gibberellin signaling in anthers and germinating seeds [39]. Our results revealed miR159 significantly induced in short fiber mutants during elongation and SCW deposition stages (Fig. 4), suggesting involvement of miR159 in regulation of elongation. Previous studies in Arabidopsis demonstrated that the miR164 family guide the mRNA cleavage of five NAC TF genes that are required for boundary establishment and maintenance, lateral root emergence, formation of vegetative and floral organs, and age-dependent cell death [25, 40–42]. In this study, NAC domain TF (Gh_D11G0347) was among the predicted targets of miR164 (Table 1). Transcript level of miR164 was significantly up-regulated in the Li 1 and Li 2 fiber cells while its target NAC TF was down-regulated in mutant fibers (Fig. 4), suggesting the potential regulatory role of miR164 in fiber development.

There are very few studies that have performed a comparison of miRNAs expression profiles among different cotton genotypes with the exception of stress-related studies [43, 44]. So far only one report explored regulatory role of miRNAs in genotype-dependent traits in cotton [45]. In that study, the authors demonstrated that miRNAs have different expression patterns in different cotton varieties, which implicated their different phenotypic traits. In the current study we assessed whether miRNAs expression may be involved in the regulation of fiber length. Of the 20 tested miRNAs, 4 of them including miR167, N1, N4 and N7 had significant negative correlations with fiber lengths of 11 diverse Upland cotton lines (Fig. 6). The majority of fiber miRNA studies have focused on miRNA identification and expression analysis as well as target prediction and validation [37, 38, 46–49]. In addition to these observations the present study introduced correlation analysis with fiber lengths of commercially important cotton varieties that can be useful for plant molecular breeding. Our data demonstrated these 4 miRNA families were expressed more highly in the short fiber mutants at multiple time points of fiber development compared to the WT NILs (Fig. 4).

Auxin Response Factors (ARFs) were among the predicted targets of miR167. These proteins bind to the auxin response elements in the promoter regions of numerous early auxin-inducible genes [50, 51]. Exogenous auxin is required to promote fiber cell development from unfertilized ovules in culture [52]. Genetically engineered increase of auxin level in the epidermis of cotton ovules at the fiber initiation stage substantially increased the number of lint fibers and consequently fiber yield [53]. In Arabidopsis miR167 controls ARFs 6 and 8 expression patterns and affects the fertility of ovules and anthers [54]. Transgenic tomato plants over-expressing miR167 exhibited reductions in leaf size and internode length as well as shortened petals, stamens, and styles [55] that may cause infertility. In a comparative study of small RNA abundance between the wild-type and fuzz/lintless mutant, the expression of miR167 was significantly up-regulated in mutant fibers compared to the wild type [37]. A critical role of miR167 in cotton fiber elongation was suggested in another study exploring small RNA expression in developing fiber cells from 5 to 20 DPA [46]. Our data demonstrated that miR167 was significantly up-regulated in the Li 1 and Li 2 fiber cells while its target, ARF6 (Gh_A12G0813), was down-regulated in mutant fibers (Fig. 4). Also expression level of miR167 and its ARF6 target correlated with the fiber length of 11 cotton varieties (Fig. 6). This suggests that miR167 might be involved in regulation of fiber elongation.

Using correlation analysis we detected 3 more novel miRNAs which might regulate fiber length. Plasma membrane intrinsic protein PIP2;8 (Gh_A01G0019) was detected as a target of novel miRNA N1 by degradome analysis (Table 1). PIPs constitute a plasma-membrane specific subfamily of major intrinsic proteins or aquaporins which are associated with water transport and play important roles in fiber elongation. In our previous study, RNAseq analysis revealed that aquaporins were one of the most significantly over-represented gene families among down-regulated genes in Li 1 and Li 2 fibers [10]. The higher concentrations of inorganic ions detected in saps of fiber cells of Li 1 – Li 2 provided indirect evidence of reduced influx of water into fiber cells due to low expression of aquaporins and consequently a reduction in fiber cell elongation in mutants [10]. Our data have shown that the target of N1 miRNA Gh_A01G0019 is highly expressed during fiber elongation from 5 – 16 DPA in wild type fiber and exhibited greatly reduced expression in short fiber mutants (Fig. 4). The gene product of Gh_A01G0019 has 95 % amino acid sequence identity to a previously characterized PIP (PIP2;4), and it was shown that suppression of PIP2;4 expression by RNA interference markedly slowed down fiber elongation [56]. Therefore novel miRNA N1 represents a good candidate gene for further investigation of its role in regulation of PIP2;8 and fiber elongation.

The degradome detected target of novel miRNA N4 was glycine-rich RNA binding protein Gh_D13G2550. This gene has not been characterized in cotton and shows 82 % amino acid identity to Arabidopsis glycine-rich RNA binding protein 7 (AtGRP7). The AtGRP7 protein is regulated by circadian clock [57], involved in response to cold stress [58], and pathogen defence [59]. Transgenic Arabidopsis plants ectopically expressing AtGRP7 showed a dwarf phenotype due to distortions in gibberellin biosynthesis [60]. Our data have demonstrated that miRNA N4 transcript abundance was significantly higher in short fiber mutants during elongation – SCW deposition (8 – 20 DPA). The transcript abundance of its target was significantly reduced during the same period of time in mutants (Fig. 4). Therefore the miRNA N4 is another candidate for further investigations of its involvement in regulation of fiber elongation.

The novel miRNA N7 showed the most significant correlation probability with fiber length among tested miRNAs (Fig. 6). The degradome detected target of miRNA N7 was 3-ketoacyl-CoA synthase (KCS, Gh_D01G1810). KCS catalyses the initial condensation reaction during fatty acid elongation using malonyl-CoA and long-chain acyl-CoA as substrates [61]. Very long chain fatty acids significantly promoted cotton fiber cell elongation with several KCS genes highly up-regulated during cotton fiber development [62]. Our data have shown that the expression pattern of miRNA N7 revealed negative relationship with its target since the transcript abundance of KCS Gh_D01G1810 was highly increased during cotton fiber elongation (5–16 DPA) and significantly decreased in short fiber mutants (Fig. 4). This suggests that miRNA N7 might be involved in regulation of fiber elongation by targeting KCS.

Conclusions

We identified 24 conservative and 147 novel miRNA families in small RNA libraries isolated from fiber cells of the cotton short fiber mutants Li 1 and Li 2 and their respective WT near-isolines. Fiber gene expression analysis of 20 selected miRNAs revealed differences in the expression profiles of short fiber mutants compared to WT during fiber development, which might reflect different transcript regulation in mutant lines comparing to WT fiber cells. Further investigations of these differentially expressed miRNAs will contribute to better understanding of the regulatory mechanisms of cotton fiber development. Of the 20 selected miRNAs, the expression patterns of 4 miRNA families showed significant correlations with fiber length of 11 eleven diverse Upland cotton lines. These miRNAs represent good candidates for further investigations of miRNA regulation of important genotype dependent fiber traits. The results of this study will contribute to further understanding of the role of miRNAs in cotton fiber development and will provide a tool for plant molecular breeding.

Methods

Plant materials

Two mutant lines Li 1 and Li 2 in a near-isogenic state with the WT Upland cotton line DP5690 were developed in a backcross program at Stoneville, MS as described before [6, 7]. A total of 150 Li 1 , 100 Li 2 , and 100 WT plants were grown in a field at the USDA-ARS Southern Regional Research Center, New Orleans, LA in the summer of 2013. First position flowers were tagged on DOA and bolls harvested at 0, 3, 5, 8, 12, 16, and 20 DPA. Bolls were randomly separated into 3 replicates with about 15–30 bolls per replicate. The ovules were carefully excised, immediately immersed in liquid nitrogen and stored at 80 °C.

Eleven diverse Upland cotton lines from across the United States were used for correlation analysis to compare the expression levels of miRNAs and the length of fiber. Lines were ‘Acala Ultima’, developed by California Planting Cotton Seed Distributors (Shafter, CA); ‘Tamcot Pyramid’, developed in the Multiple Adversity Resistant program by the Texas Agriculture Experiment Station, College Station. TX (Thaxton and El-Zik, 2004); ‘Coker 315’, developed by Coker Pedigreed Seed Co. (Hartsville, SC); ‘Stoneville 825’, developed by Stoneville Pedigreed Seed Co. (Stoneville, MS); ‘Fibermax 966’, developed by Bayer Crop Science (Lubbock, TX); M-240RNR, a root knot nematode resistant line developed by the ARS scientists Shepherd et al. [63]; ‘Paymaster HS-26’, a Texas High Plains cultivar developed by Paymaster Technologies, Inc. (Plainview, TX); ‘Deltapine Acala 90’, developed by Delta and Pine Land Co. (Scott, MS); ‘Sure-Grow 747’, developed by Sure-Grow Co. (Centre, AL) [64]; ‘Phytogen PSC 355’, developed by Mississippi Agriculture and Forestry Experiment Station (Mississippi State, MS) and licensed to Phytogen Seeds (Indianapolis, IN); and ‘Stoneville 474’, developed by Stoneville Pedigreed Seeds. Pedigrees for all except M-240RNR can be found in Bowman et al. [65].

For fiber quality measurements, seeds of the 11 cotton lines were planted as three replicates in a randomized complete block on the Plant Science Research Farm at Mississippi State, MS, in 2009, 2010 and 2011. Standard field practices were applied during the plant growing seasons. Twenty five healthy looking naturally opened bolls from the central part of a plant were hand harvested from each plant in 3 years. Boll samples were ginned on a 10-saw laboratory gin, and fiber properties were measured by Cotton Incorporated’s fiber measurement laboratory using a High Volume Instrument (HVI, USTER Technologies Inc., Charlotte, NC).

For RNA isolation from developing fibers seeds of 11 lines were planted in a field in Stoneville, MS, in 2015. Plants were in two- or four-row plots with ~70 plants per row. For the four-row plots, there were a total of three plots per cotton cultivar with each separate plot used as a biological replication. For the two-row plots there were a total of six plots with two plots used for each biological replication. First position flowers were tagged on the day of anthesis and bolls harvested at 10 DPA. Each biological replication consisted of about 8–15 bulked bolls. Once harvested, the bolls were placed immediately on ice in the field and transported to the laboratory. The bolls were dissected and the ovules with fibers attached were quickly frozen in liquid nitrogen and stored at −80 °C.

RNA isolation and RT-qPCR

Total RNA was isolated from detached fibers [66] using the Sigma Spectrum Plant Total RNA Kit (Sigma-Aldrich, St. Louis, MO) with the optional on column DNase1 digestion according to the manufacturer’s protocol. The concentration of each RNA sample was determined using a NanoDrop 2000 spectrophotometer (NanoDrop Technologies Inc., Wilmington, DE). The RNA quality for each sample was determined by RNA integrity number (RIN) using an Agilent Bioanalyzer 2100 and the RNA 6000 n Kit Chip (Agilent Technologies Inc., Santa Clara, CA) with 250 ng of total RNA per sample. RNA from each of the above mentioned time-points was used for RT-qPCR analysis. A detailed description of reverse transcription and qPCR for quantification of mRNA transcripts was previously reported [7]. 18S rRNA was used as the endogenous reference gene for relative quantitation of the gene expression data.

A protocol published by Cirera and Busk [67] was used to quantify miRNA transcripts. Briefly, 100 ng of total RNA was incubated with 1 μL of 10x reaction buffer of E. coli poly(A) polymerase (New England BioLabs Inc., Ipswich, MA), 0.1 mM dNTP, 0.1 mM ATP, 1 μM universal RT primer (5’-CAGGTCCAGTTTTTTTTTTTTTTTVN), 1 U of E. coli poly(A) polymerase (New England BioLabs Inc.), and 100 U of M-MuLV reverse transcriptase (New England BioLabs Inc.) in 10 μL reaction mixture for 1 h at 42 °C. The reaction was inactivated by heating at 95 °C for 5 min and cDNA was diluted 50 times before being used in qPCR. Micro RNA-specific primers were designed with the miRprimer software [68]. The qPCR reactions were performed with iTaq™ SYBR® Green Supermix (Bio-Rad Laboratories Inc., Hercules, CA) in a Bio-Rad CFX96 real time PCR detection system. 5.8S rRNA was used as the internal reference gene for normalization of RT-qPCR data. The relative expression levels were calculated using the 2-ΔΔCt [69]. Sequences of primers are listed in Additional file 7.

Small RNA sequencing and processing

Small RNA libraries preparation and sequencing were conducted by LC Science (Houston, TX). RNA samples from three biological replicates extracted from developing fiber cells at 8 DPA were pooled together for preparation of WT, Li 1 and Li 2 small RNA libraries, respectively. The small RNA libraries were constructed using 1 μg of total RNA according to the TruSeq® Small RNA sample preparation guide (Illumina, San Diego, CA). The general process is as follows: first, the total RNA was ligated to RNA 3’ and RNA 5’ adapters. Second, reverse transcription followed by PCR was performed to create cDNA constructs based on the small RNAs ligated with 3’ and 5’ adapters. Third, small cDNA fractions that range from 22 nt to 30 nt in length were isolated by using 6 % denaturing polyacrylamide gel electrophoresis. Fourth, cDNA construct was purified, and the library was validated. The libraries were sequenced using Illumina Hiseq 2500 platform.

Identification of conserved and novel miRNAs

Clean reads were run through miRPlant software for identification of plant miRNAs from small RNA sequencing data [70]. Gossypium hirsutum TM-1 genome [71] was used for mapping reads with software’s default parameters. Predicted by miRPlant miRNAs were BLASTed against the miRBase database (version 21, http://www.mirbase.org/) to identify conserved and previously reported miRNAs. Matched sequences with no more than two mismatches were considered as candidate conserved or previously reported miRNAs and were assigned to the corresponding miRBase family. Predicted miRNAs with more than 2 mismatches were considered as potential novel miRNAs because they lack sufficient similarity to assign to a miRNA family [72]. Statistical significance of differential expression of miRNAs in the sequencing data was established with the Audic & Claverie statistic using IDEG6 software [73, 74].

Degradome library construction, sequencing, data analysis, and target identification

A degradome library was constructed from pooled RNA samples isolated from developing fiber cells at 8 DPA of WT, Li 1 and Li 2 . The protocol is based on the method previously described by German et al. [35] and Addo-Quaye [30]. Briefly, poly(A)-enriched RNA was ligated to a 5’- RNA adapter with 3’ a EcoP15 I recognition site. Reverse transcription was performed to generate first-strand cDNA, followed by PCR amplification and EcoP15 I digestion. After digestion with EcoP15 I, a PAGE-gel was used to purify the EcoP15 I-cleaved fragments. The gel-purified products were ligated to a 3’-double-strand DNA adapter, followed by PAGE-gel purification to obtain the ligated products. PCR amplification was performed, and PAGE-gel was used for the third time to purify the corresponding gel bands containing the final products. Finally, the purified cDNA library was ready for deep sequencing. The library preparation and sequencing were conducted by LC Science (Houston, TX).

Single-end sequencing reads of 50 nucleotides were obtained using Illumina Hiseq 2500 platform. The adaptor sequences were trimmed from the raw reads and the reads shorter than 10 bases were excluded. Then these clean reads were mapped against the primary transcripts of Gossypium hirsutum TM-1 [71] using Bowtie [75]. A computational pipeline, CleaveLand [33], with its default parameters was used for the detection of cleaved miRNA targets from degradome data.

5’ RACE of miRNA cleavage

RLM-RACE was performed using the GeneRacer Kit (Invitrogen, Carlsbad, CA, USA) following the manufacturer’s instructions with minor modifications. Total RNA from developing fibers at 0, 3, 8 and 16 DPA was combined for mRNA isolation. Poly(A) mRNA was purified from total RNA using NucleoTrap mRNA Kit (Clontech, Mountain View, CA, USA) according to the manufacturer’s instructions. The GeneRacer RNA Oligo adapter was directly ligated to mRNA without calf intestinal phosphatase and tobacco acid pyrophosphatase treatment, which would have restricted analysis to full length mRNA. The GeneRacer Oligo dT primer was then used to synthesize first-strand cDNA. Two sets of reactions were performed: 1) with the GeneRacer 5’ Primer and gene-specific primers; and 2) with the GeneRacer 5’ Nested Primer and gene-specific nested primers (Additional file 7). After amplification, 5’RACE products were gel-purified and cloned into the pCR 4-TOPO vector, and approximately 10 independent clones were randomly chosen and sequenced.

Availability of supporting data

The small RNA sequencing and degradome sequence data were deposited into the NCBI with accession PRJNA307581. RNAseq reads used for expression data are available at NCBI short reads depository with accession PRJNA273732.