Transcriptional profiling of the Arabidopsis abscission mutant hae hsl2 by RNA-Seq
© Niederhuth et al.; licensee BioMed Central Ltd. 2013
Received: 16 July 2012
Accepted: 14 January 2013
Published: 17 January 2013
Skip to main content
© Niederhuth et al.; licensee BioMed Central Ltd. 2013
Received: 16 July 2012
Accepted: 14 January 2013
Published: 17 January 2013
Abscission is a mechanism by which plants shed entire organs in response to both developmental and environmental signals. Arabidopsis thaliana, in which only the floral organs abscise, has been used extensively to study the genetic, molecular and cellular processes controlling abscission. Abscission in Arabidopsis requires two genes that encode functionally redundant receptor-like protein kinases, HAESA (HAE) and HAESA-LIKE 2 (HSL2). Double hae hsl2 mutant plants fail to abscise their floral organs at any stage of floral development and maturation.
Using RNA-Seq, we compare the transcriptomes of wild-type and hae hsl2 stage 15 flowers, using the floral receptacle which is enriched for abscission zone cells. 2034 genes were differentially expressed with a False Discovery Rate adjusted p < 0.05, of which 349 had two fold or greater change in expression. Differentially expressed genes were enriched for hydrolytic, cell wall modifying, and defense related genes. Testing several of the differentially expressed genes in INFLORESCENCE DEFICIENT IN ABSCISSION (ida) mutants shows that many of the same genes are co-regulated by IDA and HAE HSL2 and support the role of IDA in the HAE and HSL2 signaling pathway. Comparison to microarray data from stamen abscission zones show distinct patterns of expression of genes that are dependent on HAE HSL2 and reveal HAE HSL2- independent pathways.
HAE HSL2-dependent and HAE HSL2-independent changes in genes expression are required for abscission. HAE and HSL2 affect the expression of cell wall modifying and defense related genes necessary for abscission. The HAE HSL2-independent genes also appear to have roles in abscission and additionally are involved in processes such as hormonal signaling, senescence and callose deposition.
Abscission is the programmed separation of plant organs [1–5]. It is both a developmental and environmental response, allowing non-functioning or infected organs to be discarded. Abscission occurs at a specialized cell layer called the Abscission Zone (AZ), which develops at the base of the abscising organ. Secretion of hydrolytic and cell wall modifying enzymes in the AZ results in the breakdown of the pectin-rich middle lamella, leading to organ separation. This is followed by the formation of a protective scar-layer of waxy substances over the newly exposed AZ.
In Arabidopsis thaliana the floral organs abscise soon after pollination. Abscission requires the function of two receptor-like protein kinases (RLK) encoded by the genes HAESA (HAE) and HAESA-LIKE 2 (HSL2). While single mutants appear phenotypically normal, the double mutant has a complete loss of abscission [6–8]. Plants with mutations in INFLORESCENCE-DEFICIENT IN ABSCISSION (IDA) have similar phenotypes to hae hsl2 mutants [6, 9]. IDA is predicted to encode a small secreted protein  and has been placed upstream of HAE and HSL2, which suggests IDA is the putative ligand for HAE and HSL2 [6, 7]. Plants expressing a double RNAi transgene targeting the MAP Kinase Kinases MKK4 and MKK5 also are defective in abscission and constitutively active versions of either MKK4 or MKK5 can rescue hae hsl2 and ida mutants. The MAP Kinases MPK3 and MPK6 are known targets of MKK4/MKK5 and expression of dominant negative versions of MPK6 in a mpk3 mutant display abscission defective phenotypes . This evidence suggests a pathway consisting IDA, HAE, HSL2, and a MAP kinase cascade regulate the initiation of abscission.
The roles of polygalacturonases (PG), xylogulcan endo-transglycosylase/hydrolases (XTHs), and cellulases have been associated with abscission in tomatoes, cotton, and roses [10–13]. A previous microarray study of stamen abscission zones in Arabidopsis has shown that expression of many genes encoding hydrolytic and cell wall modifying enzymes are increased prior to abscission [14, 15]. These include the PG encoding genes PGAZAT/ADPG2 [16, 17] and QRT2. Interestingly adpg2/qrt2 double mutant plants have a delayed abscission phenotype  which provides functional evidence these PGs have a role in abscission.
Transcriptional profiling using high throughput next-generation sequencing (RNA-Seq) has emerged in recent years as a superior alternative to microarrays . Here we report the use of RNA-Seq to identify differentially expressed genes in hae hsl2 flower receptacles. Our work suggests that HAE and HSL2 act to promote the expression of hydrolytic and cell wall modifying enzymes necessary for abscission and that disruption of the HAE HSL2 signaling pathway results in reduced expression of these enzymes explaining the loss of abscission phenotype in the hae hsl2 mutants.
Arabidopsis flower development can be broken down into 20 developmental stages based on morphology . Abscission studies typically focus on stages 12-17 (Additional file 1), with organ loss occurring at stage 17. At stage 15 the floral organs are still attached, but by stage 16 the organs have begun to wither and will fall if force is applied. By stage 17 all the floral organs have abscised from the still green siliques. Expression of HAE, HSL2, and IDA in floral AZs increases from stage 12 reaching its peak in the latter parts of stage 15 . Based on these observations we hypothesized that initiation of abscission occurs during stage 15 and that expression differences between wild type and hae hsl2 would be observed at this point.
RNA-Seq reads and mapping statistics
Uniquely Mapped to a Single Gene
Percent Uniquely Mapped to a Single Gene
hae hsl2 1
hae hsl2 2
hae hsl2 3
The global transcriptome changes observed when comparing wild type to hae hsl2 floral receptacles suggest several conclusions. First, the high correlation between all samples shows that there was little variation in the collection of floral receptacles and that these were technically consistent. Secondly, expression of the majority of genes is unaffected in hae hsl2, suggesting that this pathway targets a limited number of genes. This is supported morphologically as stage 15 hae hsl2 flowers appear normal and have normal looking AZs . Finally, the majority of the differentially expressed genes are lower in the mutant relative to wild type. This suggests the HAE HSL2 signaling pathway primarily activates expression of target genes.
A global view of the functions of the differentially expressed genes and the underlying biology can be obtained by examining their gene ontology (GO) . This is a system that categorizes genes into groups of terms based on their predicted or experimentally derived Molecular Function, Biological Process, and Cellular Component. Molecular Function includes the most genes and is usually based on sequence similarity. The Biological Process categories are typically derived empirically and as a result tend to be more stringent, but also limited, having fewer annotated genes. The Cellular Component is largely a prediction of where the gene product is localized. Looking for overrepresentation of GO terms indicates that more genes in a list are represented in that category than what would be expected by random chance and can reveal trends in the data.
Genes showing higher expression in the mutants relative to wild type were enriched for lipid metabolism and transporter activity (Figure 2D). The genes that are annotated to have a role in transporter activity are primarily involved with either ion transport or secondary active transmembrane transport (Additional file 3B).
The genes with lower expression in hae hsl2 largely fall in one of two categories based on GO term enrichment; those involved with middle lamella/cell wall degradation and remodeling, and those involved in defense against pathogens. Many of the hydrolases identified as having lower expression in the hae hsl2 samples are also predicted to be localized to the cell wall or endomembrane system (Figure 2C), and are annotated as being involved in cell wall modification (Additional file 3A). In particular, this group of genes encodes the enzymes that breakdown pectin, the PGs, pectinesterases, and pectin lyases, that are necessary for cell separation [1–5]. In the case of the PGs, two have been previously shown to be involved in abscission [16–18], PGAZAT/ADPG2 and QRT2. These results show that HAE and HSL2 regulate the breakdown of the middle lamella during abscission. Shedding of an organ also exposes a fresh surface that is potentially susceptible to infection. It has been previously proposed  that plants activate defense responses as a protective measure prior to abscission. This idea is supported by our results where failure to abscise coincides with relatively lower expression of defense related genes that have roles in response to biotic stressors, particularly bacteria and fungi. After the organ has abscised protective layers of suberin and lignin form over the AZ . Several genes involved in the biosynthesis of the phenylpropanoid precursors of these compounds have lower expression in hae hsl2 mutants. For example, the fatty acyl-coenzyme A reductases FAR4 and FAR5 and the acyltransferase GPAT5 are essential to the biosynthesis of suberin [24, 25] and show lower expression in hae hsl2, along with three genes involved in lignin biosynthesis (Additional file 3A). The possible roles of the genes that are expressed at relatively higher levels in the hae hsl2 mutants are less clear. Many of the genes are predicted to function in ion and secondary active transmembrane transport. Potentially these may be involved in the transport of nutrients to the floral organs and are switched off prior to or during abscission.
Previous microarray studies have examined gene expression changes in stamen AZs across flower development stages 12, 13 and early, mid, and late stage 15 . In their analysis Cai and Lashbrook  used a linear mixed model to identify 551 genes in eight expression clusters that showed the most significant changes across all developmental stages. Because this analysis focused on changes in gene expression over time many of the genes they identified as having differential expression are not differentially expressed in hae hsl2 relative to wild type. We reanalyzed the raw microarray data to determine the wild type expression patterns of the differentially expressed genes identified from the RNA-seq experiments. The Arabidopsis ATH1 Genome Array used by Cai and Lashbrook with TAIR10 annotations has probes for 21,144 genes. Only 219 of the 277 genes with lower expression in the double mutant and 53 of the 72 genes with higher expression are found in the microarray expression data and all subsequent analysis was done using these smaller datasets.
Cluster 2 (Figure 3B), with 33 genes, shows relatively low expression in stages 12 and 13, increases through early and mid-stage 15 and decreases slightly in late stage 15. Transmembrane transport and oxidation reduction are found in the biological process category for cluster 2 (Additional file 6B). Similarly in the molecular function there is enrichment for electron carrier activity, oxidoreductase activity, and transporter activity, particularly ion transport. Many of these genes are localized to the endomembrane system or to the membrane and these terms show enrichment in this cluster for the cellular component.
Cluster 3 (Figure 3C) has 42 genes which show relatively higher expression in stage 12, reduced expression through early and mid-stage 15 and an increase in expression in late stage 15. Although Cluster 3 has peak expression in stage 12, the expression profiles across stages 13 to late stage 15 is similar to Cluster 1. This cluster shows the least overrepresentation of GO terms (Additional file 6C). In biological process there is enrichment for oxidation reduction, carbohydrate metabolic process, and cellular amino acid and derivative metabolic process. There was no enrichment in either cellular component or molecular function.
The dynamics of expression in these clusters and the functions of their genes show that most genes affected by HAE HSL2 have relatively lower expression in earlier flower development stages when HAE, HSL2, and IDA expression is also low. In later developmental stages, the first genes to that display an increase in expression tend to be those involved in transmembrane transport and oxidative reduction. It is not until late in stage 15 that there is activation of the genes associated with cell separation, defense responses, and the formation of the protective scar tissue over the newly exposed AZ.
Genes involved in several hormonal signaling pathways show increased expression at late stage 15 (Additional file 9A). Of particular interest are those involved in ethylene biosynthesis and signaling, as ethylene regulates the timing of floral organ abscission [1, 3]. This includes several ethylene biosynthesis genes such as the s-adenosylmethionine synthetase SAM1, the ACC synthase ACS8, and two ACC oxidases ACO1 and ACO2. The ethylene receptor ETR2 also increases in expression and is known to be inducible by ethylene signaling . Expression of genes associated with abscisic acid signaling also increase, such as the genes encoding protein phosphatase 2C ABA INSENSITIVE 2 (ABI2) and the transcription factors ABI4 and ABI5.
Both ethylene and abscisic acid regulate floral senescence  and markers of senescence show enrichment in the HAE HSL2-independent gene set. This corresponds to the senescence and withering of the floral organs during abscission. Callose deposition and localization also shows enrichment and is another response of abscisic acid signaling . Physiologically this is supported by work that has shown increases in callose deposition in abscising leaves of Phaseolus vulgaris  and may be necessary to prevent water loss after abscission.
Jasmonic acid signaling and biosynthesis genes decrease in expression (Additional file 9B) from stage 12 to 15. This includes the allene oxide synthase AOS, the allene oxide cyclases AOC3 and AOC4, and the lipoxygenases LOX2 and LOX4. Interestingly aos mutants have a delayed abscission phenotype that is enhanced synergistically by mutations in the ethylene insensitive ein2 and the ABA-deficient mutant aba2 . Also decreasing in expression are cellulose biosynthesis genes such as CELLULOSE SYNTHASE 5 (CESA5), CELLULOSE SYNTHASE CATALYTIC SUBUNIT 7 (CESA7), and multiple cellulose synthase-like genes. This suggests there is inhibition of cell wall biogenesis while at the same time that cell wall modification is occurring. While pectinesterases increase in expression in a HAE HSL2-dependent manner during stage 15, pectinesterase inhibitor genes decrease in expression independently of HAE HSL2.
There is likely cross talk between HAE HSL2-dependent and HAE HSL2-independent processes, particularly by ethylene (Figure 6). Ethylene signaling may affect the timing and degree of IDA expression . Additionally ethylene-dependent abscission signaling could be modulated by MAP kinase signaling  either through regulation of the ACC synthases and ethylene biosynthesis  or through the transcription factor EIN3 . Some classes of genes are found to varying extents in both the HAE HSL2-dependent and HAE HSL2-independent categories. For example there are HAE HSL2-independent defense response, lignin biosynthesis, and suberin biosynthesis genes. There are even some HAE HSL2-independent hydrolases, particularly XTHs, which may explain the observation that hae hsl2 and ida mutants initially undergo a decrease in the force required to remove the floral organs before increasing again .
Abscission is a tightly regulated process resulting in the breakdown of the middle lamella at the AZ and shedding of the abscising organ. This process is regulated by the two RLKs HAE and HSL2 which operate through a MAP kinase cascade. However the targets of this signaling pathway were previously unknown. Our work reveals the HAE HSL2 signaling regulates the expression of hydrolytic and cell wall modifying enzymes necessary for the breakdown of the pectin rich middle lamella and the expression of defense response genes during abscission. Disruption of HAE HSL2 signaling prevents normal expression of these genes, resulting in the loss of abscission phenotype observed in hae hsl2 mutants. These genes form three distinct expression clusters with most genes having peak expression in late stage 15, right before organ separation. Testing of differentially expressed glycosyl hydrolases in ida mutants shows that many of the same genes are co-regulated by IDA and HAE HSL2 and support the role of IDA in the HAE and HSL2 signaling pathway. Comparison to stamen AZ microarrays shows there are HAE HSL2-independent processes that includes the ethylene, abscissic acid, and jasmonic acid hormonal signaling pathways. Responses of these pathways, such as senescence and callose deposition increase prior to abscission, while other processes like cellulose synthase decrease. This demonstrates the complexity of abscission and how multiple pathways integrate to lead to organ loss.
Arabidopsis thaliana Columbia-0 wild type, hae-3 hsl2-3, and ida-2 plants were grown in growth chambers at 22C under 16 h light 8 h dark. At 7 weeks of age, stage 15 flowers were collected and the flower receptacles were immediately dissected and frozen in liquid nitrogen. Receptacles were dissected by cutting off the stems from the base of the flower and slightly above the base of the stamen with a surgical knife. 20 receptacles were pooled per sample, each receptacle from a separate plant. All samples were collected between 4 and 6 pm to minimize circadian effects. For qPCR this was repeated with stage 12, 13, and 14 flower receptacles collected in addition to stage 15. Two sets of wild type samples were collected for qPCR, one grown alongside hae hsl2 the other grown later alongside ida.
RNA was isolated using TRIZOL reagent (Invitrogen). DNA was removed using DNase TURBO (Ambion) and then cleaned up using RNAeasy Mini Kit (Qiagen). RNA-Seq libraries were prepared using the TruSeq RNA sample preparation kit (Illumina) following the manufacturer’s protocol. All six samples were pooled together and run on three lanes, each on a separate flow cell of an Illumina HiSeq 2000.
Reads were quality trimmed using FASTX FASTQ Quality Trimmer version 0.0.13  with a minimum quality score of 13 and a minimum length of 32. Because adaptor sequence was present at the three prime end of some reads, these were further trimmed using CUTADAPT version 0.9.5  with a minimum overlap of 2 and a minimum length of 32. Reads were then aligned back to the TAIR10 version of the Arabidopsis genome  using TOPHAT version 1.3.1  supplied with the TAIR10 GFF at default settings. Sam and bed files were generated using SAMtools version 0.1.18  and BEDTools version v2.12.0 . Read counts for each gene was quantified using HTseq-count with the settings stranded=no, mode=union, and type=gene .
Differential Expression was determined using the DESeq version 1.5.6 . This was done using the sequence of commands: newCountDataSet, estimateSizeFactors, estimateDispersions, and nbinomTest. For the estimateDispersions function the settings used are method = “per-condition”, sharingMode = “maximum”, fitType = “parametric”.
Gene Ontologies were analyzed for term enrichment using the agriGO Single Enrichment Analysis tool  with TAIR10 GO annotations. The full GO contains thousands of specific terms while the GO Slim is a reduced set of broader higher level terms that is easier to represent graphically. Both the full GO and GO Slim were used. The hypergeometric test was used with Benjamini-Hochberg FDR correction and a p < 0.05. Lower expressed and higher expressed genes were analyzed separately.
Primers for qPCR (Additional file 10) were designed using Primer3 software  to have a Tm of 59-61 C, with a length of 22-26 nts and a product length of 60-150 bps . Reference genes were chosen from Czechowski et al.  for stability across tissue type and developmental stage. These were then checked against RNA-Seq results for stability in the hae hsl2 background. RNA was isolated using TRIZOL reagent (Invitrogen) and DNA was removed using DNase TURBO (Ambion). Samples were then cleaned up using RNAeasy Mini Kit (Qiagen). 1000 ng of RNA was used to make cDNA with the SuperScript III First Strand Synthesis Kit (Invitrogen). 200 nm of each primer were added to each well and dried overnight. For qPCR 2.5 uL of Platinum SYBR Green qPCR Supermix-UDG (Invitrogen) and 1 uL of cDNA (1 ng/uL) for a total of 5 uL was used in each well. To reduce pipetting errors master mixes of the cDNA and 2x reaction mix were made prior to dispensing. For each sample 3 biological reps were used and repeated 3 times for technical replication. Real-time PCR was done on a CFX384 Touch Real-Time PCR Detection System (BioRad) at 50C for 2 minutes, 95C for 2 mins, and 45 cycles of 95C for 15 s, 55C for 30s, and 72C for 30s followed by a melting curve analysis. qPCR was analyzed on CFX Manager Software (BioRad) using the ΔΔCt method . Statistical significance was determined using Student’s t-test.
Microarray data from Cai and Lashbrook  was downloaded from Array Express  and reanalyzed using ROBIN  using the probe logarithmic error intensity estimate (PLIER) algorithm  and the linear models from the limma package . Stage 12 was used as a baseline of comparison to later stages resulting in the following stage categories: stage 12-stage 13, stage 12-early stage 15, stage 12-mid stage 15, and stage 12-late stage 15. Differentially expressed genes were considered significant with a FDR adjusted p < 0.05.
Microarray signal intensities extracted for the list of differentially expressed genes in hae hsl2 were normalized to a range of -1 to 1 and centered on a mean of 0 using Cluster 3.0 . These were imported into R and the number of k-Means clusters was estimated to be 3 using the gap statistic gap . Final clustering was done by k-Means clustering in Cluster 3.0 using Euclidean distance as the Similarity Metric, a k = 3, and 1000 runs.
Because RNA-Seq detects a broader range of transcripts than microarrays, which are limited to the probe set, any genes not found on the array were first filtered from the RNA-Seq data. Log2 fold changes of differentially expressed genes in both RNA-Seq and mircoarray data was then compared using VennMapper software  looking for overlap in genes with a 2 fold or greater difference in expression. Genes that were expressed at lower and higher quantities in hae hsl2 relative to wild type were treated as separate categories and compared in a pairwise fashion to each stage.
Raw reads, bed alignment files, and raw gene counts, are available under the GEO accession number GSE35288 at the NCBI Gene Expression Omnibus .
INFLORESCENCE-DEFICIENT IN ABSCISSION
MAP KINASE KINASE 4
MAP KINASE KINASE 5
MAP KINASE 3
MAP KINASE 6
POLYGALACTURONASE ABSCISSION ZONE A. THALIANA/ARABIDOPSIS DEHISCENCE ZONE POLYGALACTURONASE 2
False Discovery Rate
FATTY ACYL-COENZYME A REDUCTASE 4
FATTY ACYL-COENZYME A REDUCTASE 5
S-ADENOSYLMETHIONINE SYNTHETASE 1
1-AMINO-CYCLOPROPANE-1-CARBOXYLATE SYNTHASE 8
ACC OXIDASE 1
ACC OXIDASE 2
ETHYLENE RESPONSE 2
ABA INSENSITIVE 2
ABA INSENSITIVE 4
ABA INSENSITIVE 5
ALLENE OXIDE CYCLASE 3
ALLENE OXIDE CYCLASE 4
CELLULOSE SYNTHASE 5
CELLULOSE SYNTHASE CATALYTIC SUBUNIT 7
ETHYLENE INSENSITIVE 2
ABA DEFFICIENT 2
ETHYLENE INSENSITIVE 3.
The authors would like to thank members of the DNA Core Facility at the University of Missouri-Columbia for performance of RNA-Seq library construction and sequence generation Dr. William Spollen for advice on alignments. We are thankful to Dr. Jim Birchler for comments on the manuscript. We would like to also thank the rest of the Walker laboratory for their advice and encouragement. This work is funded by a grant from NSF (MCB0743955) to JCW.
This article is published under license to BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.