Introduction

Plant hormones are known to be important in many aspects of plant growth and development. Recent evidence has begun to suggest that a much bigger suite of signaling molecules exist in plants than previously understood (Lindsey et al. 2002). This includes a diverse group of putative small proteins that may function as signaling and regulatory molecules important in various aspects of plant development and responses to external stimuli.

The first well-characterized plant peptide hormone, systemin, was discovered as an 18 amino acid (AA) peptide in wounded tomato leaves responsible for transmitting a systemic signal in response to insect herbivory (Pearce et al. 1991). Systemin is encoded as a single peptide at the C-terminal end of a larger precursor protein (about 200AA; McGurl et al. 1992). Recent work has identified a likely candidate for a systemin receptor in tomato, a receptor with an additional role in brassinolide signaling (Scheer and Ryan 1999, 2002; Montoya et al. 2002). Since the initial characterization of systemin, multiple small proteins that appear to act as hormones have been described with roles in plant development.

For example, in plant development, the CLAVATA pathway consists of CLV1, a LRR-RLK (Clark et al. 1997), CLV2, a LRR-containing protein with a transmembrane domain but lacking an intracellular kinase domain (Jeong et al. 1999), and CLV3, a 12-AA-secreted peptide (Fletcher et al. 1999; Ito et al. 2006; Kondo et al. 2006; Sawa et al. 2006). The CLV genes appear to act in a self-regulatory loop with the homeobox gene, WUSCHEL (WUS), to maintain a functional shoot apical meristem (Schoof et al. 2000; Fletcher 2002). The importance of peptides in plant development has become clear as additional peptides and peptide roles have been uncovered. Currently, over 10 families of peptides have been found to play a role in plant development (Farrokhi et al. 2008). However, our understanding of many of these roles is still limited and much additional research is required to determine the functions of these proteins in plant biology.

The DEVIL (DVL) gene family, encoding a group of putative small proteins in Arabidopsis, has been shown to play a role in plant development. The founding member of this family, devil1-1 dominant (dvl1-1D), is a dominant, activation-tagged mutant that shows pleiotropic developmental phenotypes (Wen et al. 2004). dvl1-1D plants have shorter but wider rosette leaves resulting in a rounded leaf appearance and are shorter in stature compared to WT plants. The flowering inflorescence is more tightly clustered than WT, and the individual flowers have shortened sepals and petals, which results in a protrusion of the pistil prior to flower opening. The flowers are fully fertile, and seed set does not appear to be significantly altered. The valves of the silique show an interesting horn-like protrusion on the distal end of the silique that was the namesake of this mutant.

Rotundifolia4-1D (rot4-1D) is another member of the DVL family isolated in an activation-tagging screen for modified leaf morphology in Arabidopsis (Narita et al. 2004). rot4-1D has a similar phenotype as dvl1-1D of smaller, rounded leaves. The size of the leaf cells does not appear to be altered. However, the number of cells is reduced, especially in the longitudinal axis of the leaf, resulting in the shorter leaf phenotype. Cloning of the tagged genes uncovered small ORFs encoding predicted small proteins of 51AA (dvl1-1D) and 53AA (rot4-1D). These genes belong to a larger gene family (in Arabidopsis, >20 members) with members found thus far only in the seed plants (Narita et al. 2004; Wen et al. 2004). A domain at the C-terminal end is conserved in members of this gene family, including two highly conserved cysteine residues. In Arabidopsis, the DVL family members are likely to be partly redundant in function. dvl1-1D and rot4-1D have similar phenotypes. Ectopic overexpression of other family members also results in similar phenotypes (Wen et al. 2004; Wen and Walker 2006).

Overexpression of DVL (dvlOE) was shown to result in a downregulation of the MADS-box gene, FUL/AGL8 (Wen et al. 2004). However, our understanding of the genetic interactions in which DVL plays a role is still incomplete. To gain a better understanding of these interactions, we developed a large-scale genomic screen using a dvl transgene that could be experimentally controlled. In this screen, we uncovered many responses at the mRNA level of genes that play roles in transcription processes. One of these genes, NAC1, a NAC-domain transcription factor (Olsen et al. 2005), was further investigated. A loss-of-function (LOF) of nac1 reverses a dvl overexpression phenotype to a WT-like phenotype. Furthermore, this nac1 LOF results in an epigenetic change that is stable in following generations.

Materials and Methods

To identify genes that are transcriptionally responsive to DVL overexpression, microarray analysis to profile different transgenic Arabidopsis lines in a Columbia (Col WT) background was used. These transgenic lines used a GVG construct (glucocorticoid inducible) that allows for control of the transgene expression by applications of a steroid hormone, dexamethazone (dex) to the transgenic plants (Aoyama and Chua 1997). A control line was transformed with the GVG construct but without a coding sequence inserted in the multiple-cloning site. This line was indistinguishable from Col WT with and without dex treatment or mock treatment. A second line was also transformed with the GVG construct containing the DVL4 coding sequence. Without dex treatment or mock treatment, this line was indistinguishable from Col WT but displayed a typical dvlOE phenotype with dex treatment. The GVG construct was generated as described in (Aoyama and Chua 1997) and transformed into Arabidopsis using Agrobacterium-mediated transformation (Clough and Bent 1998).

Plants were grown in artificial greenhouse potting media in standard growth chambers at 22°C and 16-h days. Plants were treated when they reached the six expanded rosette leaf stage (vegetative) with mock treatment or 30-μm dex in water. Three different treatments were completed: Empty vector control with dex treatment; GVG DVL4 with mock treatment; and GVG DVL4 with dex treatment. Three biological replicates were collected for each treatment.

At 9–10 h post-treatment, the largest rosette leaf was collected from 20 plants. These leaves were pooled and were considered one biological replicate. A total of three biological replicates (each pool of leaves from 20 plants) were collected for each treatment; thus, three microarray chips were hybridized per treatment. Total RNA was isolated from the leaf samples using TRIzol reagent (Invitrogen) and purified using the RNeasy Midi Kit (Qiagen) according to manufacturer’s instructions. cDNA synthesis and labeling with biotin was completed according to chip manufacturer’s instructions (Affymetrix). Hybridization to Affymetrix Arabidopsis ATH1 Genome GeneChips was completed according to chip manufacturer’s instructions (Affymetrix) using an Affymetrix GeneChip Hybridization Oven 640 and GeneChip Fluidics Station 450 at the DNA Core, University of Missouri, Columbia. The microarrays were then scanned and data acquired using an Affymetrix GeneChip Scanner 3000 driven by GCOS 1.2 software (Affymetrix) at the DNA core. Computational analysis was completed using dChip software (www.dchip.org). The empty vector control with dex treatment was used to correct for any transcriptional responses due to dex treatment alone. Global analysis of the datasets was completed using MapMan software, version 5.0 (Thimm et al. 2004).

Quantitative Reverse Transcriptase PCR (QPCR) was completed on total RNA extracts from flower bud tissue. Extracts were made using TRIzol reagent (Invitrogen Life Technologies) per manufacturer’s instructions. Contaminating DNA was removed by Turbo DNA-free DNase treatment (Ambion) and cDNA was made using 1 μg of total RNA and ImProm-2 Reverse Transcriptase (Promega) following the manufacturer’s instructions. QPCR was completed using an Optican 2 Real-Time PCR machine (MJ Research) and ABsolute QPCR SYBR Green Premix (ABgene) per manufacturer’s instructions. All sample c(t) values were log transformed and normalized using EF1α control reactions by first calculating the change relative to a control sample and then normalizing with the EF1α reactions. Primers are listed in Table S2.

T-DNA insertion LOF nac1 lines were obtained from the Arabidopsis Biological Resource Center (ABRC; Columbus, OH). Homozygous lines were isolated or confirmed by PCR screening. Primers are listed in Table S2. Direct sequencing identified the insertion sites.

Results and Discussion

Inducible DVL Expression System

To aid in screening for genetic interactions with the DVL genes, an inducible system in Arabidopsis using the GVG vector (Aoyama and Chua 1997) was created. For this inducible system, DVL4 was cloned and placed downstream of the GVG binding site, allowing for transcriptional activation of DVL4 with the application of dex to the plant. Multiple homozygous transgenic lines were screened for the absence of any DVL phenotype without dex application and a DVL overexpression phenotype with the application of dex (Fig. 1). One line with consistent DVL overexpression phenotypic responses following dex application was selected for further analysis. Control lines, which only contained an empty vector, were also generated. To confirm that no leaky expression of DVL4 from the transgene was present before dex application, steady-state DVL4 transcript levels were tracked using QPCR. Steady-state transcript levels of DVL4 were identical to WT before the application of dex and strongly induced following foliar dex (30 μM) application (Fig. 2).

Fig. 1
figure 1

Inducible GVG DVL4 transgenic plants. Silique phenotypic responses in inducible GVG DVL4 and control plants after dex treatment. Columbia WT (a); dvl1-1D (b); control (empty vector) plants, no dex treatment (c); control plants, dex treatment (d); inducible GVG DVL4 plants, no dex treatment (e); inducible DVL4 plants, dex treatment (f). This range of silique shapes in (f) is similar to that seen in CaMV 35S DVL transgenic plants (Wen and Walker 2006). Scale bar: 1 cm

Full size image
Fig. 2
figure 2

DVL4 transcript accumulation in dex-treated inducible GVG DVL4 plants. QPCR was used to quantify the amount of DVL4 transcript present, relative to WT Columbia. A biological replicate showed a similar trend and is presented in Fig. S1

Full size image

Gene Discovery for DVL Interactors

This inducible-DVL4 system was used in a microarray-based experiment. The inducible-DVL4 system was used because this should allow the early transcriptional responses to be uncovered without the added complexity of secondary responses that might be associated with the dvl1-1D plants. To achieve this, samples were collected shortly after induction of DVL4 in an attempt to enrich for the rapid changes in gene expression resulting from overexpression of DVL.

Control plants (empty vector-transformed plants) and inducible DVL4 plants were grown in a growth chamber to the six-rosette leaf stage. At this point, all the plants are treated with 30 μM dex or mock treated. Leaf samples were collected at 9–10 h post-treatment. Three replicates of each treatment were prepared, labeled, and hybridized to Affymetrix ATH1 Genome GeneChips. Three different treatments were completed (Table 1). The data were then analyzed using dChip software (www.dchip.org; Li and Wong 2001a, b) that generates a statistical trend for each of the probe sets. Based on the analysis, a common trend was observed in which dex alone induces the expression of many genes. Therefore, an analysis procedure was set up to find all genes that were responsive to dex-induced DVL4 expression relative to mock treatment and also responded with a significant difference in expression from that of the dex-treated control plants. The full dataset has been deposited at the National Center for Biotechnology Information Gene Expression Omnibus (http://www.ncbi.nlm.nih.gov/geo/) database under record GSE8975. A summary of the analysis is presented in (Table 2) and a list of the most responsive genes is presented in (Tables S1 and S2).

Table 1 Summary of treatments
Full size table
Table 2 Summary of microarray data
Full size table

In order to gain a clearer picture of the responses found in the dex-treated inducible DVL4 plants, the microarray datasets were further analyzed using MapMan software (Thimm et al. 2004). MapMan software graphically organizes the datasets based on functional classifications. In this analysis, it was found that genes involved in transcription regulation were overrepresented in these datasets. Indeed, transcription factors made up one of the largest subgroups of the genes involved in regulatory processes that were responsive to DVL4 overexpression (Fig. S3). QPCR was used to confirm the responses seen in the microarray samples for NAC1 (discussed below) and At2g41070, a basic region/leucine zipper motif transcription factor (Jakoby et al. 2002). At2g41070 mRNA was found to overaccumulate approximately 30-fold in the dex-induced DVL4 line relative to the dex-treated control line, confirming the trend from the microarray data. Therefore, we decided to focus in greater detail on this subgroup.

Fig. 3
figure 3

Segregation analysis of dvl1-1D and nac1 crosses. A crossing diagram and genotypic analysis of the progeny of the dvl1-1D x nac1 (Salk 098992) cross shows unexpected segregation ratios in both the F2 and F3 generations. In the F2 generation, those plants which did not carry at least one copy of the dvl1-1D allele were killed by herbicide selection. The remaining progeny were genotyped by PCR. In the F3 generation from selected F2 parents, no herbicide selection was used. An encircled X indicates self pollination. All plants in the F1, F2, and F3 generations showed only a Col WT-like phenotype

Full size image

NAC1 Loss-of-Function Mutants Disrupt the dvlOE Phenotype

One of the genes in the transcription factor subgroup was selected for further study: NAC1 (At1g56010). NAC1 is a NAC-domain transcription factor that has been shown to play a role in auxin-mediated lateral root development along with its targeting microRNA, miRNA164 (Xie et al. 2000; Guo et al. 2005). QPCR confirmed the microarray data showing NAC1 expression is increased in dex-induced DVL4 plants. Available T-DNA insertion mutagenesis pools were searched, and three putative loss-of-function nac1 lines were found. Inserts were confirmed in the NAC1-coding sequence for two of these lines (salk_098992 and salk_135734) by PCR, and the insertion sites were determined by sequencing (Fig. S2).

Homozygous lines have been isolated for both alleles. No obvious developmental phenotypes were observed for either allele. To determine if NAC1 plays a role in the dvlOE phenotype, both of these alleles were crossed to the DVL1 overexpressing line, dvl1-1D (Fig. 3). This line was used due to its stable phenotype through multiple generations. F1 plants from these crosses do not show any dvlOE phenotypes, which is unexpected since dvl1-1D is a dominant gene (Fig. 4). In a F1 generation, the heterozygous dvl1-1D normally results in a typical dvlOE phenotype. In the segregating F2 populations of the cross with salk_098992 and dvl1-1D, there is a significant disruption in expected segregation ratios (chi-square <0.005) with no nac1 −/− /dvl1-1D +/+ plants being recovered (Fig. 3). However, in the F3 lines, a nac1 +/− /dvl1-1D +/+ parent does yield nac1 −/− /dvl1-1D +/+ plants at a slightly lower, but not significantly reduced ratio. These nac1 −/− /dvl1-1D +/+ plants appear indistinguishable from Col WT plants. Surprisingly, all the plants in both the F2 and F3 generations show a Col WT-like phenotype, including the NAC1 +/+ /dvl1-1D +/+ plants, which would be expected to show a dvlOE phenotype (Figs. 3 and 4). To check for silencing of dvl1-1D, transcript levels were examined. Unexpectedly, real time analysis of the F3 NAC1 +/+ /dvl1-1D +/+ flower buds showed an overaccumulation of DVL1 transcript, similar as that found in the original dvl1-1D line (>100,000-fold overaccumulation) although the dvlOE phenotype was not present. This suggests that NAC1 likely has an epigenetic effect on dvlOE; however, this effect does not seem to function by repression of DVL transcription, at least on a tissue scale. This epigenetic effect was not found in F1 plants of dvl1-1D crossed with homozygous salk_107861, a line predicted to contain an insert in another NAC transcription factor, NAC46 (At3g04060). This suggests some specificity in this epigenetic response, occurring only in both alleles of nac1.

Fig. 4
figure 4

nac1 LOF suppresses the dvl1-1D phenotypes. dvl1-1D x nac1 F1 plants do not show any dvlOE phenotypes including the protruding pistils and silique horns (A), dvl1-1D left, dvl1-1D x nac1 (Salk 098992 shown) F1 right. In addition, the rosette phenotype of smaller, rounder leaves is also suppressed in this cross (B), Col WT left, dvl1-1D center, dvl1-1D x nac1 (Salk 135734 shown) F1 right. The LOF of nac1 results in an epigenetic effect in which the dvl1-1D phenotype is suppressed in a F3 plant from the dvl1-1D x nac1 (Salk 098992) cross with the NAC1 +/+ dvl1-1D +/+ genotype in both the F2 and F3 generations (C) Col WT left, NAC1 +/+ dvl1-1D +/+ right. Scale bar 1 cm

Full size image

Since NAC1 was found to be overexpressed following dex-induced DVL4 expression, we hypothesized that a LOF allele of NAC1 would likely change the dvlOE phenotype to a more WT-like phenotype. Indeed, this is what was observed. Crosses with NAC1 LOF alleles resulted in a WT-like phenotype. However, these unusual segregation ratios, epigenetic effects and lack of observed dvlOE phenotypes, even when the plants had segregated to a NAC1 WT genotype, is an interesting and unexpected discovery. Further insights into the mechanism underlying this epigenetic effect may provide a better understanding of the role of DVL and NAC1 in plant development. Particularly intriguing is the fact that this effect does not appear to be functioning through a repression of transcription since DVL1 was still found to be overexpressed, although the dvlOE phenotype was absent. Since a repression of transcription of DVL1 does not seem to be responsible for this observed epigenetic effect, further studies could investigate other aspects. Perhaps translation of the DVL mRNA is perturbed resulting in little production of DVL protein despite high mRNA copy number. Additionally, there could be post-translational effects in which DVL protein could be generated but improperly processed and thus not active or not present at the proper location. Indirect effects through epigenetic changes on additional unknown players that may be downstream in the response to dvlOE could also result in the observed phenotype. Thus, several possible mechanisms through which this epigenetic effect may function can be envisioned.

Conclusions

Since the DVL gene family is well conserved across many seed plants (Narita et al. 2004; Wen et al. 2004), it is likely that this family plays a significant developmental role in general for plants. Therefore, a better understanding of this gene family will not only yield results applicable to the small proteins in plants but also will likely give us greater insights into plant development in general. Global expression analysis revealed that many transcription factors show perturbed expression levels in response to DVL4 overexpression. One of these transcription factors, NAC1, was investigated in further detail. We hypothesized that a LOF of nac1 would revert a dvlOE phenotype to a more WT-like phenotype. Indeed, this was shown using LOF nac1 lines. However, unusual epigenetic effects were observed suggesting that the dvl and NAC1 interaction is more complex than expected. Further investigation into this effect may yield insights into developmental and epigenetic processes during plant development.