Abstract
Background Plant meristems are analogous to animal stem cell niches as they maintain a pool of undifferentiated cells that divide and differentiate to give rise to organs. The carpel margin meristem is a vital, multi-potent structure located in the medial domain of the Arabidopsis thaliana gynoecium, the female floral reproductive organ. The carpel margin meristem generates ovules that upon fertilization become seeds. The molecular mechanisms that specify this meristematic region and regulate its organogenic potential are poorly understood. Here, we present an analysis of the transcriptional profile of the medial domain of the Arabidopsis gynoecium highlighting the developmental stages that immediately proceed ovule initiation, the earliest stages of seed development.
Results Using a floral synchronization system and a SHATTERPROOF2 domain-specific reporter, paired with fluorescence-activated cell sorting and RNA sequencing, we assayed the transcriptome of the gynoecial medial domain with temporal and spatial precision. This analysis reveals a set of genes that are differentially expressed within the SHATTERPROOF2 expression domain that marks portions of the developing medial domain. Many members of this gene set have been shown previously to function during the development of medial domain-derived structures, including the ovules, thus validating our approach. Other uncharacterized members including differentially expressed cis-natural antisense transcripts, are potential novel regulators of medial domain development. Members of the REPRODUCTIVE MERISTEM (REM) family of transcriptional regulators were enriched in the SHATTERPROOF2-expressing cell population including a previously unrecognized REM family member. Finally, the analysis transcriptional isoforms in the medial domain identified genes that may exhibit “isoform switching” behavior during gynoecial development.
Conclusions This data set provides genome-wide transcriptional insight into the development of the gynoecial medial domain that contains the carpel margin meristem, a vital reproductive structure that gives rise to the ovules in Arabidopsis thaliana.
Background
The seedpod of flowering plants develops from the gynoecium, the female reproductive structure of the flower [1]. The gynoecium generates the ovules (the precursors of the seeds) and develops into the edible fruit in many fruiting species. As an estimated two-thirds of the calories of humankind's’ diet are derived from gynoecia and seeds, the gynoecium is a globally vital structure [2], [3].
In the flowering plant Arabidopsis thaliana, the gynoecium is a morphologically complex, multi-organ structure with a diversity of tissues and cell types [1, 4, 5]. The mature gynoecium displays morphological and functional differentiation along apical-basal, medio-lateral and adaxial-abaxial (inner-outer) axes. Stigmatic and stylar tissue form at the apex of the gynoecium, where the pollen grains are received and germinate. The stigma and style also comprise the apical-most portion of the transmitting tract, a structure that allows the pollen tube cell and sperm cells to reach the internally-located female gametophytes [4, 6, 7] (Fig. 1a,b,c). Located basal to the stigmatic and stylar tissue is the ovary portion of the gynoecium.
Ovules form within the ovary from a meristematic structure termed the medial ridge or carpel margin meristem (CMM), located in medial portions of the gynoecium [5, 8, 9] (Fig. 1a,b,c). Plant meristems are analogous to animal stem cell niches as they maintain a set of undifferentiated cells that can divide and differentiate into numerous tissues and cell types [10]. Early during floral development, patterning events divide the gynoecial primordium into medial domain that contains the carpel margin meristem and lateral domains that will form the walls of the gynoecium [5]. These domains express different sets of transcriptional regulators from early developmental time points.
Many genes that play a role in the development of the CMM and in the generation of ovules from this structure have been previously analyzed [8]. However, due to the complexity of the developing gynoecium and the heterogeneity of the gynoecial tissues, the ability to analyze the transcriptomic signature of the developing CMM or even other specific developing gynoecial structural domains has been limited. Wynn et al. previously evaluated the transcriptional properties of the gynoecial medial domain using hand-dissected gynoecial samples from the seuss aintegumenta (seu ant) double mutants that display a loss of many medial-domain-derived structures including ovules [11]. They identified 210 genes displaying reduced expression in seu ant gynoecia from floral stages 8-10 (stages according to Smyth et al. [12]). Many of these genes were shown via in situ hybridization to be preferentially expressed in the developing medial domain of the wildtype gynoecium and several of these genes have been shown to function during the development of ovules from the medial domain [8]. It is, however, difficult with this approach to obtain samples from gynoecia younger that stage 8 and thus to assay the earliest gynoecial patterning events.
An alternative approach to investigate the transcriptional properties of specific cellular populations utilizes Fluorescence-Activated Cell Sorting (FACS) of protoplasted cells to isolate specific-cell populations based on patterns of gene expression. This approach has been successfully applied to the Arabidopsis Shoot Apical Meristem (SAM) [13, 14] and roots [15, 16], [17], [18] as well as to developing cell lineages within the Arabidopsis leaf epidermis [19].
Here, we developed a novel FACS-based system for the transcriptomic analysis of a specific cellular population from the developing gynoecium, specifically the population of cells expressing the transcriptional regulator SHATTERPROOF2 (SHP2). SHP2 encodes a MADS-domain transcription factor that is expressed early within the developing CMM and thus functions as a marker for the meristematic population of cells that generate the transmitting tract and ovules [20-23]. In order to focus our analysis on early stages of gynoecium development during which key patterning events occur, we generated a SHP2-domain-specific reporter in a genetic background that allowed the synchronization of floral development. FACS-based protoplast sorting procedures, coupled with RNA sequencing, provided a unique temporal and spatial precision to assay the transcriptional signature of the gynoecial SHP2-expression domain.
Our system provides the ability to isolate a large numbers of cells from a temporally- and spatially-restricted gynoecial domain. We apply this system to investigate the transcriptomic signature of the medial domain of the gynoecium at the developmental stages when key patterning events and ovule initiation occur. Our analysis reveals many genes that are expressed preferentially within the developing medial portions of the gynoecium including members of the REPRODUCTIVE MERISTEM (REM) family of transcriptional regulators [24, 25]. We also take advantage of strand-specific RNA sequencing technology to find coding protein genes and non-coding RNAs (ncRNAs) as well as to examine isoforms and naturally occurring antisense transcripts that are preferentially expressed in the medial domain.
This work complements and extends previous analyses of medial domain development and generates a list of potential novel regulators of medial domain development that are strong candidates for future functional analyses. Furthermore, global analyses of the transcriptomic dataset indicate a similarity of the pSHP2-expressing cell population to previously characterized meristematic domains, further supporting the meristematic nature of this gynoecial tissue.
Results and Discussion
FACS-based protoplast sorting allows the collection of the SHP2-expressing cell population from a temporally restricted inflorescence sample
The transcriptional regulator SHP2 is preferentially expressed in the medial domain of the gynoecium and in a subset of the medial-domain derived tissues [21-23] (Fig. 1d,e). SHP2 plays an important role in the development of the medial domain and in the specification of ovule identity [22, 26-29]. To better characterize the molecular mechanisms of the medial domain and ovule development, we sought to identify transcripts that are differentially expressed within the medial domain of the Arabidopsis thaliana gynoecium relative to the rest of the inflorescence. To enable this, we generated a transgenic line containing a two-component reporter system, in which a pUAS-3xYPET reporter was driven by a pSHP2-GAL4 driver construct (Methods). Throughout this manuscript we refer to this two-component reporter as pSHP2-YFP.
A system for the collection of temporally- and spatially-restricted cell populations from the Arabidopsis thaliana gynoecium. a Microscopic image of a mature wild type Arabidopsis gynoecium. The stigma (stg), style (sty), carpel valve (cv), abaxial replum (abr), gynophore (gn), and ovary (ovy) are false colored. b False-colored confocal cross section of a stage-8 gynoecium. Medial and lateral domains of the Arabidopsis gynoecium are indicated. The carpel margin meristem/medial ridge (CMM) is false colored pink. c False-colored stage-11 cross-section. Ovules (ov), septum (s) and carpel valves (cv) are indicated. d Confocal microscope image of the pSHP2-YFP two-component reporter in the ap1; cal; pAP1::AP1:GR background. YFP expression from the pSHP2-YFP reporter is chiefly confined to the medial domain of the gynoecium at late stage 7/early stage 8, although weak, non-medial domain expression can be detected in portions of the stamens. Sepals (se) and stamens (st) are labeled. e Z-stack composite 3D projection image of a gynoecium isolated from the flower at mid-stage 8. YFP expression from the pSHP2-YFP reporter is detected in the medial domain and at the apex of the gynoecium. f Chloral hydrate image of an inflorescence of an ap1; cal; pAP1::AP1:GR plant after mock treatment. Inflorescence-like meristems do not transition to floral meristems. g Chloral hydrate image of an inflorescence of an ap1; cal; pAP1::AP1:GR plant 125 hours after spray application of Dexamethasone synthetic hormone (Dex). Samples were enriched for stages 6-8. h Percentage of flowers at a given stage from inflorescences used for FACS-sorting. Stages 6, 7, 8p (pre-ovules) and 8s (post-ovules) are indicated in the X-axis as St6, St7, St8p, St8s, respectively. Stage 8p is before any visible morphological manifestation of ovule primordia upon observation under DIC microscopy. Stage 8s ovule primordia were observed and were at ovule stage 1-I or 1-II according the Schneitz et al. [125]. i Confocal microscopy of YFP fluorescence of protoplasted cells after FACS. Panels a, b and c are adapted from Azhakanandam et al. [43] (with permission).
To better understand the early specification of medial and lateral gynoecial domains and in the earliest stages of ovule primordium initiation, we focused our transcriptomic analysis on floral stages 6-8, when these key developmental events occur [5]. In order to increase our ability to collect a large number of pSHP2-YFP-expressing cells from this specific bracket of developmental stages, we crossed the pSHP2-YFP reporter into an ap1 cal-based floral synchronization system that allows the collection of large numbers of semi-synchronized flowers at roughly the same developmental stage [30, 31]. The expression of the pSHP2-YFP reporter in the floral synchronization system was largely similar to that observed in wild-type inflorescences [20, 22, 23], and was confined chiefly to the medial domain and medial domain-derived tissues (Fig. 1d,e). Some expression was observed in non-medial domain tissues. The most apparent of this was expression in the apex of the developing gynoecium where both medial and lateral domains express the pSHP2-YFP reporter. Additionally, expression could be observed in a small number of cells within the stamens (Fig. 1d) and occasionally in the edges of sepals that appeared to have undergone a homeotic transformation toward a carpelloid fate (data not shown). Thus, the vast majority of the pSHP-YFP reporter expression reflected the endogenous pSHP2 expression domain (in the medial and apical portions of the gynoecium). A minority of the expression outside of the gynoecium may reflect ectopic expression of the reporter due to genetic background or transgene insertion site or limitations of the regulatory sequences used in the pSHP-YFP reporter construct.
Microscopic examination of our semi-synchronized inflorescence samples indicated that flowers ranged between floral stages 1 and early stage 8, with a strong enrichment for floral stages 6 through early 8 (Fig. 1g,h). Flowers that had developed beyond late stage 8 were not detected in our samples. Thus, our biological sample is strongly enriched for transcripts that are expressed during early patterning of the gynoecium and the earliest stages of ovule development (initiation) and does not include later floral developmental stages where SHP2 is expressed in stigma, style and valve margin tissues. Additionally, as the initial expression of the pSHP-YFP reporter is detected at late stage 5 or early stage 6 [23], we expect that the population of YFP-expressing protoplasts derived from this material will be highly enriched with cells from the stage 6-8 medial domain. FACS-sorting of protoplasts derived from these inflorescences yielded three populations of sorted cells (collected in biological quadruplicate): “YFP-positive”, “YFP-negative” and “all-sorted” (Additional file 1: Figure S1). The “all-sorted” sample included all protoplasts recovered (regardless of YFP expression) after sorting gates were applied to remove debris and broken cells (Methods). We additionally collected (also in biological quadruplicate) “non-sorted” samples from entire non-protoplasted inflorescences to measure the abundance of transcripts in the biological starting material before protoplast generation and FACS-sorting. In order to evaluate the purity of the YFP-positive protoplasts during a preliminary FACS run, YFP-positive cells were resorted. Ninety six percent of the YFP-positive cells were found to resort into the YFP-positive gate, indicating a high degree of enrichment and purity in the YFP-positive sample (Additional file 1: Figure S1). Confocal microscopy (Fig. 1i) also revealed an enriched population of intact YFP-positive protoplasts after FACS.
Sorting/re-sorting YFP-positive cells yielded high cell-type specific purity. Panel a Sorting gates used to select for fluorescent YFP-positive, YFP-negative and all-sorted protoplasts. b Re-sorting YFP-positive protoplasts to assess sample purity.
We used real time PCR (qRT-PCR) to estimate the degree of enrichment of the endogenous SHP2 and NGATHA1 (NGA1) transcripts in RNA samples derived from the YFP-positive and YFP-negative samples. NGA1 is expressed in the adaxial portions of the gynoecium starting at stage 7 in a domain that partially overlaps with the SHP2 expression domain [32, 33] and thus provides an additional benchmark to estimate the enrichment of medial domain-expressed transcripts. The normalized level of the SHP2 transcript was ~30 fold higher in the YFP-positive samples relative to the YFP-negative samples (p <0.001) while the NGA1 transcript was ~4 fold higher in the YFP-positive sample (p <0.05). The difference in the levels of the TUBLIN6 was not found to be statistically significant (p = 0.4) between the YFP-positive and YFP-negative samples (Additional file 1: Figure S2).
Enrichment of medial-domain expressed genes, SHATTERPROOF2 (SHP2) and NGATHA1 (NGA1), assessed by quantitative real time PCR (qRT-PCR) from FACS-sorted protoplasts corresponding to YFP-positive and YFP-negative samples. The differences in the level of expression between the YFP positive and YFP negative samples transcripts were statistically significant for SHP2 and NGA1, p <0.001 and p <0.05, respectively. The difference in the levels of expression for the TUBLIN6 gene was not found to be statistically significant (p = 0.4). The expression levels of the ADENINE PHOSPHORIBOSYL TRANSFERASE1 (APT1) transcript was used for normalization.
Transcriptomic analysis of the gynoecial SHP2 expression domain and identification of candidate regulators of gynoecial medial domain development
To investigate the transcriptomic profile of the gynoecial SHP2 expression domain, we performed high-throughput RNA-sequencing from the collected protoplasts and non-protoplasted inflorescences samples. We expect that the identification of differentially expressed genes (DEGs) between the YFP-positive and YFP-negative samples (referred to as “YFP-positive/YFP-negative” or “YFP+/-”) will provide insight into the set of transcripts differentially expressed in the gynoecial medial domain relative to the rest of the inflorescence. Additionally, DEGs identified in the all-sorted and non-sorted comparison (referred to as “all-sorted/non-sorted”) are expected to reveal transcripts that are differentially represented as a result of the protoplasting/FACS-sorting protocol.
Two lanes of the HiSeq2500 Illumina sequencing platform yielded 320 million raw reads with an average of 20 million reads (MR) per library. Nearly 11 MR were filtered out after removing barcode-adapters and low quality sequences. The remaining 306 MR were aligned against the Arabidopsis thaliana TAIR10 reference genome [34] with more than 90% of them successfully mapping to the genome sequence. Among the mapped reads, 244 MR mapped uniquely to only one location and were used for subsequent analyses. A detailed breakdown is shown in Additional file 2: Table S1.
Sample names are listed in the first column (YFP-positive, YFP-negative, all-sorted, non-sorted). The 2nd and 3rd column correspond to million DNA reads (MR) in each technical (T1 and T2) and biological replicate (B1, B2, B3, B4), respectively. Filtered reads (Q30L50) per sample are found in the 4th column and 5th column (technical and biological replicate, respectively). The 6th column indicates the number of the MR filtered out after removing adapters and low quality sequences. The 7th and 8th columns indicate MR mapped to the Arabidopsis TAIR10 genome. The 9th and 10th column include uniquely mapped reads to the Arabidopsis TAIR10 genome and 11th the MR that aligned more than in one place in TAIR10 genome. The 12th and 13th column represent reads mapped to only annotated TAIR10 genes (technical and biological replicates). The 14th and 15th columns correspond to uniquely mapped reads to annotated genes.
We used three different programs to determine expressed and differentially expressed protein coding genes in our dataset: Cufflinks [35], edgeR [36] and DESeq2 [37] (Methods) (Non-protein coding gene models were considered separately and are presented below). Here, the term “differentially expressed gene (DEG)” is used to indicate a gene whose steady-state transcript level differs significantly at a false discovery rate (FDR) of <0.001 and shows a fold change of four or more between the two compared RNA samples. To identify potential regulators of gynoecial medial domain development, a ‘stringent’ criteria was used to select a subset of the YFP+/-DEGs for downstream analysis. For a gene to be selected from the YFP+/-comparison, we required that the transcript is identified as differentially expressed by all three independent software packages (Fig. 2b). Alternatively, to identify DEGs in response to the protoplasting/FACS-sorting procedure, a ‘less stringent’ criterion was used. Transcripts in the union set of all the non-sorted/all-sorted DEGs were considered to be potential protoplast-induced genes even if they were identified by only one software program (Fig. 2a). Only 48 transcripts were found in common between the YFP+/-DEGs and the all-sorted/non-sorted DEGs (Fig. 2c), indicating a high degree of specificity in the DEGs identified in each comparison. We then removed these 48 transcripts from our analysis to eliminate any that might be differentially expressed as a result of the protoplast generation or FACS-sorting procedures, leaving 363 “cleaned” protein coding DEGs (Fig. 2c). The expression profiles of these 363 YFP+/-DEGs, including data from the allsorted and non-sorted samples, are represented in a heatmap (Additional file 1: Figure S3). This gene set includes 95 DEGs whose transcript levels were higher in the YFP-positive samples (“enriched”) and 268 DEGs whose transcript levels were lower (“depleted”) in the YFP-positive samples, relative to the YFP-negative samples (Additional file 3: Table S2).
Transcriptional profile in Reads Per Kilobase of transcript per Million mapped reads (RPKM) of the 363 selected differentially expressed genes across all the samples used in this study (YFP POS = YFP-positive, YFP NEG = YFP-negative, ALL SORT = all-sorted, and NO SORT = non-sorted). Biological replicates are indicated as B1, B2, B3, and B4.
Table 1 - List of 363 differentially expressed genes (YFP+/-) with FDR <0.001 and fold change > 4 co-identified by three independent software packages (Cufflinks, DESeq2, and edgeR). Table 2 - List of 410 differentially expressed genes (all-sorted/non-sorted) with FDR <0.01 and fold change > 4 co-identified by Cufflinks, DESeq2 and edgeR. Table 3 - List of 48 genes differentially expressed in both the YFP+/- comparison and the non-sorted/all-sorted comparison.
Venn diagram of differentially expressed genes (DEGs) using Cufflinks, edgeR and DESeq2 (FDR<0.001, Fold Change >4). a Venn showing DEGs identified between the all-sorted/non-sorted samples with the 3 programs used for differential expression analysis of RNA-seq expression profiles. b Venn showing DEGs between YFP+/-samples identified in the 3 programs. c Intersection of the DEGs (48) from both datasets (a and b). DEGs (363), after removing DEGs induced by the protoplsting/FACS-sorting stress, were used for downstream analysis.
For the 95 DEGs that were enriched in the YFP-positive sample (at a fold change > 4), we expected many to be preferentially expressed in the medial portions of the gynoecium at floral stages 6-8. To test this, we examined the literature to determine the expression patterns of members of this gene set. From the top 15 of the 95 YFP-positive enriched DEGs (ranked by fold change), five have previously been reported to be preferentially expressed in the gynoecial medial domain via in situ or reporter gene analysis [i.e. HECATE1 (HEC1), HEC2, SHP1, SHP2 and STYLISH1 (STY1)] [20-22, 38, 39] and three others are previously described as enriched in medial domain-derived tissues in published transcriptomic datasets (i.e., AT1G66950, AT5G14180, and AT1G03720) [40, 41] (Table 1). An additional gene from this list, CRABS CLAW (CRC), has been shown via in situ hybridization to be expressed in portions of the medial gynoecial domain, as well as non-medial portions of the gynoecium [42, 43]. The expression pattern of the remaining six genes from this gene list have not yet been assayed in the gynoecium. Thus, as predicted, the set of 95 genes enriched in the YFP-positive sample is enriched for genes that are preferentially expressed in the gynoecial medial domain.
Top 15 most differentially expressed genes from the YFP+/-comparison (FDR <0.01), enriched in the YFP-positive sample ranked by fold change (FC). Arabidopsis gene ID is shown in the 1st column, gene name (TAIR10 annotation) is shown in the 2nd column and the 3rd column reference available reporter lines and/or in situ hybridization for the top 15 DEGs in this study. Reads Per Kilobase of transcript per Million mapped reads (RPKM) values are indicated for each sample (YFP_NEG = YFP-negative, YFP_POS= YFP-positive) and each biological replicates (B1, B2, B3, B4).
Published functional analyses of HEC1, HEC2, SHP1, SHP2 and STY1 indicate that these genes function during the development of the medial domain or medial domain-derived tissues [22, 28, 29, 38, 39]. Many other genes in the set of 95 DEGs enriched in the YFP-positive sample have been previously shown to play a role in medial domain development (e.g. NGA family members [32, 33], SPT [44], and CUC2 [45]). Other genes within this list are interesting candidates for future functional studies. This includes members of the REM family of transcriptional regulators [24, 25], several auxin synthesis or signaling-related genes such as LIKE AUXIN RESISTANT 1 (LAX1) (AT5G01240) [46] and YUCCA4 (YUC4) (AT5G11320) [47], as well as transcription factors regulating other developmental processes such as MATERNAL EFFECT EMBRYO ARREST 3 (MEE3) (AT2G21650) [48] and GLABROUS 3 (AT5G41315) [49].
It is important to note that the 48 DEGs that were identified in both the YFP+/- and all-sorted/non-sorted comparisons (Fig. 2c) should not be discounted as potential medial domain regulators. These genes may be both preferentially expressed in the YFP-positive cell population as well as induced in response to the protoplasting procedure (Additional file 3: Table S2). Indeed, some of these genes, including the transcription factors HECATE3 and BR-ENHANCED EXPRESSION1 (BEE1), have been reported to be preferentially expressed in medial domain-derived tissues and to function in gynoecium development [38], [50]. However, we chose to use the “cleaned” set of 363 YFP+/-DEGs for downstream analyses in order to reduce the likelihood of the inclusion of genes whose expression was altered significantly by the protoplasting process.
REPRODUCTIVE MERISTEM family members are differentially expressed in the SHP2-expression domain
In order to look for enriched categories of transcription factors within the set of “cleaned” 363 YFP+/-DEGs (Fig. 2c), we used the online Transcription Factor Enrichment Calculator tool [51]. Members of the ABI3/VP1 transcription factor family that includes the REM family TFs and NGA family TFs were found to be statistically over-represented (Additional file 8: Table S7) (corrected p <9.97E-06). The REMs belong to the plant-specific B3 superfamily of transcription factors and expression of many REM family members is observed in meristematic tissues such as the inflorescence meristem, floral meristem and the CMM [11, 24, 25, 52-54]. The numerical designations used to describe the REM family members in this manuscript are taken from Romanel et al. [24]. In our study, six REM members were amongst the 363 statistically significant YFP+/- DEGs; five were found to have enriched expression in the YFP-positive sample, while one, REM25 (AT5G09780), was ~4 fold less abundant in the YFP-positive sample. REM13 (At3g46770) transcript level is enriched ~12 fold in the pSHP2-YFP expressing cells. REM13 was previously predicted to be preferentially expressed in the inner integument, ovule primordia and medial domain based on transcriptomic data [40]. We employed in situ hybridization to assay the expression pattern of the REM13 transcript during gynoecial development (Figure 3). Using a REM13 antisense probe, we detected signal in the medial portions of the gynoecium corresponding to the carpel margin meristem as early as stage 7. Expression was also observed in the initiating ovule primordia in stage 8 gynoecia and then continued to be detected in portions of the ovules at later developmental stages.
Candidate medial domain regulator REM13 (At3g46770) is expressed within the medial gynoecial domain and developing ovules. Results from an RNA in situ hybridization with REM13 probe. a-d antisense probe. e sense strand probe. a Hybridization signal is detected in the carpel margin meristem (adaxial portions of the medial gynoecial domain) in the stage 7 longitudinal section. b, c and d In transverse gynoecial sections REM13 expression is detected in the ovule primordia; stage 7 (panel b) stage 8 (panel c) and stage 9 (panel d) gynoecia. e A stage 8 section hybridized with a REM13 sense strand probe. (ov) - ovules, (cmm) - carpel margin meristem. Scale bars for each panel represent 5 microns.
REM34/ATREM1 (At4g31610) [24, 52], REM 36 (At4g31620)[53], and VERDANDI (VDD/REM20) [53, 55] also displayed enriched expression levels in the YFP-positive sample of ~8 fold, ~9 fold and ~6 fold, respectively. Published in situ hybridization patterns indicate enriched medial domain expression patterns for REM34/ATREM1 and VDD/REM20 [11, 52, 55]. Additionally, expression of At5g60142, a previously unnamed member of the REM family, is enriched ~11 fold in the YFP-positive sample (Additional file 4: Table S3). At5g60142 is an interesting candidate for functional studies that is located on chromosome V in tandem to REM11 (At5g60140) and REM12 (At5g60130) and shares a high degree of sequence similarity with these two genes, as well as REM13 [24, 53]. We propose to designate At5g60142 as REM46.
Gene Set Enrichment Analysis
To gain global insights into underlying biological mechanisms of medial domain development and function, Gene Set Enrichment Analysis (GSEA) was performed for the 95 YFP-positive enriched DEGs in the medial domain. This analysis identified 147 GO terms that were statistically overrepresented (p < 0.01), including “gynoecium development” (GO:0048467) and “flower development” (GO:0009908), “response to gibberellin” (GO:0009739) and “auxin homeostasis” (GO:0010252) (Fig. 4 and Additional file 7: Table S6). This GSEA analysis further suggests that the set of 95 genes enriched in the YFP-positive sample function as regulators of medial domain development.
GO term overrepresentation of SHP2-domain enriched genes suggests a role for this set of genes in floral, gynoecial and ovule development. BiNGO/Cytoscape representation of overrepresented GO terms from the 95 YFP+/- DEGs displaying enriched expression in the YFP-positive samples. Edges represent the parent/child relationships of the GO terms [126], while color of the nodes indicates the degree of statistical significance (p <0.01) as reported by BiNGO [123]. To unclutter the figure, given the large number of significant GO terms, selected nodes and edges have been removed from this graphical representation.
In contrast, when performing GSEA with DEGs identified between the all-sorted/nonsorted samples, a different set of 304 overrepresented GO terms were identified, including “response to stress” (GO:0006950) and “response to wounding” (GO:0009611), suggesting that many of the genes identified as differentially expressed between the all-sorted/non-sorted samples reflect stress-induced changes in gene expression during protoplast/FACS-sorting.
The transcriptomic signature of the SHP2-expressing cell population shares commonalities with transcriptional signatures of other meristematic samples
In order to gain insight into the characteristics of the 363 YFP+/- DEGs identified from the SHP2 expression domain, we compared the expression profile of this set of genes across several different tissues. Using existing Arabidopsis RNA-seq transcriptomic datasets from whole flowers [56], aerial seedlings tissues (GEO accession: GSE54125), as well as from Laser Capture Microdissected (LCM) inflorescence meristems, floral meristems and stage-3 flowers [54], a Spearman rank correlation analysis was performed. In the sample-wise hierarchical clustering (Fig. 5a), the transcriptomic profiles from the SHP2-expressing (YFP-positive) sample clustered more closely with the meristematic samples, while the YFP-negative and all-sorted samples clustered more closely with the whole-flower and whole-seedling samples. This suggests that the expression signature of the YFP-positive sample is more similar to that of the floral and inflorescence meristems and young flowers, than it is to whole flowers or young vegetative seedlings (Fig. 5a).
The transcriptomic signature of the SHP2-expressing domain is more similar to the transcriptomes of other meristematic samples than it is to whole flower. a Dendrogram based on hierarchical clustering using the Spearman rank correlation using RNA-seq (RPKM) expression values from flowers and other tissues. b Comparison of RNA-seq and affymetrix ATH1 arrays samples including transcriptomic data from whole flower, shoot apical meristem and seedling. WT = wild type, MT= mutants. Data from Mizzotti et al. [56](1), Mantegazza et al. [53](2), GEO accession: GSE54125(3) and Yadav et al. [13, 14](4) were used for comparison. Samples corresponding to this study are color coded red in both dendrograms.
Further supporting the similarity of the SHP2-expressing domain to other meristematic samples, the expression levels of GA20OX1 (AT4G25420) and GA20OX2 (AT5G51810) were both significantly depleted in the YFP-positive sample, relative to the YFP-negative sample (Additional file 11: Table S10). GA20OX1 and GA20OX2 encode key biosynthetic enzymes of the plant hormone gibberillic acid (GA) [57]. Levels of expression of GA20OX1 and GA20OX2 are low in the shoot apical meristem (SAM) relative to expression in the juxtaposed young organ primordia and high levels of GA synthesis interfere with the maintenance of meristematic fate in the SAM [58, 59]. These data suggest that low levels of GA may also be associated with the meristematic nature of the carpel margin meristem. Although not discussed here, expression values of genes annotated with a role in ethylene signaling are found in Additional file 11: Table S10.
We additionally compared the medial domain transcriptional signature to datasets generated with the Affymetrix ATH1 array allowing comparisons to transcriptomic signatures of a variety of cell types including vascular and meristematic cell types from the Arabidopsis SAM isolated via FACS [13, 14]. When these additional samples are included, the hierarchical clustering dendrogram (Fig. 5b) shows the YFP-positive sample is more similar to the SAM cell-types, rather than to the vascular procambium (AtHB8) and phloem cell types (S17). This again suggests the meristematic character of the YFP-positive sample (Fig. 5b). One should be cautious, however, to interpret the results of this (or any) cross-platform (array/RNA-seq) comparison until validated crossplatform comparisons methods are available. To the best of our knowledge, there is no clear consensus in the literature of a standard cross-platform comparison practice [60], [61], [62], [63]. Indeed, many researchers have used both platforms (array/RNA-seq) in the same experiment comparing final results rather than finding a way to directly compare the two technologies [64], [65], [66], [63], [67]. Here, we employ a Spearman rank correlation as it is less sensitive than the Pearson correlation to strong outliers, makes no assumptions about data distribution, and does not inflate type I error rates. This approach fits well with the data in this work as samples do not cluster based on technology platforms but rather cluster based on the apparent cell-type similarities of gene RPKM (Reads Per Kilobase of transcript per Million mapped reads) expression levels.
Transcriptomic analysis of the SHP2 expression domain complements existing medial domain and CMM data sets
Wynn et al. previously carried out a related transcriptomic study and identified many genes that were shown via in situ hybridization to be preferentially expressed in the developing medial domain of the wild-type gynoecium [11]. When comparing the 95 enriched DEGs from our RNA-seq experiment (Additional file 3: Table S2 and Additional file 1: Figure S3) with a set of 210 medial domain enriched genes from Wynn et al., 23 genes were found in common (Table 3). The 24% overlap of these two gene sets is significantly higher than expected by chance (hypergeometric test; p = 3.15 × 10−30) [68]. Members of the REM, HECATE and NGA gene families, as well as several auxin-homeostasis-related genes were among the set of 23 genes identified in both experiments (Table 3).
Reyes-Olalde et al. recently performed a comprehensive literature survey of genes that function during CMM development [8]. They reported 86 protein-coding genes corresponding to transcription factors, hormonal pathways, transcriptional co-regulators, and others of widely diverse functions. While all 86 are expressed in our dataset, fifteen of these CMM developmental regulators are found within the set of 363 YFP-positive DEGs (hypergeometric test; p = 3.3 × 10−13) [68] (Fig. 6). The expression profiles of the 86 genes reported by Reyes-Olalde et al. within the medial domain-enriched dataset from this work, as well as within data from floral meristem enriched samples [54], is displayed in a heatmap in Figure 6 (RPKM values can be found in Additional file 10: Table S9).
Heatmap representation of the expression profiles of previously identified regulators of Carpel Margin Meristem development. Expression profiles in Reads Per Kilobase of transcript per Million mapped reads (RPKM) of the 86 genes reported by Reyes-Olalde et al. [8] with functional role during CMM development. Transcriptional profiles from this study (YFP POS = YFP-positive, YFP NEG = YFP-negative, ALL SORT = all-sorted, and NO SORT = no-sorted) as well as Mantegazza et al. [54] corresponding to flower stage 3 (FL.STAGE 3), floral meristem (FL.MERISTEM) and inflorescence meristem (IN.MERISTEM) are included. Genes color-coded in red are those identified as DEGs between YFP-positive and YFP-negative samples (FC >4 and FDR <0.001) while genes that displayed a statistically significant expression level (FDR <0.01) between YFP-positive and YFP-negative (regardless of their fold change) are indicated with ***.
Transcript isoforms in the Arabidopsis medial domain
One utility of transcriptome analysis through RNA-seq is the identification of novel alternative spliced transcripts, alternative transcription start sites (TSS), and instances of isoform switching [69]. To further characterize the transcriptome of the SHP2-expression domain at the isoform level, we first selected isoforms that showed a significant (α <0.01) change in their expression between YFP+/-samples using Cufflinks/Cuffdiff. For this analysis we did not apply a fold magnitude cutoff, thus capturing all isoforms with α <0.01. To avoid transcripts that were affected by the cell-sorting procedure, we removed all isoforms that showed a significant (α <0.01) expression level change between all-sorted/non-sorted samples. This resulted in 4555 YFP+/- differentially expressed isoforms (Additional file 9: Table S8). Within this set of isoforms differentially expressed between the YFP+/- samples, we sought to highlight multi-isoform genes that showed major changes in the relative frequency of individual isoforms between the YFP-positive and YFP-negative samples. To this end, we estimated the relative frequency of each isoform as a percentage of the total expression for the gene. Among the 4555 significantly differentially expressed isoforms, only 52 isoforms from multi-isoform genes displayed changes of 20% or more in their relative frequency. The major isoform (most highly expressed isoform) differed between YFP+/- samples for only 15 genes (Table 2). Remarkably, the transcriptional co-regulator SEU (At1g43850), previously implicated in medial domain development [70], [43], showed a significant increase of isoform At1g43850.1 in the YFP-positive samples, while its second isoform At1g43850.2 did not significantly change between samples. As a result, isoform 1 was the major (predominant) isoform in YFP-positive cells, and isoform 2 was the major (predominant) isoform in the other samples. The functional significance, if any, of this isoform switching is currently unknown.
List of significant isoforms (a=0.01) between the YFP+/- samples and nonsignificant (a=0.01) between the all-sorted/non-sorted comparison. These genes were 20% or more enriched in the medial domain for a given isoform. TSS = transcriptional start site. Match between the Cufflinks transcripts and TAIR10 genome are indicated with class code ‘=’ for complete transcript match and ‘j’ for potentially novel isoform (fragment) [35].
List of overlapping differentially expressed genes (DEGs) between the 95 DEGs from this study (FC > 4, FDR <0.001) and DEGs from the transcriptomic array data of Wynn et al. [11] derived from the seuss aintegumenta (seu ant) double mutant.
The regulation of gene expression through alternative promoter usage or use of alternative TSS is frequently observed in multicellular organisms [71]. Using the same pipeline and criteria we employed to select differentially expressed isoforms in the YFP+/- samples, we identified 93 isoforms that were differentially expressed as a result of the use of alternative promoter/transcriptional start sites (Additional file 9: Table S8). Interestingly, one such promoter/transcriptional start site switch was found for the
REVERSIBLY GLYCOSYLATED POLYPEPTIDE 5 (RGP5) gene (isoform). Members of the RGP family (RGP1 and RGP2) involved in sugar metabolism are expressed in other Arabidopsis meristematic tissues, such as the root tip and the apical meristem of young seedlings [72]. In our work, the transcript level of RGP5 isoform 2 (At5g16510.2) in the YFP positive sample is 61% higher relative to the level of this isoform in the YFP-negative sample, while the level of isoform 1 (At5g16510.1) is 75% lower (Fig. 7a and Additional file 9: Table S8).
Differential expression of REVERSIBLY GLYCOSYLATED POLYPEPTIDE 5 (RGP5) isoforms as well as TRANS-ACTING siRNA3 (TAS3) and AUXIN RESPONSE FACTOR genes. a Promoter/transcriptional start site switch found for the RGP5 gene (At5g16510). The isoform 2 (At5g16510.2) increases its expression in the YFP-positive domain while isoform 1 (At5g16510.1) of the same gene decreases its expression in the same domain. b Expression of the AUXIN RESPONSE FACTORS (ARFs) (ARF2, ARF3, ARF4) and TAS3 transcripts. Expression levels of ARF2, ARF3, ARF4 are significantly enriched in the YFP-positive sample at FDR <0.01. Expression levels of the TRANS ACTING siRNA3 (TAS3) genes At5g49615 and At3g17185, that negatively regulate the expression of ARF2, ARF3, and ARF4 expression [85], are significantly reduced (FDR <0.01) in the YFP-positive sample.
Auxin homeostasis and the development of the gynoecial medial domain
Auxins are a class of plant hormones that regulate growth and development [73, 74]. The most common plant auxin is Indole-3-Acetic Acid (IAA). The regulation of auxin homeostasis (including synthesis, response, transport, inactivation and degradation) plays an essential role in patterning the gynoecium and other lateral organs [74, 75]. The role of auxin during the development of the medial and lateral domains of the gynoecium is less clearly defined, however recent studies suggest that auxin homeostasis mechanisms are likely to be distinct in medial and lateral domains [23, 75, 76].
To better analyze auxin homeostatic mechanisms during medial domain development, we examined the expression of 127 genes with an annotated function in auxin homeostasis. Of these 127 genes, 80 were expressed in our dataset and 60 were differentially expressed at a FDR of < 0.01 in the YFP +/- comparison, without applying a fold enrichment filter (Additional file 11: Table S10). The expression levels of TRYPTOPHAN AMINOTRANSFERASE OF ARABIDOPSIS 1 (TAA1) and YUC4, two genes encoding proteins in the auxin synthetic pathway, were strongly enriched (> 4 fold) in the YFP positive samples as was predicted from previously published expression patterns indicating enriched expression within the medial portions of the gynoecium [33, 47, 77–80]. Within the PINFORMED (PIN) family of polar auxin transporters, the expression levels of PIN1, PIN3 and PIN7 were significantly enriched in the YFP-positive sample (Additional file 11: Table S10). This is consistent with the reported expression patterns at the protein level of these PIN transporters within the medial domain of the gynoecium [23, 76, 81, 82].
Auxin induces gene expression through a family of transcription factors called AUXIN RESPONSE FACTORS (ARFs) [74]. At a fold change level of 1.5 fold and FDR of <0.01, ten ARFs were enriched in the YFP-positive sample (ARF1, ARF2, ARF3/ETTIN, ARF4, ARF5, ARF6, ARF7, ARF8, ARF16 and ARF18), while no ARFs were identified as depleted in the YFP-positive sample (Additional file 11: Table S10). Our data suggests these ARF family members may be preferentially expressed in the medial domain and play a role during development of this meristematic tissue. Previous studies have documented gynoecial developmental defects in arf3/ettin mutants [4] as well as arf6 arf8 [83, 84] double mutants. Interestingly, the levels of the precursor transcripts for two TRANS-ACTING SIRNA3 (TAS3) genes (At5g49615 and At3g17185) were significantly reduced (FDR <0.01) in the YFP-positive sample (Fig. 7b Additional file 9: Table S8). The trans-acting siRNAs that are encoded by the TAS3 genes negatively regulate the levels of ARF2, ARF3, and ARF4 transcripts [85]. Thus the enrichment of ARF2 ARF3 and ARF4 transcript levels in the SHP2-expression domain may in part be due to a reduction in the level of expression of the TAS3-encoded tasi-RNAs in the medial domain.
SQUAMOSA PROMOTER BINDING PROTEIN-LIKE3 (SPL3) and the cis-NAT antisense gene At2g33815
The SQUAMOSA PROMOTER BINDING PROTEIN-LIKE (SPL) genes function in the regulation of the transition from juvenile to adult growth phases, and regulation of shoot regenerative capacity [86–89]. In our study, the expression of SPL3 (At2g33810) was more than four fold lower in the YFP-positive sample relative to the YFP-negative sample. SPL3 encodes a DNA-binding protein directly regulating APETALA1 (At1g69120), a key regulator of floral-meristem-identity specification [90]. Interestingly, the expression of the cis-NAT antisense gene At2g33815, complementary to portions of the SPL3 gene, was also significantly reduced ~4.5 fold in the YFP-positive sample (Cufflinks data in Additional file 4: Table S3 and Additional file 9 Table S8). This is perhaps in contrast to the expected pattern of expression, where the expression levels of the targeted SPL3 transcript might be expected to go up as the levels of cis-NAT antisense At2g33815 go down. The expression of another regulator of SPL3 activity, the miRNA157D (At1g48742), was also significantly reduced in the YFP-positive samples. The miRNA157D reduces translation of the SPL3 transcript by acting through a miRNA156/157-responsive element in the SPL3 3’UTR [89, 91]. These data suggests that the miRNA156/157/SPL module may act during medial domain development and may be regulated by the cis-NAT antisense gene At2g33815. A complete list of differentially expressed natural-antisense, transposable-element and other non-protein coding transcripts identified as differentially expressed by Cufflinks, DESeq2 and edgeR is found in Additional file 9: Table S8.
Protoplasting-induced stress genes
While the predominant focus of this work was to perform transcriptomic analysis in medial-domain-enriched cells (YFP+/-), transcripts induced by the protoplasting and sorting process (all-sorted/non-sorted) were also identified (Methods). To facilitate the visualization of all samples, we generated an interactive 6-way Venn diagram using the web-based tool ‘InteractiVenn’ [92]. By uploading the Additional file 12 to InteractiVenn [93], mousing over, and clicking on the numbers in the Venn diagram, researchers will find gene ID from DEGs between YFP+/- and all-sorted/non-sorted samples (3 programs and two comparisons). As expected, when comparing such different types of samples (all-sorted/non-sorted and YFP+/-), few DEGs (26) overlapped across the 6 samples (Fig. 8). The lack of overlap of DEGs across the entire experiment indicates that the YFP+/- DEGs reported here are not a result of protoplasting-stress-induced processes.
Six-way venn diagram image showing detailed overlap from all the differentially expressed gene (DEGs) datasets. The total number of DEGs under each condition and for each program are indicated in parentheses. CTR= DEGs between all-sorted/non-sorted and YFPs= DEGs between YFP+/-. Cuff= Cufflinks, edg=edgeR, Des=DESeq2. The interactive tool can be accessed online using the ‘InteractiVenn’ webtool [93] and uploading Additional file 12.
When comparing the protoplasting-induced gene set from this work (all-sorted/nonsorted DEGs) with those induced due to FACS-sorting methodology in shoot apical meristem by Yadav et al. [13] and in roots as reported by Birnbaum et al. [16] few DEGs were found in common (seven across all datasets) (Additional file 1: Figure S4) indicating that different tissues and/or different protoplasting techniques generate different sets of protoplast-induced gene-expression changes. Thus, appropriate controls should be included to control for condition-specific protoplasting-induced geneexpression changes.
Conclusions
Despite the importance of the gynoecial medial domain in ovule development, no domain-specific transcriptome has been previously reported, mainly, due to the difficulty of isolating the meristematic cells from which ovules are derived. In this work, we developed a novel FACS-based system using the SHP2-expression-domain-specific using a GAL4/pUAS-based two-component system that, when combined with flower synchronization and flow cytometry, allowed for the efficient isolation of medial-domain cells expressing SHP2. The quality and quantity of biological samples that can be recovered with our system enables cell-type and strand-specific RNA-seq transcriptomic analysis and opens up possibilities for small RNAs, metabolomic and proteomic analyses [94–98]. This approach, coupled with high-throughput RNA-sequencing, has yielded a unique and novel snapshot of the gynoecial medial domain transcriptome and a set of candidate regulators of medial domain development for future functional analysis.
Methods
Construction of pSHP2-GAL4//pUAS-3xYpet dual construct lines
The SHP2 promoter fragment was amplified from Columbia wild type genomic DNA using the primers proSHP2gwF1 (5’CACCATCTCCAACGCATTGTTACG3’) and proSHP2gwR1 (5’CATTTCTATAAGCCCTAGCTGAAG3’). This fragment contains the sequences from -2170 to +1 relative to the SHP2 ATG and includes the 5’UTR, the first intron and the first Met codon of SHP2. This promoter previously was shown to mimic the endogenous SHP2 expression pattern [22]. This genomic fragment was cloned into the pENTR/D-TOPO vector (Invitrogen) to create plasmid LJ001, and then shuttled via gateway LR reaction (Invitrogen) into the destination vector JMA859 (i.e. pEarleygate303-GAL4) to create plasmid AAS003. Transgenic Arabidopsis lines were created by Agrobacterium-mediated transformation of the AAS003 plasmid into the S. No. 1880 seed stock that contained the pGWB2-pUAS-3xYpet responder construct (see below) generating the pSHP2-GAL4; pUAS-3xYpet dual construct line (S. No. 1896), referred to as pSHP2-YFP. The pSHP2-YFP plants were crossed to the ap1 cal1 Wellmer floral induction system [31] as described below.
JMA859 (pEarleygate303-GAL4) is a modified pEarleygate303 [99] plasmid in which the reporter was replaced by the coding sequences from the GAL4 yeast transcriptional activator. To achieve this, pEarleygate303 was cut with NcoI (New England Biolabs) and SpeI (New England Biolabs). Then fusion PCR was used to create the insert that fused the GAL4 sequences to the deleted portions of pEarlygate303. This required three PCR reactions: 1st PCR with primers pEarl303NcoIFor
(5’TGGCCAATATGGACAACTTCT3’) and pEarl303Rev_GAL4 (tale) (5’ATGGAGGACAGGAGCTTCATACACAGATCTTCTTCAGAGA 3’); 2nd PCR with primers GAL4F_pEarl303 (tale)
(5 ’TCTCTGAAGAAGATCTGTGTATGAAGCTCCTGTCCTCCAT3 ’) and GAL4Rev_SpeI (5’ CCGGACTAGTCTACCCACCGTACTCGTCAA3’), and then a fusion PCR joining these two fragments using the external primers to amplify. The product of the fusion PCR was double-digested with NcoI/SpeI and ligated into NcoI/SpeI-cut pEarleygate303.
JMA382 (pUAS-pGWB2) was created from pGWB2 [100] by replacing the p35S sequences in pGWB2 with pUAS sequences (HindII/XbaI sites used). A Gateway LR reaction was then used to move the 3xYpet cassette from JMA710 (pENTR/D-TOPO-3xYpet) into JMA382, creating vector JMA721 (i.e. pGWB2-pUAS-3xYpet). Homozygous single insertion-site transgenic lines harboring JMA721 were then generated (S.No 1880).
Plant material
In a wild-type inflorescence, cells expressing SHP2 represent a small percentage of the total cells. Additionally, wild-type inflorescence contains a full range of developmental series of floral stages. The Wellmer floral synchronization system [31] was used to maximize the amount of gynoecial tissue from floral stages 6-8 [12]. The Wellmer group kindly provided pAP1-AP1::GR; apl; cal seeds (KanR in Ler background - S.No. 1927). pSHP2-GAL4; pUAS-3xYpet dual construct plants (S.No. 1896) were crossed to pAP1-AP1::GR; ap1; cal. Lines homozygous for er, ap1, cal and the transgenes were selected in F2 and F3 generations (generating S. No. 2060). Because of the mixed ecotype cross (Col and Ler), lines that were erecta homozygous mutant and gave consistent YFP expression pattern and consistent inducibility of the AP1-GR activity were selected before the generation of protoplasts. Plants were grown under constant light and temperature at 22 °C to minimize circadian transcriptional fluctuations. To induce flowering in the transgenic plants, 20 μm of the synthetic steroid hormone dexamethasone (DEX) (Sigma, USA) in 0.015 % silwet was applied directly (spray application) ~30 days after planting [31]. Inflorescences were collected for protoplast generation ~ 120 h after DEX-induced floral synchronization. When collecting samples for protoplast preparation, 5-6 inflorescence heads were fixed for chloral hydrate clearing and DIC microscopy to determine the developmental stages of the flowers of the inflorescence samples. Additionally, before protoplasting, whole inflorescences were also collected and frozen immediately in liquid nitrogen for analysis of the transcriptional starting state of the non-protoplasted tissue (non-sorted samples, see Experimental design).
Experimental design
Material for RNA samples was gathered from batches of plants grown at one-week intervals to generate biological replicates (material from each week was considered as a biological replicate). To reduce variability between bioreplicates due to environmental heterogeneity within the growth chamber, each bioreplicate was drawn from a pool that contained plants grown within three different chamber positions. Four biological replicates of each of four tissue samples (YFP-positive, YFP-negative, all-sorted, and non-sorted) were collected (16 samples total). Whole inflorescences were collected for non-sorted samples and immediately frozen in liquid nitrogen before RNA isolation (i.e. these samples were not subjected to protoplasting nor FACS-sorting). The all-sorted samples represented the total population of protoplasts that come off the FACS machine after debris and broken cells are removed based on sorting gates (Additional file 1: Figure S1). The YFP-positive and YFP-negative protoplast populations are processed equivalently to the all-sorted samples except that a final FACS-sorting gate is used to divide the all-sorted protoplasts into YFP-positive and YFP-negative samples (Additional file 1: Figure S1). RNA was isolated from these three protoplast populations, as well as from entire non-protoplasted inflorescences (“non-sorted”). The YFP-positive, YFP-negative and all-sorted samples were prepared and collected as described below (Protoplast recovery and cell sorting).
Protoplast recovery and cell sorting
Protoplasts from the S. No. 2060 plants were generated according to the protocol of Birnbaum et al. [16], with adaptations for inflorescence plant material. Inflorescences (~200) were hand-collected with forceps and/or scissors and chopped with a “Personna double edge prep blade” (American Safety razor company; 74-002) within a 15 min period. Cell-wall polysaccharides were digested by immersing the chopped plant material in 10 ml of filter-sterilized solution B in a 50 ml falcon tube. Solution B (prepared according to Birnbaum et al.) is prepared from Solution A (10 mM KCl, 2 mM MgCl2, 0.2M MES, 600 mM Mannitol) to which cell wall digesting enzymes were added [final concentrations of 1.5% Cellulase (Yakult, Japan), 1% Pectolyase (Yakult, Japan) and 1% Hemicellulase (Sigma, USA)]. This mixture is then dissolved by gently swirling, covered in foil, and warmed in a water bath at 55 °C for ten minutes to inactivate DNAses and proteases. After cooling to room temperature, CaCl2 (2 mM final) and BSA (0.1% final) were added and the solution was filter-sterilized through a 25-micron filter.
After 1 h of incubation at room temperature with occasional gentle agitation, 10 ml of the protoplast-rich solution B was filtered through a 70-micron filter basket to a 50 ml falcon tube. A 10 ml rinse of solution A was applied directly to the material left in the 70-micron filter basket to rinse through any protoplasts left behind. Protoplasts were spun at 500 g, 10 °C for 10 min; the majority of the supernatant was removed by aspiration being careful not to disturb the protoplast pellet which is typically not tightly compacted. Protoplasts were resuspended in 25 ml of Solution A as a rinse step to remove cell-wall-digesting enzymes. Protoplasts were filtered again through a 50-micron filter mesh to a new tube adding 8 ml of solution A to again rinse through any protoplasts stuck in the filter. Protoplasts were then spun again at 500 g for 10 min. The majority of the supernatant was removed leaving 2 ml of the protoplasts in solution after the second centrifugation step. Propidium Iodide (5 micrograms/ml final) was added to the protoplasts (to allow separation of broken protoplasts) and a final filtering step though a 30-micron mesh filter (CellTrics, Partec) was carried out before loading onto the FACS machine.
Flow cytometry through FACS-sorting (Moflo XDP; Beckman Coulter Inc.) was used to isolate the YFP expressing cells from the total pool of cells. The FACS machine was equipped with a cooling device (set to 10 °C) and fitted with a 100-μm nozzle. Protoplasts were sorted at a rate of up to 10,000 events per second at a fluid pressure of 25 psi. Four sorting gates were set in an effort to collect the cleanest set of protoplasts and to eliminate debris and broken cells. A first gate based on size and granularity using side-scatter (SS) and forward-scatter (FS) parameters was used to select for intact protoplasts. Then a second gate was used to select for single cells and remove “doublets”. A third gate was used to select for cells that were negative for propidium iodide (PI) signal, as broken protoplasts and debris are preferentially stained by PI, which is excited by the 488 nm laser and emits at 617 nm. The total population of protoplasts that came off the FACS sorter machine after these gates constituted the all-sorted sample. In parallel, the YFP-positive protoplasts and YFP-negative protoplasts were separated into two collection tubes using the gates described above and one additional sorting gate based on the level of emission intensity in the green channel (529nm/28nm filter). Preliminary experiments with protoplasts that did not express the YFP transgene were used to set this gate and determine the levels of auto-fluorescence of the protoplasts. Protoplasts were collected directly into 14 ml tubes containing 4 ml of Trizol (Invitrogen/Life Technologies) and occasionally agitated during the approximately 40 min of sorting required to collect the protoplasts. Trizol was the method of choice as it maintains a high level of RNA integrity during tissue homogenization while also disrupting and breaking down cells and cell components. In order to minimize artifactual changes to transcript levels, the entire process of cell wall digestion, protoplast generation and FACS-sorting was kept under three hours. This procedure typically yielded between 300,000 and 500,000 YFP-positive protoplasts. These YFP-positive protoplasts typically represented approximately 0.5% of the total FACS sorting events. On average from four sorting trials representing four biological replicates, the number of cells collected and processed for each sample was: 575K for the YFP-positive, 1000K for the YFP negative and 493K for all-sorted samples.
RNA extraction and quantitative RT-PCR
Total RNA was extracted from sorted protoplasts collected in Trizol (keeping a 3:1 ratio of Trizol to sorted cells) and by modifying the Plant RNeasy Mini Kit, Qiagen protocol, as follow: collected cells in Trizol (4 ml total) were vortexed for 5 min at room temperature (RT) and 1 ml of chloroform (Sigma) was then added. The solution was vortexed again for 1 min at RT and centrifuged at 4,000 rpm for 10 min at 4 °C to separate phases; RNA from the aqueous phase (top layer) was carefully sucked up and mixed with 700 μl of Qiagen RLT buffer (Plant RNeasy Mini Kit, Qiagen) and 7μl of B-Mercaptoethanol (Sigma). 500 μl of 100% ethanol was added, solution was then transferred to a Qiagen MinElute column (Plant RNeasy Mini Kit, Qiagen) and spun in a 2 ml microfuge tube for 15 sec at ~ 10,000 rpm. 500 μl of RPE (Plant RNeasy Mini Kit, Qiagen) was added to the spin column, spun for 15 sec at ~10,000. 750 μl of 80% ethanol was added to the MinElute column and spun at ~10,000 rpm for 15 sec (twice) to ensure removal of all guanidine salts that may inhibit downstream applications. A final 5 min spin at top speed with the cap off was performed to remove trace amounts of ethanol. Total RNA was then eluted with 10 μl of RNAse-free water. A second elution was performed with another 10 ul of RNAse-free water. It is worth noting that one biological replicate (4th biological replicate) from the YFP-positive protoplasts was lost at this point, leaving only 3 biological replicates for this tissue sample and yielding a total number of 15 samples sequenced in two lanes and used for the experiment.
Prior to high-throughput sequencing, quantitative RT-PCR (qRT-PCR) was conducted on YFP-positive and YFP-negative samples using the 2−ΔΔCT method as suggested by Schmittgen and Livak [101] to assess relative gene expression of specific medial domain markers, SHATTERPROOF2 and NGATHA1. Total isolated RNA was quantified using fluorometric quantitation (Qubit RNA Assay Kit, Life Technologies, Inc.) for both YFP-positive and YFP-negative samples [~ 100 ng]. SuperScript III First-Strand Synthesis System (Invitrogen/Life Technologies) was used to generate cDNA (cDNA diluted 1:4 prior qRT-PCR analysis) from total RNA. qRT-PCR experiment assay was performed (Thermal Cyclers from Applied Biosystems) using a SYBR green mix (QuantiTect SYBR Green PCR Kits, Qiagen). Three biological replicates of the YFP-positive and YFP-negative samples were included and each biological replicate was assayed in triplicate. The expression levels of the ADENINE PHOSPHORIBOSYL TRANSFERASE1 (APT1) (At1g27450) gene was used for normalization.
Barplots
Barplots graphs were constructed using the ‘R’ package bear [102] and plyr [103] to calculate mean, standard error and confidence intervals and ggplot2 [104] to generate the plots.
Library preparation and mRNA sequencing
Total RNA isolated was quantified using fluorometric quantitation (Qubit RNA Assay Kit, Life Technologies, Inc.) and RNA quality was assessed using Agilent 2100 Bioanalyzer (Agilent). The RNA integrity number (RIN) for the 15 samples was higher than 7.3, which is above the Illumina threshold for library construction (> RIN 7). Strand-specific cDNA libraries were constructed from approximately 100 ng of total RNA using a NEB Ultra Directional Library Prep Kit for Illumina (New England Biolabs). The average size of the cDNA fragments was ~ 250 bp. The 15 bar-coded libraries were pooled and single-end sequencing was performed in a HiSeq 2500 Illumina (Illumina, Inc.) with ‘HiSeq SR Cluster Kit v4’ for the flow-cell and ‘HiSeq SBS v4’ for sequencing reagents. cDNA libraries were sequenced in 125-cycle plus 7-cycle for multiplexed samples. Sequencing was performed in two lanes of a flow-cell; all 15 libraries were sequenced twice and the results from the two independent lanes were analyzed as technical replicates. As no lane-specific effects were observed during data analysis, the reads from each lane were pooled for analysis of DEGs (see Table counts and technical replicates).
Bioinformatics analysis
All bioinformatics analyses were performed on a server cluster with 128 GB (gigabytes) of RAM, 16 cores (CPUs) and Ubuntu Linux-Distribution 12.04 operating system using ‘Simple Linux Utility Resource Management’ (SLURM) queue management system at the Bioinformatics Research Center (BRC) at the North Carolina State University, Raleigh, NC, USA.
Read Processing
Quality control and preprocessing of metagenomic data was performed using FastQC software [105]. Adapters and low quality sequences were filtered out with Ea-Utils software [106]. Reads with phred-like quality score (Q-score) > 30 and read length > 50-bp were kept and aligned against the TAIR10 Arabidopsis reference genome.
Sequence alignment to the Arabidopsis genome
Splice junction mapper TopHat2 (version 2.0.10) [107] was used to align filtered RNA-seq reads to the Arabidopsis thaliana TAIR10 genome (Ensembl annotation) downloaded from the iGenome database [108]. Default parameters for TopHat2 were used except for strand specificity (--library-type=fr-firststrand) to match to the first strand of cDNA synthesized (anti-sense to the mRNA) and maximal intron length (--I 2000), as it has been shown that the large majority of the known introns are smaller than the selected threshold [96]. To align reads solely and exclusively against TAIR10 annotated gene models, the arguments ‘--T’ (transcriptome only) and ‘--no-novel-juncs’ (no novel junction) were also included. Uniquely mapped reads were extracted from the TopHat2 output binary (BAM) file using samtools [109] and selecting for the “NH:i:1” two-character string-tag. Only uniquely mapped reads were used for downstream analysis.
Table counts and technical replicates
The ‘HTSeq: Analyzing high-throughput sequencing data with Python’ software [110] was used with default parameters except for the ‘stranded=reverse’ mode to generate tables-counts for downstream differential expression analysis for the ‘R’ packages edgeR [36] and DESeq2 [37].
Using edgeR, we assessed the gene level variance versus log gene expression level among technical replicates (corresponding to two lanes in the flow-cell of the Illumina HiSeq 2500). A linear-dependent Poisson distribution was observed for technical replicates (Additional file 1: Figure S5), in accordance with several studies [36], [67], [111]. Thus, differential gene expression analysis was performed using pooled technical replicates.
Plot of pooled gene level variance versus log gene expression level among technical replicates.
Gene expression and differential gene expression
Gene expression and differential gene expression analysis was carried out using ‘R’ packages edgeR [36] and DESeq2 [37] and the Linux-based Cufflinks program (v2.2.1) (G option) [35], for differentially expressed genes and transcripts [35]. To facilitate future use of these datasets, all the expressed genes identified and their expression values (F/RPKM) in YFP+/- (Additional file 4: Table S3) and all-sorted/non-sorted (Additional file 5: Table S4) are included as supplementary material.
Filters were applied to determine if a gene was detected, abiding by the suggestions of statisticians and bioinformaticians [112], [113], [114], [115], [37], [116] as a means to enrich for true DEGs, to reduce type I error and to improve P-value adjustment. The edgeR function [36] ‘cpm’ (counts per million) was used to discard those genes whose cpm was lower than a threshold of 2 reads per gene in at least 3 biological replicates, as suggested in the edgeR vignette. For cufflinks, a minimum RPKM of 5 was set for a gene to be expressed, following Suzuki et al. criteria [117]. According to Sims et al. 80% of genes can be accurately quantified with FPKM > 10 [69]. DESeq2 performs independent filtering using the ‘results’ function, as described in the DESeq2 vignette [37]. An FDR cutoff of < 0.01 was used to determine differentially expressed genes in all three programs. The Gene Regulatory Information Server (AGRIS) was used to identify transcription families in the dataset [118]. Enriched categories of transcription factors within the set of “cleaned” 363 YFP+/- DEGs was assessed with the online Transcription Factor Enrichment Calculator tool [51].
Venn Diagrams and heatmaps
Venn diagrams were constructed using the ‘R’ package VennDiagram [119] and the web-based tool package InteractiVenn [92]. Heatmaps were produced using the ‘R’ package pheatmap [120]. RPKM normalization by gene length and library size values were produced using the ‘rpkm’ function from edgeR [36]. To calculate gene length, a TAIR10 gene length list (CDS plus UTRs) was constructed by extracting length information from the TAIR10 GFF file with homemade Perl script. Genes with multiple isoforms were collapsed and length was calculated using the longest one. RPKM values were then calculated for clustering purposes and to have an intermediate point of comparison between Cufflinks, edgeR and DESeq2. Samples were clustered (default clustering) with parameters provided in the software. The ‘R’ package colorRamp [121] was used to produce a gradient of color values corresponding to gene-fold change values.
Gene set enrichment analysis
Gene Ontology (GO) enrichment tests were performed using the ‘R’ package topGO [122], with the ‘classic’ algorithm (where each GO category is tested independently) and the ‘fisher’ statistic test for ‘biological processes’, ‘molecular function’ and ‘cellular component’. Enrichment analysis was performed separately for all the genes that were differentially expressed between the YFP+/- samples and between the all-sorted/nonsorted samples. Network analysis of GO terms was performed using BiNGO [123] plugin for Cytoscape [124]. GO terms for the 268 genes identified as depleted in the YFP-positive sample, as well cellular component (CC) and molecular function (MF) for the YFP+/-sample can be found in Additional file 7: Table S6.
Dendrograms
The ‘R’ Dist function was used to compute a distance matrix using the spearman method (Spearman test rank correlation) and the ‘R’ Cor function to compute the variance of the matrix. To perform hierarchical clustering, the hclust function in ‘R’ was used. All statistical analyses were performed in ‘R’ v.3.0.2. Dendrogram plots were built using the ‘R’ ape package with edge.color = "blue".
Confocal microscopy
Confocal microscopy was performed using a Zeiss LSM 710 (Carl Zeiss, Inc. Thornwood, NY), microscope model (Zeiss Axio Observer Z.1), objective type Plan-Apochromat 20x/0.8 M27. Z-stack intervals were set to 2 μm and the total thickness of the stack was 62 μm.
Chloral hydrate clearing and Differential Contrast (DIC) Microscopy
Inflorescence samples were fixed in a solution of 9 parts ethanol: 1 part acetic acid for two hours at room temperature, and then washed twice in 90% ethanol for 30 min each wash. Inflorescences were transferred to Hoyer’s solution (70% Chloral hydrate w/v, 4% glycerol, 5 % gum Arabic) and allowed to clear for several hours to overnight. Samples were then dissected in Hoyer’s solution. The dissected inflorescence heads were mounted in Hoyer’s under coverslips and examined with DIC optics on a Zeiss Axioskop 2 to determine the floral stages.
In situ hybridization
For in situ hybridization analysis, Arabidopsis thaliana Col-0 flowers were fixed and embedded in paraffin as described previously [11, 70]. Sections of plant tissue were probed with digoxigenin-labeled antisense and sense RNA probes (Roche). Probes corresponded to nucleotides +686 to +920 of REM13 relative to the transcriptional start site of the CDS using the following oligos to amplify the template: REM13_ISH_Fwd 5’ AAAATAGAACGCGCATACCG 3’ and REM13_ISH_Rev 5’ TCGTGAACCAAACCGTGATA 3’. Hybridization and immunological detection were performed as described previously [11, 70].
Data availability
Illumina sequencing raw data (fastq) have been submitted to the Gene Expression Omnibus (GEO) database (accession GSE74458).
- List of abbreviations
- bp
- base pair
- CMM
- carpel margin meristem
- DEG
- differentially expressed gene
- DEX
- dexamethasone
- FACS
- fluorescence-activated cell sorting
- FDR
- false discovery rate
- FPKM
- fragments per kilobase of transcript per million mapped reads
- GSEA
- gene set enrichment analysis
- LCM
- laser capture microdissection
- nt
- nucleotide
- PI
- propidium iodide
- qRT-PCR
- quantitative RT-PCR
- RPKM
- reads per kilobase of transcript per million mapped reads
- ssRNA-seq
- strand-specific RNA sequencing
- TMM
- trimmed mean of M-values
- TSS
- transactional start site
- UTR
- untranslated region
- YFP
- yellow fluorescent protein
Competing interests
The authors declare that they have no competing interests.
Authors’ contributions
RGF, GHV, ES and SH coordinated and conceived of the study. AS, JV, LR built the GAL4/pUAS-based two-component reporter system. GHV, MFV and BS performed FACS-sorting, wet-lab and microscopy analyses. SM and LC performed in situ hybridization. GHV, QH and SH performed RNA-seq and related bioinformatic and computational analysis. GHV and RGF interpreted the data and drafted the manuscript. All of the authors revised and approved the final manuscript.
Description of additional data files
The following additional data are available with the online version of this paper. Additional data files Figures S1-S5 contain sorting gates used to select YFP samples and the re-sorting of the YFP-positive cells to assess sample purity (Fig. S1), qRT-PCR enrichment of medial domain genes SHP2 and NGA1 and the gene TUB (FIG. S2), expression profiles for the 363 differentially expressed genes (FC >4, FDR <0.001) across all 4 samples (YFP-positive, YFP-negative, all-sorted, non-sorted) (Fig. S3), Venn Diagram comparison of stressed induced genes due to protoplast/FACS-sorting procedure (Fig. S4) and gene level variance versus log gene expression level among technical replicates (Fig. S5). Additional file 2: Table S1 containing summary RNA-seq data (number of reads, mapped reads, uniquely mapped, etc.). Additional file 3 Table: S2 contains differentially expressed genes (DEGs) from the YFP+/- and all-sorted/nonsorted comparison. Additional file 4: Table S3 and Additional file 5: Table S4 contain all the expressed genes identified with three different programs between all the YFP+/- samples and all-sorted/non-sorted samples, respectively. Additional file B: Table S5 corresponds to raw high-throughput count data for YFP+/- and all-sorted/non-sorted comparison. Additional file 7: Table S6 contains the set enrichment analyses (GSEA) for YFP+/- and all-sorted/non-sorted comparison, including Biological Process (BP), Molecular Function (MF) and Cellular Component (CC). Additional file 8: Table S7 lists the transcription factors families identified in the DEGs from YFP+/- and their statistical enrichment Additional file 9: Table S8 contains isoforms expression, regulation of gene expression by alternative promoters and antisense transcripts identified by Cufflinks, edgeR and DESeq2. Additional file 10: Table S9 corresponds to the expression profile (RPKM) of the 86 genes described by Reyes-Olalde et al. [8] expressed in the medial domain. Additional file 11: Table 10 contains hormone (Auxin, GA, Ethylene) related-genes present in our dataset. Additional file 12 is the datafile to upload the web-based tool package “InteractiVenn”.
Table counts produced with ‘HTSeq: Analyzing high-throughput sequencing data with Python’ software (HtSeq) for all the samples (YFP-positive, YFP-negative, all-sorted, and non-sorted) including all biological replicates. Table 1 (YFPs) - YFP-positive (YFP_POS) and YFP-negative (YFP_NEG), Table 2 - allsorted (ALL_SORT) and non-sorted (NO_SORT). Biological replicates are indicated as B1, B2, B3, B4.
List of GO terms including biological process (BP), cellular component (CC) and molecular function (MF) categories identified in the gene set enrichment analyses (GSEA) for the enriched differentially expressed genes (95) and for the depleted DEGs (238) between YFP+/- from the 363 DEGs list (**p-value <0.01). DEGs. Over-represented GO categories (**p < 0.01) between all-sorted/non-sorted are also presented in Table 2, labeled as “ALL vs. NON_SORTED”.
Table 1 - List of 75 transcription factors and associated transcription factor family designations identified in the 363 YFP+/- DEGs. Table 2 - enriched categories of transcription factors within the set of the 363 DEGs using the online Transcription Factor Enrichment Calculator tool [51].
Table 1-Differential isoform expression that are significant (a=0.01) between the YFP+/- comparison and non-significant between the all-sorted/nonsorted comparison. Table 2- List of genes differentially expressed (a=0.01) by TSS/alternative promoter usage. Differentially expressed non-coding RNAs between the YFP+/- samples are reported in Table 3, 4, 5 for edgeR, DESeq2 and Cufflinks, respectively.
Expression profiles in Reads Per Kilobase of transcript per Million mapped reads (RPKM) of the 86 genes described by Reyes-Olalde et al. [8] in our dataset; YFP-positive (YFP_POS), YFP-negative (YFP_NEG), non-sorted (NO_SORT), all-sorted (ALL_SORT) and expressed in the medial domain of the arabidopsis gynoecium. Data from Mantegazza et al. [54] are also included; inflorescence meristem (IM_AVG), floral meristem (FM_AVG) and floral stage 3 (ST3_AVG).
List of genes associated with auxin, gibberellin and ethylene, their expression values, fold changes, and significance in our dataset.
Data file to upload the web-based tool package “InteractiVenn” [93].
Acknowledgements
We thank Sarah Schuett (CVM, flow cytometry facility, NCSU) and the Genomic Sciences Laboratory (GSL) research facility (NCSU) for library preparation and Illumina sequencing. We also would like to thank William Thomson (NCSU) and Emily Wear (NSCU) for FACS-sorting assistance; Frank Wellmer and Diarmuid O’Maoileidigh (Smurfit Institute, Trinity College of Dublin, Ireland) for the pAP1:AP1:GR ap1 cal1 floral synchronization system; Colleen Doherty (NCSU) for help with Cytoscape/BiNGO analysis; Maria Angels De Luis Balaguer (NCSU) for assistance with cross-platform expression correlation approaches; José Alonso and Ross Sozzani (NCSU) for thoughtful comments on the manuscript; Jigar Desai, Dmitry Grinevich and Colleen Doherty (NCSU) for help with the transcription factor enrichment analysis. Aureliano Bombarely (Virginia Tech) for help with GO terms analysis; and Eva Johannes (CMIF, Molecular Imaging Facility, NCSU) for laser scanning confocal microscope assistance. This work was funded by a grant from the National Science Foundation to RGF and SH (NSF IOS-1355019) and the FP7-PEOPLE-2013-IRSE FRUIT LOOK program (to ES, LC and RGF).
Footnotes
Gonzalo. H. Villarino: ghvillar{at}ncsu.edu
Miguel Flores-Vergara: maflores{at}ncsu.edu
Qiwen Hu: qhu{at}ncsu.edu
Bhupinder Sehra: bsehra{at}ncsu.edu
Linda Robles: lrobles1{at}uncc.edu
Javier Brumos: jbrumos{at}ncsu.edu
Anna Stepanova: atstepan{at}ncsu.edu
Silvia Manrique: silvia.manrique{at}unimi.it
Lucia Colombo: lucia.colombo{at}unimi.it
Eva Sundberg: eva.sundberg{at}slu.se
Steffen Heber: sheber{at}ncsu.edu
Robert G. Franks: rgfranks{at}ncsu.edu
References
- 1.↵
- 2.↵
- 3.↵
- 4.↵
- 5.↵
- 6.↵
- 7.↵
- 8.↵
- 9.↵
- 10.↵
- 11.↵
- 12.↵
- 13.↵
- 14.↵
- 15.↵
- 16.↵
- 17.↵
- 18.↵
- 19.↵
- 20.↵
- 21.↵
- 22.↵
- 23.↵
- 24.↵
- 25.↵
- 26.↵
- 27.
- 28.↵
- 29.↵
- 30.↵
- 31.↵
- 32.↵
- 33.↵
- 34.↵
- 35.↵
- 36.↵
- 37.↵
- 38.↵
- 39.↵
- 40.↵
- 41.↵
- 42.↵
- 43.↵
- 44.↵
- 45.↵
- 46.↵
- 47.↵
- 48.↵
- 49.↵
- 50.↵
- 51.↵
- 52.↵
- 53.↵
- 54.↵
- 55.↵
- 56.↵
- 57.↵
- 58.↵
- 59.↵
- 60.↵
- 61.↵
- 62.↵
- 63.↵
- 64.↵
- 65.↵
- 66.↵
- 67.↵
- 68.↵
- 69.↵
- 70.↵
- 71.↵
- 72.↵
- 73.↵
- 74.↵
- 75.↵
- 76.↵
- 77.↵
- 78.
- 79.
- 80.↵
- 81.↵
- 82.↵
- 83.↵
- 84.↵
- 85.↵
- 86.↵
- 87.
- 88.
- 89.↵
- 90.↵
- 91.↵
- 92.↵
- 93.↵
- 94.↵
- 95.
- 96.↵
- 97.
- 98.↵
- 99.↵
- 100.↵
- 101.↵
- 102.↵
- 103.↵
- 104.↵
- 105.↵
- 106.↵
- 107.↵
- 108.↵
- 109.↵
- 110.↵
- 111.↵
- 112.↵
- 113.↵
- 114.↵
- 115.↵
- 116.↵
- 117.↵
- 118.↵
- 119.↵
- 120.↵
- 121.↵
- 122.↵
- 123.↵
- 124.↵
- 125.↵
- 126.↵