Differential sensitivity of the Arabidopsis thaliana transcriptome and enhancers to the effects of genome doubling


Author for correspondence:
Luca Comai
Tel: +1 530 572 8485
Email: lcomai@ucdavis.edu


  • Two fundamental types of polyploids are known: allopolyploids, in which different parental chromosome sets were combined by ancestral hybridization and duplication; and autopolyploids, which derive from multiplication of the same chromosome set. In autopolyploids, changes to the nuclear environment are not as profound as in allopolyploids, and therefore the effects of genome doubling on gene regulation remain unclear.
  • To investigate the consequences of autopolyploidization per se, we performed a microarray analysis in three equivalent lineages of matched diploids and autotetraploids of Arabidopsis thaliana. Additionally, we compared the expression levels of GFP transgenes driven by endogenous enhancer elements (enhancer traps) in diploids and autotetraploid of 16 transgenic lines.
  • We expected that true ploidy-dependent changes should occur in independently derived autopolyploid lineages. By this criterion, our microarray analysis detected few changes associated with polyploidization, while the enhancer-trap analysis revealed altered GFP expression at multiple plant life stages for 25% of the lines tested. Genes on individual traps were coordinately regulated while endogenous gene expression was not affected except for one line.
  • The unique sensitivity of enhancer traps to ploidy, in contrast to the observed stability of genes, could derive from lower complexity of regulatory pathways acting on traps versus endogenous genes.


The effects of polyploidy on gene regulation are of interest in plant systems because of the widespread occurrence of polyploid species and populations and the mounting evidence for ancient polyploidization events in the ancestry of apparent diploids (Cui et al., 2006). Plants display higher tolerance to the consequences of genome doubling than vertebrates (Otto & Whitton, 2000; Doyle et al., 2008). Two fundamentally different types of polyploids exist: allopolyploids, which are derived from interspecies hybrids in which two or more parental genomes are maintained in duplicated forms; and autopolyploids, which are derived by multiplication of the same chromosome set (Comai et al., 2000; Comai, 2005). These two types of polyploids are different at the molecular and genetic levels. For example, in autopolyploids, recombination occurs between the multiple homologous chromosomes while in allopolyploids recombination is limited to the pair of homologous chromosomes derived from a single ancestral genome. Because diverged parental genomes are combined in allopolyploids, multiple regulatory changes are expected from the interaction of two disparate systems (Osborn et al., 2003), an expectation validated in multiple studies (Adams & Wendel, 2005; Chen, 2007; Doyle et al., 2008). Predictions are not as obvious for autopolyploidy, where changes to the nuclear environment are expected to be less profound than in allopolyploids (Comai, 2005). Accordingly, evidence for gene regulatory changes in autopolyploids is not clear. A study examining the expression of 18 maize (Zea mays) genes in haploid, diploid, triploid and tetraploid plants showed that for most genes the transcript level per cell was directly proportional to the structural gene dosage. A few genes, however, displayed negative or positive (greater than expected) correlation with ploidy (Guo et al., 1996). The molecular basis for this response is not yet understood. Interestingly, inbred maize autotetraploids display stronger inbreeding depression than diploids (Riddle et al., 2006). Recently, analysis of the transcriptome in an inbred maize ploidy series found frequent but low-level changes (Riddle et al., 2010). Comparison of diploid to triploid and tetraploid maize revealed that the most common trend was a decrease in expression, mirroring the decrease in size observed in maize autopolyploids. In Arabidopsis thaliana autotetraploids, however, no changes in vigor are visible. Depending perhaps on accession and on experimental conditions, phenotypic changes have been described as limited to trichome structure. (Mittelsten Scheid et al., 2003; Yu et al., 2009). Autotetraploids of multiple accessions displayed an increase in flower size and stomatal size, a likely reflection of increased nuclear content and cell size. To a casual observer, however, autotetraploids are nearly undistinguishable from the corresponding diploids (D. Pignatta et al., unpublished). Autopolyploids of multiple species display frequent aneuploidy and this may induce phenotypic alterations that should vary with the type of chromosome imbalance (Comai, 2005). Compelling evidence for tetraploidy-induced changes in epigenetic gene regulation exists for a transgenic locus of A. thaliana, which demonstrates striking ploidy responses, such as the ability to transfer its silencing state from one allele to the other (Mittelsten Scheid et al., 1996, 2003). This phenomenon is reminiscent of paramutation, where changes in chromatin states are mediated by an RNA signal (Chandler, 2007). We examined the possibility that another RNA-mediated phenomenon, transgene-induced post-transcriptional silencing, could also respond to ploidy (Pignatta et al., 2008). Comparing the efficiency of silencing of the endogenous gene CHALCONE SYNTHASE between tetraploids and their progenitor diploids, we found that a change in ploidy neither enhanced nor suppressed the silencing action of dsRNA-generating hairpin transgenes (Pignatta et al., 2008). Therefore, no evidence was found that the small RNA pathway responsible for post-transcriptional transgene silencing was affected by ploidy changes.

Whole-transcriptome expression analysis might discover cases of endogenous genes that respond to ploidy. In potato (Solanum phureja), 10% of the c. 9000 genes analyzed using microarray displayed changes in expression level within a twofold range, with no ‘ploidy-regulated’ gene identified (Stupar et al., 2007). In Saccharomyces cerevisiae (budding yeast), there is conflicting evidence about whether ploidy-regulated genes exist or not. Indeed, the first genome-wide transcriptome analysis in S. cerevisiae appeared to support the hypothesis of ploidy-regulated genes (Galitski et al., 1999). However, a more recent whole-genome transcriptome analysis of haploids and isogenic tetraploids resulted in very similar transcription profiles, although tetraploids failed to up- or down-regulate some genes important for stationary-phase survival (Andalis et al., 2004). The similarity of the gene expression profiles of diploid and tetraploid budding yeast was again confirmed in another study, and further supported by the steady-state protein levels for selected proteins (Storchova et al., 2006).

In A. thaliana, the effects of genome doubling on nonadditive gene expression remain elusive. As part of a study by Wang et al. on nonadditive gene regulation in allopolyploids (Wang et al., 2006), a transcriptome analysis of diploid vs autotetraploid A. thaliana identified only a small percentage of genes significantly differently expressed between the two. Importantly, the study focused on only one lineage of A. thaliana ecotype Landsberg erecta (Ler) (Wang et al., 2006).

At least two problems could complicate the analysis of regulatory changes during autopolyploidy. The first is the pervasive presence of aneuploids in autopolyploid populations (Bond et al., 2004; Henry et al., 2005, 2006). Depending on the species and background, polyploid populations can include 25–50% of aneuploid individuals. For example, in a tetraploid population the observed aneuploids may include pentasomics or trisomics for one to a few chromosomes, with hyperploidy being more common. These chromosomal variants could introduce regulatory variations that depend on dosage alterations rather than on ploidy (Henry et al., 2006). Also, the microarray analyses reported thus far may have lacked the statistical power to detect relatively small changes, because they were too subtle, masked by noise, or tissue-specific.

Here, we addressed this problem using a twofold approach. First, we carried out a microarray analysis in three presumably equivalent lineages of matched diploids and tetraploids of A. thaliana accession Columbia (Col-0). Independent tetraploid events derived from the same starting genotype should allow resolution of lineage-specific changes, such as those attributable to aneuploidy or to the colchicine treatment, from changes that derived from the diploid to tetraploid transition. Secondly, we used transgenic A. thaliana lines engineered with an enhancer gene trap T-DNA (GAL4-VP16- GFP; Fig. 1a). We reasoned that a tool that more efficiently displays changes in the spatial distribution of gene expression based on the sensitive green fluorescent protein (GFP) marker would be more likely to identify subtle changes or changes that may be more visible at specific developmental stages. The A. thaliana transgenic lines (accessions Col-0 and C24) used here carried the enhancer trap in different genomic locations. Expression of the chimeric transcriptional activator GAL4-VP16 gene was activated by an enhancer element present in the plant genome, which was different for each independent transgenic line. This strategy thus separated the effect of the structural part of the gene from that of the genomic positions. Our microarray analysis revealed few changes in gene expression between diploid and tetraploid plants. By contrast, the enhancer-trap analysis revealed that multiple loci in the genome are potentially subject to ploidy regulation. The inconsistency of the answers provided by the two methods is discussed.

Figure 1.

 Enhancer trap scheme and genotype of transgenic Arabidopsis thaliana lines. (a) T-DNA elements: the chimeric transcriptional activator GAL4-VP16; the neomycin phospho-transferase resistance II (NptII) gene; the 5× upstream activator sequence (UAS); the mGFP-ER reporter gene encoding the green fluorescent protein targeted to the endoplasmic reticulum. RB and LB, right and left borders, respectively; (p), the 35S CaMV minimal promoter; p, the nos promoter; E, endogenous A. thaliana enhancer in flanking genomic region. (b) GFP expression was compared between diploid individuals (2×), which carry two copies of the trap, and autotetraploid individuals (4×), which carry four copies. For simplicity, only one chromosome type is depicted. (c) Single-locus T-DNA enhancer trap lines of accession Columbia (Col-0) with linked genes. For each line, a schematic representation of the T-DNA insertion locus and the copy number are given. The length of linked genes (in black) and the distance from the T-DNA insertion locus (in green) are expressed in nucleotides (nt). The trap sequence is represented as a black rectangle. Arrows indicate the direction of the transcription.

Materials and Methods

Biological material for microarray analysis and RNA preparation

Three independent tetraploid lineages of Arabidopsis thaliana (L.) Heynh accession Col-0 were generated with colchicine treatment as follows. Seedlings were germinated on 0.5 × MS plates. At 10 d post germination they were flooded with colchicine solution (0.25–1%) for 15–60 min. The plants were rinsed in distilled H2O and grown to maturity. From each individual C1 plant, seeds were harvested and sown, and the ploidy of adult plants was assessed by flow cytometry using propidium iodide fluorescence following established protocols (Henry et al., 2005). The ploidy of each plant varied depending on the ploidy of the sector from which the gametes arose. Only sets of diploid and tetraploid plants that arose from the same colchicine-treated individual were considered a ‘matched pair’ and used in the subsequent analysis. Therefore, individuals in a matched pair are related by being derived from the same shoot apical meristem. As we have no information on the dispersion and transport of colchicine in different cells, we do not know that the members of a pair were exposed to the same amount of colchicine. Defining the treated plants as first generation, the third or further generation sibs from a single family were used for each line in the following experiment. Individual plants were grown in single 4 × 4 cm pots, with pots randomized within flats for each diploid/tetraploid set, in a growth chamber under 16 h of artificial daylight provided by fluorescent lights (TL80 bulbs; Philips, Eindhoven, the Netherlands) at 22°C (± 3°C). The above-ground portion of the plant was harvested whole at 4 wk after sowing. Plant order during harvest was randomized within each set so that no time differential existed between diploid and tetraploid materials. Tissue from 12 plants was pooled for each biological replicate. Frozen tissue was weighed and then ground to a powder. The powder was placed in a beaker and for each gram of tissue 4 ml of Trizol reagent (Invitrogen, Carlsbad, CA, USA) was added and the mixture thawed while stirring. The slurry was placed in centrifuge tubes and oriented horizontally on a nutating shaker for 15 min at room temperature. The solid material was pelleted by centrifugation for 10 min at 12,000 g in an SS-34 rotor. Supernatant was transferred to a fresh tube and 1 ml of chloroform was added for each 3 ml of solution, the mixture was vortexed for 15 s and let stand for 3 min. The organic and aqueous phases were separated by centrifugation for 15 min at 12,000 g in an SS-34 rotor. The aqueous supernatant was transferred to a fresh tube and the chloroform extraction was repeated. The aqueous phase was transferred to a fresh tube and RNA was pelleted with a minimal co-sedimentation of polysaccharides by adding 250 μl of 1.2 M NaCl in 800 mM sodium citrate, mixing, and then adding 250 μl of isopropanol for each millilitre of transferred supernatant. The mixture was incubated for 20 min and then spun for 20 min at 27,000 g in an SS-34 rotor. The RNA pellet was washed three times in 70% ethanol, dislodging the pellet from the wall of the centrifuge tube each time. The pellet was air-dried for several minutes and then re-suspended in 100 μl of TE buffer (10 mM Tris-Cl, pH 7.5, 1 mM EDTA).

Microarrays and statistical analysis

The A. thaliana whole-genome array-ready oligo set version 1.0 from Operon (Eurofins MWG Operon, Huntsville, Al, USA) was employed for these experiments, as previously described (Wang et al., 2006). The data were analyzed using a linear model. Two analysis of variance (ANOVA) models were utilized to identify differentially expressed genes between diploid and tetraploid individuals. The first model used a common variance assumption (i.e. all genes have the same variance):

image(Eqn 1)

where = 1, …, 8; = 1, 2; g = 1, …26 107, and r = 1, …, ng; μ is the grand mean, and A, D, T and G are the array, dye, treatment and gene effects, respectively. Moreover, AG, DG and TG are the interactions between array and gene, dye and gene, and treatment and gene, respectively. ɛijkgr are error terms which are independent random variable from a normal distribution with mean 0 and variance σ2. Differential expression was tested using hypotheses:

image(Eqn 2)

The second model used a per-gene variance assumption, and thus acknowledges each gene’s variation:

image(Eqn 3)

where μg, A, D and T are the average gene intensity, array, dye and treatment effects for gene g, respectively. ɛijkgr are error terms which are independent random variable from a normal distribution with mean 0 and variance σg2. Because each gene is tested for differential expression, the hypotheses are in terms of the treatment effects (i.e. T1g denotes tetraploid and T2g denotes diploid):

image(Eqn 4)

The multiple testing problem was addressed using Benjamini–Hochberg false discovery rate (FDR) and Holm sequential testing.

The adjusted Wald method at http://www.measuringusability.com/wald.htm was used to calculate a confidence interval for the occurrence of ploidy-dependent regulation of traps.

Enhancer trap lines characterization

Transgenic lines of A. thaliana accessions Col-0 and C24 were generated by S. Poethig and J. Haseloff and obtained from the ABRC (http://enhancertraps.bio.upenn.edu/default.html) (Fig. 1c). The T-DNA is an enhancer trap (Fig. 1a), with a GFP reporter gene driven by a minimal 35S CaMV promoter and under the regulation of a GAL4-VP16 gene, driven by the same promoter (http://www.plantsci.cam.ac.uk/Haseloff). We determined the complete nucleotide sequence of all the T-DNA elements (Supporting Information Fig. S1) with a series of PCRs using outward-facing primers designed with the Web-based Primer3 software (Table S1). PCR was followed by cloning into the pCR®2.1-TOPO vector (Invitrogen) and sequencing.

A series of crosses was performed to characterize single-locus insertion lines as described in Fig. S2. For each segregating line, 20 individuals from the backcross 1 (BC1) population were genotyped. DNA was extracted from rosette leaves as previously described (Comai et al., 2000) and used as a template in the PCR with primers targeting the GFP gene (Table S1). Lines were genetically and molecularly characterized for the presence of single-locus T-DNA insertion using inverse PCR and Thermal asymmetric interlaced PCR (TAIL-PCR) (Ochman et al., 1988; Liu & Whittier, 1995). A set of line-specific primer pairs was designed to confirm the insertion site and for use in multiplex PCR together with GFP primers to distinguish between homozygous and hemizygous individuals in the BC1 family (Table S1).

Southern blot analysis was used to determine the number of transgene copies inserted in a single locus for the lines segregating 3 : 1 as described previously (Pignatta et al., 2008). The probe corresponded to the mGFP5-ER gene (nt 192–739). The copy number for each line is reported in Fig. 1c.

GFP-homozygous diploid plants were tetraploidized as described previously (Josefsson et al., 2006). Briefly, several individuals of a line were treated with colchicine at the four to eight leaf stage and later allowed to self. Seeds from each treated plant were germinated and individual plants were examined phenotypically to identify prospective tetraploids (wild-type phenotype with larger flowers). The nuclear DNA content was confirmed by flow cytometry (Fig. S2). Tetraploid individuals were obtained from independent diploids of the same line.

Seeds derived from the crosses were surface-sterilized (20% bleach) and planted on an agar-solidified medium containing 1× MS salts, 1% sucrose and 50 mg ml−1 kanamycin sulfate (pH 6). Ten-day-old seedlings (c. 20) grown on vertical plates were directly scored for GFP expression with a Leica MZ16F manual fluorescence Stereomicroscope (Leica Microsystems Gmbh, Wetzlar, Germany) with filters designated GFP1 (excitation-emission; 395–455 nm) and GFP3 (450–490 nm) and a Leica 106 z lamp. Leaves, flowers and siliques were collected from at least five individuals, put onto plates with 0.8% agar solidified medium and scored for GFP expression. Images were acquired using a digital camera (Leica DFC 280) with the software Leica FireCam 1.9.1.

Trap- and trap-linked gene expression quantification

Total RNA extracted from flowers (three biological replicates) with the RNeasy Mini Kit (Qiagen Inc., Valencia, CA, USA) was DNAse-treated and used as a template in RT-qPCR as previously described (Pignatta et al., 2008). Flowers were chosen as the most convenient of the tissues that displayed consistent enhancer trap expression changes. 18S ribosomal RNA was used as a standard, consistent with the rationale described in a previous study on ploidy-induced changes (Guo et al., 1996). Expression of the GAL4-VP16, neomycin phospho-transferase resistance II (NptII) and GFP genes and the endogenous trap-linked genes was quantified using the primer pairs listed in Table S1.

DNA methylation analysis

DNA methylation analysis on the Gal4 and GFP genes was performed with McrBC digestion followed by PCR, as previously described (Lippman et al., 2003). Briefly, 1 μg of genomic DNA isolated from flowers of diploid and tetraploid plants was digested for 3 h at 37°C with McrBC, followed by heat-inactivation. Undigested DNA was used as a positive control. A 5-μl volume of sample was amplified by PCR with primers flanking the 35S-5 × upstream activator sequence (UAS) and complementary to GFP (Table S1). MULE transposon (AtMu1) primers were used to amplify a known methylated sequence and thus provide a control for McrBC activity (Lippman et al., 2003).


Polyploidization has subtle effects

To investigate the effects of genome doubling per se on gene expression, we created three independent tetraploid lineages of the A. thaliana accession Col-0 and matched each to diploid lineages derived from the same colchicine treatment. Comparison of diploid vs autotetraploid lineages should therefore be insensitive to stochastic colchicine effects. Reproducible changes that are caused by autotetraploidy should be found in each of the three independent autotetraploid lineages, while changes caused by unknown factors unrelated to ploidy should be unlikely to occur in all three. The data were analyzed using a linear model under two different scenarios, a common variance assumption and a ‘per gene’ variance assumption (see Materials and Methods section). With the ‘per gene’ variance the variation specific to a gene is used to determine whether such a gene deviates from the null hypothesis of no change. By using gene-specific data, the effect of expression level and gene type are taken into account rather than relying on averaging measurements for all genes. Accordingly, ‘per gene’ variance is expected to be a better tool than ‘general’ variance.

The individual analysis of each of the three lineage pairs yielded genes that were significantly differently expressed between matched diploids and autotetraploids (Fig. 2). Thus, from each individual analysis it could be concluded that at least some genes affected by ploidy had been discovered. However, comparing the three paired lineages we found no significant common change induced by autopolyploidy when using the stringent Holm’s procedure to control for multiple test errors (Fig. 2). Using the less stringent Benjamini and Hochberg false discovery rate (FDR) procedure, we found that the changes attributable to autopolyploidy were minor (Fig. 2). Using the ‘per gene’ variance assumption, only one gene, At2g32210, was significantly differentially expressed (up-regulated) in all three paired lineages (Fig. 2, Table 1). Under the more unrealistic scenario of common gene variance, seven genes were differentially expressed (Fig. 2, Table 1). Only two of the seven genes were differentially expressed in the same direction in the three paired lineages (Table 1). Pair-wise plots of expression changes illustrate the lack of correlation between lineage replicates, with correlation values ranging from 0.09 to 0.01 (Fig. S4). Furthermore, comparing the 88 differentially regulated genes reported by Wang et al. (2006) from analysis of diploid vs tetraploid Ler with the total of eight genes in the present study, we found none in common.

Figure 2.

 Venn diagrams summarizing the microarray statistical analysis for the three Arabidopsis thaliana lineages (36, red; 32, black; 17, blue). Data were analyzed using two linear models based on the per gene (top panels) and common variance assumptions (bottom panels). Both the Benjamini–Hochberg’s false discovery rate (FDR) and Holm’s approach were employed to control the significance level at 0.05 when testing the 26 107 genes. The intersection of these results summarizes the number of significantly differentially expressed genes in common between diploids and autotetraploids in the three A. thaliana lineages. Only one gene (At2g32210) was significantly differentially expressed in all the three lineages based on per gene variance analysis.

Table 1.   List of genes significantly differentially expressed in autotetraploids versus diploids common to the three Arabidopsis thaliana lineages (32, 17 and 36)
 4× vs 2×321736
  1. The gene list was derived using two linear models: per-gene and common variance assumptions. The Benjamini-Hochberg's false discovery rate (FDR) approach was employed to control the significance level at 0.05 when testing 26,107 genes. Lfc is log fold change, SD is standard deviation.

Per gene variance; FDR
Common variance; FDR

In summary, microarray analyses did not identify a reliable set of ploidy-regulated genes. This finding was consistent with two general possibilities. First, genome doubling may have no effects on gene regulation. Secondly, the effects may be subtle or hidden by noise, rendering the analytical tools unable to detect statistically significant changes.

Polyploidization can alter enhancer-activated GFP expression

A more sensitive measurement of individual genes might succeed in revealing effects of polyploidization that were not detected by the microarray analysis. Enhancer traps achieve expression of a transgenic reporter by fusion of a minimal promoter in the trap to an anonymous activating element residing in the host genome. We obtained transgenic A. thaliana lines engineered with an enhancer trap T-DNA encoding a sensitive GFP reporter (Haseloff, 1999) and characterized 16 independent transgenic lines: we confirmed or determined the insertion locus by PCR and sequenced the junction genomic region and determined the number of transgene copies at the single insertion locus by Southern blot analysis (Fig. 1). Tetraploids were generated by colchicine treatment of GFP-homozygous diploid seedlings for each line and the ploidy level was confirmed by flow cytometry analysis (data not shown).

We compared the expression of the GFP gene between diploids and tetraploids (Fig. 1b) at four developmental stages (seedlings, leaves, flowers and siliques) with a fluorescence stereomicroscope. For a majority of lines (12 of 16) we did not observe any difference in GFP expression following tetraploidization (Table 2). Four lines, all derived from Col-0, responded to polyploidization (Table 2, Fig. 3). These responses were not random, as in all cases tetraploids generated from different plants of the same GFP line displayed the same response. In lines cs70267 and cs70007, GFP was over-expressed in tetraploids compared with diploids. In particular, the increase in GFP fluorescence intensity was observed in all the developmental stages in the case of line cs70267 (Fig. 3a), while only in flowers and siliques for line cs70007 (Fig. 3b). Tetraploids of line cs70091 had a different pattern of expression in leaves compared with their respective diploids (not shown), while in seedlings, flowers and siliques the GFP fluorescence intensity was higher (Fig. 3c, and not shown). Tetraploids of line cs70016 were the only ones to display suppression of GFP expression as compared with their diploid counterparts. This pattern was observed in leaves (not shown), flowers and siliques (Fig. 3d).

Table 2.   Summary of microscopic observations of GFP fluorescence in autotetraploids and diploids at four developmental stages (seedling, leaf, flower and silique) in transgenic Arabidopsis thaliana lines from accessions Columbia (Col-0) and C24
 GFP expression in tetraploids compared with diploids
  1. Up, GFP expression was higher in autotetraploids; down, GFP expression was suppressed in autotetraploids; diff. pattern, GFP appeared to be equally expressed in terms of intensity but with a different pattern.

Accession Col-0
CS70091UpDiff. patternUpUp
Accession C24
Figure 3.

 Response of enhancer traps to genome doubling at multiple plant life stages. Images of GFP expression in diploids and autotetraploids were acquired using a digital camera fitted to a manual fluorescence stereomicroscope. Transgenic Arabidopsis thaliana lines: (a) cs70267, (b) cs70007, (c) cs70091 and (d) cs70016. For each line, two developmental stages are shown.

These results, which were robust and reproducible in independently generated tetraploids, indicated that genome doubling per se in A. thaliana can affect expression in multiple developmental stages of c. 25% of active enhancer traps.

Enhancer trap genes coordinate regulation and endogenous genes expression

Each enhancer trap consists of a complex transgenic locus inserted in a different genomic site. The T-DNA of the trap carries three genes (Fig. 1a): the regulator Gal4-VP16 is at the edge of the trap and its minimal promoter is available for fusion to an endogenous regulatory element. Activation of GAL4-VP16 leads to trans-activation of the GFP reporter fused to a chimeric promoter containing multiple Gal4-VP16 binding sites (5× UAS). A third gene, NptII, is driven by the relatively weak nos promoter. The regulation of three trap genes can thus be compared to that of flanking loci. After identifying sensitive traps by the changes in GFP fluorescence, we used quantitative RT-PCR to compare the expression of GFP, of its regulator Gal4-VP16, of NptII and for each trap of the respectively most closely linked plant gene (Fig. 4) in diploids vs tetraploids. Regulation of GFP is generally assumed to reflect regulation of GAL4-VP16 transcription. It is possible, however, that the interaction of the GAL4-VP16 trans-activator with the GFP gene upstream activating sequence (5× UAS) may itself be sensitive to ploidy. Several lines of evidence indicate that this is not the case: first, three-quarters of the lines studied did not display changes; secondly, one of the four ploidy-sensitive line displayed suppression while the remaining three displayed activation of GFP; thirdly, the changes measured in GAL4-VP16 mRNA levels were consistent with the changes observed in GFP mRNA (Fig. 4). As a control, we quantified the expression of GAL4-VP16 and the NptII gene in flowers of diploids and tetraploids for three of the 12 lines (cs70074, cs70078 and cs70116) that did not exhibit fluorescence changes upon polyploidization. Consistent with microscope observations, qPCR results indicated no difference in expression (not shown). These results indicate that changes in GFP expression reflect the regulatory action of neighboring plant enhancers on Gal4-VP16 transcription.

Figure 4.

 Enhancer trap and trap-linked gene expression quantification in diploids (solid columns) vs autotetraploids (open columns). RNA was extracted from flowers of transgenic and wild-type Arabidopsis thaliana individuals, respectively. Three biological and three technical replicates were tested in qPCR. Data were normalized to 18s RNA expression and are presented as fold change. Bars represent overall expression levels. Error bars indicate standard deviation.

In two of the ploidy-sensitive trap lines (cs70267 and cs70016; Fig. 4), the NptII mRNA level was also affected, indicating that the whole transgenic domain was influenced by a change in ploidy. Line cs70016 displayed 10–15-fold suppression of GFP and Gal4-VP16 expression and even stronger suppression of the NptII gene (c. 60-fold; Fig. 4). The presence in the enhancer trap of the 5× UAS repeated element may trigger a silencing response that methylates short tandem repeats of DNA (Martienssen, 2003; Kinoshita et al., 2006). To investigate the role of DNA cytosine methylation on this suppression, we tested McrBC susceptibility of the promoters and activator sequences regulating Gal4-VP16 and GFP. We found no difference between the diploids and tetraploids (Fig. S3) and concluded that cytosine methylation was not involved in down-regulation of this enhancer trap.

Trap activity could reflect regulation of neighboring genes by polyploidy. To evaluate this possibility, we quantified the expression of the trap-linked genes (Fig. 1c) in diploid vs tetraploid A. thaliana Col-0 but found no difference dependent on ploidy level (Fig. 4). We further compared the expression of the trap-linked genes in cs70267, cs70007 and cs70016 to determine whether it was affected by the presence of the enhancer trap. The genes in the first two lines, in which the trap was induced by tetraploidy, displayed no changes and were therefore consistent with the expression pattern of the wild type. In the trap line that displayed suppressed GFP expression in tetraploids (CS70016), the trap is inserted in the 5′ transcribed region of the transposable SADHU2.1 element (Fig. 1c). While SADHU2.1 is expressed in wild-type flowers regardless of ploidy, its expression was suppressed in tetraploids of the trap-tagged line in a manner similar to that of the GFP reporter (data not shown). Therefore, suppression of this flanking genomic region is ploidy dependent but requires the presence of the trap. This case of tetraploid-dependent silencing is reminiscent of that reported by Mittelsten Scheid et al. (2003), although we found no evidence of differential methylation in the tested regions of the trap (Fig. S3).


Changes between diploids and tetraploids of the same species have been reported previously (Galitski et al., 1999; Andalis et al., 2004; Storchova et al., 2006; Wang et al., 2006; Stupar et al., 2007). What is not clear is whether these are true responses to genome doubling, because independent replication of the polyploidization process was not employed. True responses to genome doubling should be reproducible. We applied this criterion by analyzing the transcriptome in three independent lineages of autotetraploid A. thaliana Col-0 and their matched diploids. We also tested the response of transcriptional enhancers by examining the regulation of GFP enhancer traps upon genome doubling. Our results are seemingly paradoxical: microarray analysis of ploidy-induced changes failed to provide a set of robustly regulated genes, yet ploidy-induced changes could be easily demonstrated by the use of enhancer traps. If enhancer traps reproduce the regulation of ploidy-sensitive endogenous genes, why did we not find them in the microarray study? One possibility is that the microarray assay may be biased or not sufficiently sensitive. Inconsistent with this possibility, this is the same microarray system that was used to detect common changes in independently produced synthetic allopolyploids of A. thaliana and Arabidopsis arenosa (Wang et al., 2006). Another possibility is that ploidy-regulated genes are expressed at very low level, or in specific tissues, and thus escape detection (i.e. are in the range of the experimental noise). This is unlikely as GFP activation and suppression were found in multiple tissues of the four responding trap lines and the whole aerial part of 4-wk-old plants was sampled for the microarray experiment (see Materials and Methods section). Moreover, differences in NptII, GFP and GAL4-VP16 were readily found by RT-PCR, which utilizes whole tissue preparations of RNA, but none of the eight endogenous genes compared between diploid and tetraploid wild types exhibited expression changes in response to ploidy. If one assumes that each regulated enhancer trap represents a gene, the total number of ploidy-regulated genes would lie between 2800 and 17 000. This 1% confidence interval (CI) was derived from the observation that four of 16 enhancer traps tested responded positively or negatively to autopolyploidy (CI 0.07 to 0.57 with α = 0.01, or 2800 to 17 000 genes out of c. 30 000). If traps reproduce perfectly the regulation of genes, thousands of genes would be expected to be ploidy sensitive. Moreover, the observed changes in expression levels are of sufficient magnitude (×4 to over ×600) that at least some of the thousands of these ploidy-sensitive genes should be identified in a microarray analysis. We believe, therefore, that our microarray analysis was sufficiently sensitive, but, consistently with previous investigations, few endogenous genes displayed ploidy-dependent regulation.

Our analysis is inconsistent with the induction of widespread, reproducible and large changes in the expression of genes by a change in ploidy. Nonetheless, the frequent and reproducible response of enhancer traps to ploidy changes suggests that autopolyploidy induces changes in the regulatory environment of the cell. These diverging observations would be reconciled if enhancer traps exhibit unique sensitivity to ploidy while endogenous genes are buffered against it. Hypersensitivity of enhancer traps to ploidy may result from reduced complexity of their regulation. Enhancer trap activation requires the capture of positive regulatory signals from genomic sites. Several cases where traps were analyzed in depth (Springer, 2000; Fridborg et al., 2004; Baroux et al., 2005; Leon-Kloosterziel et al., 2005; Slotkin et al., 2009) indicate that their regulation is lower in developmental and probably biochemical complexity than the complete cis-regulatory complement acting on genes, which are subject to complex regulation involving multiple pathways. If multiple positive and negative signals converge on the promoter of a gene, the other pathways may compensate changes in one pathway.

Alternatively, establishment of an altered regulatory environment in response to ploidy changes may affect many genes, but with low penetrance. One explanation of our observations could be that upon polyploidization a large set of genes are potentially affected, but only a smaller and random number become altered, perhaps with unpredictable direction. Indeed, this may explain why only two of the seven genes that were detected to be differentially expressed under the common variance assumption (Fig. 2, Table 1) were regulated in the same direction.

The dependence of enhancer traps on the formation of a T-DNA transgenic locus may also affect another layer of gene regulation: the formation of insulated domains (Ong & Corces, 2009). We observed that in two of four cases the regulation of the NptII gene varied in harmony with GFP. It is possible that the T-DNA disrupts existing domains, resulting in poor insulation and increased susceptibility to perturbation, perhaps because of the spreading of neighboring chromatin states. ‘Position effects’ have been previously described (Singh et al., 2008) and different positions in the genome may be differently sensitive to a change in ploidy. Consistent with this possibility, two different transgenes driven by the same 35S-CaMV promoter, but located in different regions of chromosome 5, reacted differently upon triplication of the chromosome (Huettel et al., 2008). One transgene was strongly down-regulated in the Chr5 trisomics as compared with diploids, while the average expression of the other transgene remained approximately the same. The suppressed transgene was inserted near a cluster of silent transposon-related sequences and tRNA genes, giving rise to numerous small RNAs. The enhancer trap that displays ploidy-dependent suppression, CS70016 (Fig. 4), is inserted inside the putative transposon SADHU2.1 (Fig. 1c), which is in turn located in the 3′ nonterminal inverted repeats of a Mutator-like element (non-TIR MULE) named KAONASHI (Hoen et al., 2006; Rangwala et al., 2006). Interestingly, SADHU2.1 becomes sensitive to ploidy in the enhancer trap line, providing additional evidence for domain-wide regulatory changes. The negative regulation of this enhancer trap line by ploidy may underlie an interaction with suppressive heterochromatin. However, in lines cs70267, cs70091 and cs70007 the trap was inserted into protein-coding genes, which are moderately or highly expressed (Schmid et al., 2005). Thus, the formation of poorly insulated domains could result in susceptibility to multiple perturbations including ploidy.

It is possible that ploidy-regulated changes may exhibit a genome-specific threshold of sensitivity, or that they may be sensitive to the cause of polyploidization. Maize, for example, may be sensitive to autotetraploidy (Riddle et al., 2006, 2010) because of its genomic characteristics – perhaps as a result of the widespread interspersion of genes and transposons. The A. thaliana genome, in contrast, may be more resilient when tetraploid, but be subject to regulatory problems at higher ploidy with connected phenotypic depression of growth. By colchicine treatment of diploids we have produced hundreds of polyploids of A. thaliana (B. P. Dilkes et al., unpublished). Using flow cytometry, in all cases we found tetraploids. The absence of hexaploids and octaploids may result from impaired proliferation of hexaploid and octaploid cells in the meristem of colchicine-treated plants or from failure of gametes with ploidy higher than 2N. Indeed, octaploid A. thaliana are unstable (Yu et al., 2009) and severely compromised in growth (Tsukaya, 2008) and therefore A. thaliana resembles maize in its response to ploidy. Perhaps a higher ploidy threshold is required to trigger deleterious ploidy-related effects in A. thaliana.

Autopolyploidization may occur through different mechanisms. This study used a spindle poison, colchicine. The maize study (Riddle et al., 2006) employed tetraploids produced by nitrous oxide treatment of diploids, which also interferes with spindle function. In nature, tetraploids can arise from spontaneously tetraploidized sectors or from rare 2n gametes and in this latter case pass through a triploid bridge, which may entail an aneuploid stage (Henry et al., 2005). Therefore, different stresses may be imposed on cells and result in different regulatory outcomes.

In summary, using microarray analysis it was not possible to find regulatory changes that could be clearly attributable to changes in ploidy. This was in contrast to the readily demonstrable, large and frequent changes we observed in an analysis of enhancer trap expression. The response of enhancer traps indicates that regulatory pathways acting on transcription can be affected but no effects of ploidy were seen on neighboring endogenous genes in comparisons of diploid and tetraploid wild-type plants. While it is possible that the difference in the results obtained with these two methods might lie in the insufficient sensitivity of the microarray system compared with the exquisite sensitivity of the GFP reporter, we argue that a better explanation is higher complexity of regulatory pathways affecting genes compared with those acting on enhancer traps, or better insulation of chromosome domains from perturbing effects compared with enhancer traps. Buffered regulation of genes with respect to genomic ploidy may have selective advantages and might help explain the pervasive occurrence of polyploidy in the plant kingdom.


This research was supported by NSF Plant Genome Program award DBI 0733857, Functional genomics of plant polyploids, to Z.J.C., L.C., R.W.D., and A.M. We are grateful for the many useful discussions at the biennial meetings of the NSF polyploid project, and to Drs R. Gaeta, Z. Ni, and J. C. Pires, L. Tian, and J. Wang for help with the printing and hybridization of the microarrays. We thank Brian Watson and Steven H. Reynolds for technical help.