Plant Molecular Biology

, Volume 74, Issue 3, pp 293–305

Both the constitutive Cauliflower Mosaic Virus 35S and tissue-specific AGAMOUS enhancers activate transcription autonomously in Arabidopsis thaliana


DOI: 10.1007/s11103-010-9673-9

Cite this article as:
Singer, S.D., Cox, K.D. & Liu, Z. Plant Mol Biol (2010) 74: 293. doi:10.1007/s11103-010-9673-9


The expression of eukaryotic genes from their cognate promoters is often regulated by the action of transcriptional enhancer elements that function in an orientation-independent manner either locally or at a distance within a genome. This interactive nature often provokes unexpected interference within transgenes in plants, as reflected by misexpression of the introduced gene and undesired phenotypes in transgenic lines. To gain a better understanding of the mechanism underlying enhancer/promoter interactions in a plant system, we analyzed the activation of a β-glucuronidase (GUS) reporter gene by enhancers contained within the AGAMOUS second intron (AGI) and the Cauliflower Mosaic Virus (CaMV) 35S promoter, respectively, in the presence and absence of a target promoter. Our results indicate that both the AGI and 35S enhancers, which differ significantly in their species of origin and in the pattern of expression that they induce, have the capacity to activate the expression of a nearby gene through the promoter-independent initiation of autonomous transcriptional events. Furthermore, we provide evidence that the 35S enhancer utilizes a mechanism resembling animal- and yeast-derived scanning or facilitated tracking models of long-distance enhancer action in its activation of a remote target promoter.


Enhancer 35S promoter AGAMOUS second intron Transcription initiation Long-distance activation of gene expression Arabidopsis thaliana 


The majority of eukaryotic genes exhibit unique levels and spatiotemporal patterns of expression through the combinatorial action of multiple regulatory elements that exert their control either locally or over long genomic distances. One such key regulatory element is the transcriptional enhancer, which is classically defined as a cis-acting DNA sequence that stimulates transcription of RNA polymerase II-transcribed genes independently of their position and orientation with respect to the transcriptional initiation site (Banerji et al. 1981). Within a genomic context, there is very little correlation between the location of these stimulatory elements and that of their target gene. For example, the wing margin enhancer of the Drosophila cut locus resides 85 kb upstream of its cognate promoter (Jack et al. 1991), whereas the Drosophila β3-tubulin and Arabidopsis AGAMOUS enhancers lie within introns of their respective transcriptional units (Busch et al. 1999; Deyholos and Sieburth 2000; Hinz et al. 1992). Conversely, enhancers for the rat and human myosin light chain 1/3 genes are located downstream of their polyadenylation signals (Donoghue et al. 1988; Rosenthal et al. 1990) and in certain instances, chromosome pairing even allows an enhancer to activate transcription from an allelic promoter on a separate chromosome (Morris et al. 1998). Furthermore, this activation of transcription is not restricted to constitutive induction; indeed, a large number of enhancers direct very precise tissue-specific expression of their target genes (for example Emerson et al. 1987; Gillies et al. 1983; Jacquemin et al. 1994; Sieburth and Meyerowitz 1997).

The current view of enhancer action supports the premise that these elements function to increase the probability that a gene will be active in a given cell (Fiering et al. 2000; Sutherland et al. 1997; Walters et al. 1995) through the recruitment and delivery of numerous factors involved in transcription initiation to cis-linked promoters via protein–protein interactions (Keaveney and Struhl 1998; Ptashne and Gann 1997). However, only a few metazoan enhancers and yeast upstream activating sequences (UASs), which are the functional equivalents of enhancer elements, have been studied in any detail to date. Therefore, we lack a general picture of transcriptional regulation by the majority of these elements and many questions remain unanswered regarding the mechanism by which enhancers promote transcription.

In an attempt to explain the means of gene control by distant enhancers, a number of different models have been proposed (Blackwood and Kadonaga 1998; Bulger and Groudine 1999; Dorsett 1999), among which the looping, scanning and facilitated tracking models are principal. The looping model stipulates that transcription initiation is promoted by the direct interaction of an enhancer with promoter elements via DNA bound proteins, with looping out of intervening chromatin (Rippe et al. 1995; Shang et al. 2002, Spicuglia et al. 2002; Tolhuis et al. 2002). The scanning model suggests that enhancers recruit and bind RNA polymerase II and/or associated transcription factors, which then migrate in either direction along the DNA until they encounter a competent promoter and instigate the formation of a transcription initiation complex (Louie et al. 2003; Travers 1999; Tuan et al. 1992). This scanning by RNA polymerase II often results in the generation of RNA transcripts derived from the intervening DNA (Dobi and Winston 2007; Ho et al. 2006; Masternak et al. 2003; Rogan et al. 2004; Tchurikov et al. 2009; Zhu et al. 2007). The facilitated tracking model incorporates aspects of both the looping and scanning models in that both the enhancer and its anchored factors progress along the chromatin fiber until a stable interaction is formed with promoter-bound proteins, generating a loop of intervening DNA that enlarges progressively during the tracking process (Blackwood and Kadonaga 1998; Hatzis and Talianidis 2002). Although there appears to be evidence favoring each of these models, it has been difficult to provide clear-cut substantiation in support of any single one of them and current data suggests that one model may not apply to all enhancers.

While various mechanisms are in place within eukaryotic genomes to prevent inappropriate enhancer-mediated activation of nearby promoters (reviewed in Kadauke and Blobel 2009), this is not the case in transgenic constructs. In plants, transgenic technology often requires the use of tissue-, organ- or developmental stage-specific and strong constitutive promoters to drive transgene expression exclusively in targeted tissues, and transformation vectors with the ability to harbor multiple gene units to enhance the performance of diverse agronomic traits are becoming the norm. Unfortunately, enhancer–promoter crosstalk and the subsequent alteration of promoter specificity and/or strength can cause serious consequences for precisely engineering gene function in transgenic plants. For example, the co-existence of the CaMV35S enhancer with the root-specific LRP1 promoter (Smith and Fedoroff 1995), vascular tissue-specific AAP2 promoter (Hirner et al. 1998), carpel-specific AGL5 promoter (Savidge et al. 1995), or tapetum-specific TA29 promoter in the same transgene can activate these promoters ubiquitously in all plant tissues (Jagannath et al. 2001; Zheng et al. 2007). To date, very few approaches have been developed that prevent these interactions (Hily et al. 2009; Singer et al. 2010), which is likely due, at least in part, to our lack of understanding of the mechanism underlying enhancer–promoter communication in plant systems.

To gain insight concerning the mechanism by which enhancers operate in plants, we endeavored to characterize the activation function of both the constitutive CaMV 35S enhancer and highly tissue-specific enhancer derived from the AGAMOUS second intron (AGI) in Arabidopsis thaliana. Both of these enhancers are used regularly to drive the expression of a downstream reporter gene when fused to a minimal promoter (for example Busch et al. 1999; Deyholos and Sieburth 2000; Kay et al. 1987; Liu and Liu 2008); however, the manner in which they operate is as of yet unknown. Since transcriptional activation by enhancer elements is typically described as being linked to the presence of a cognate promoter, we investigated the ability of the 35S and AGI enhancers to activate transcription of the β-glucuronidase (GUS) coding sequence both in the presence and absence of a target promoter. Unexpectedly, both enhancers were able to drive the expression of a nearby GUS transgene in the correct spatiotemporal pattern regardless of the presence of a cis-linked promoter. Analyses of the enhancer-activated GUS transcripts indicated that in the absence of a target promoter, these two superficially disparate enhancer elements functioned in a very similar manner by initiating transcription autonomously. Furthermore, our results provide evidence for a transcription-based model of long-distance enhancer–promoter communication within an artificial system in plants; information that may be of potential value when designing transgenic constructs which minimize enhancer–promoter interactions in the future.

Materials and methods

Plasmid constructs and plant transformation

Representations of the transformation vectors utilized in this study can be found in Fig. 1. All plasmids were produced using standard protocols and are present in a pBIN19 (Bevan 1984) background. The 3,048 bp AGI (flanked by 84 bp of upstream MADS-box-derived exonic sequence and 11 bp of downstream intervening region-derived exonic sequence) was isolated from Arabidopsis as described previously (Liu and Liu 2008) and inserted immediately upstream of the GUS coding region and nopaline synthase transcriptional terminator (nos-t) to generate AGI::GUS. The promoterless GUS plasmid (::GUS) was produced by removing the AGI fragment from the AGI::GUS vector via HindIII digestion and re-ligation. The AGIP::GUS vector, which includes the AGI sequence fused to a minimal promoter (P), was generated as described by Liu and Liu (2008). Briefly, the AGI was fused to the minimal CaMV35S promoter (−60 position), yielding AGIP, which was inserted 5′ of the GUS coding region and nos-t. The production of 35S/AGI::GUS, which includes a partially duplicated 35S promoter (bearing two copies of the enhancer element to ensure strong expression of a downstream transgene; Kay et al. 1987) fused to the enhanced green fluorescent protein (eGFP) coding sequence and nos-t along with an AGI::GUS::procyclin-associated gene 7 (pAG7) transcriptional terminator cassette in a head-to-head orientation, has been described in detail elsewhere (previously termed JM69; Hily et al. 2009). Plasmid 35S/AGIP::GUS was generated by replacing the AGI::GUS::pAG7 cassette of the 35S/AGI::GUS vector with the AGIP::GUS::nos-t cassette from the AGIP::GUS plasmid.
Fig. 1

Schematic diagram of constructs utilized in this study. All constructs shown are drawn to scale and are present in a pBIN19 background. Vector ::GUS is a negative control vector containing a promoterless/enhancerless GUS reporter gene (a). Constructs AGI::GUS and AGIP::GUS include the GUS coding region fused to an upstream AGI, with the minimal 35S promoter present between the two in the latter construct (b). Plasmids engineered to study 35S enhancer function are shown in (c). The 35S/AGI::GUS vector contains the 35S::eGFP and AGI::GUS cassettes in a head-to-head orientation. In contrast, the 35S/AGIP::GUS construct bears the minimal 35S promoter between AGI and GUS sequences. Triangles indicate directionality of the various regulatory elements. Black lines represent background vector sequence. 35S: partially duplicated 35S promoter of the Cauliflower Mosaic Virus; AGI: AGAMOUS second intron; P: 35S minimal promoter; M: 84 bp of AGAMOUS exon 2 (including a 3′ portion of the MADS-box); I: 11 bp of AGAMOUS exon 3 (including a 5′ portion of the I-box); GUS: β-glucuronidase coding sequence; T: transcriptional terminator; eGFP: enhanced green fluorescent protein coding sequence

Each binary vector was introduced into Agrobacterium tumefaciens strain GV3101 via electroporation and the resulting recombinant bacteria were used to transform Arabidopsis thaliana ecotype C10 using the floral dip method (Clough and Bent 1998), which has been shown previously to result in the introduction of a single T-DNA insert in over 50% of transgenic lines (Alonso et al. 2003; Rosso et al. 2003). All transgenic plants utilized in this study were phenotypically normal primary transformants and were confirmed by PCR.

Histochemical staining for GUS activity

Histochemical staining for GUS activity was carried out basically as described by Jefferson et al. (1987). Leaf and floral tissues from a selection of transgenic lines bearing each construct were incubated in 1 mM 5-bromo-4-chloro-3-indolyl-β-d-glucuronide (X-gluc; in 100 mM phosphate buffer, pH 7.0, 10 mM EDTA, 0.5 mM potassium ferrocyanide and 0.1% Triton X-100) at 37°C for 72 h. Stained tissue was subsequently depigmented in a series of 70% ethanol washes.

Detection of GFP fluorescence

A Typhoon Trio fluorescence scanner with the control v5.0 software (Amersham Bioscience, Piscataway, NJ) was used to scan floral tissue for visualization of fluorescence emitted from either GFP or chlorophyll as described by Hily and Liu (2009). The resulting images were overlaid for visualization of both GFP and chlorophyll fluorescence. Acquired images were extracted using the Image Quant TL v2005 software.

RT-PCR analysis of GUS and GFP expression

Total RNA was isolated from flower and leaf tissues using the RNeasy Plant Mini kit (Qiagen, Valencia, CA) and contaminating DNA was removed using the TURBO DNA-free system (Ambion, Austin, TX). RT-PCR analyses were performed using the Superscript III first-strand cDNA synthesis kit according to the manufacturer’s instructions (Invitrogen, San Diego, CA). In each case, 1 μg of RNA was reverse transcribed with an oligo(dT) primer and 2 μl of the resulting cDNA mixture were utilized as template in PCR assays using HotStart GoTaq polymerase (Promega, Madison, WI) in a final volume of 50 μl. To determine whether GUS transcripts were present in leaf and floral tissues from the various lines analyzed, primers GUSRTF2 and GUSRTR2 were used to amplify a 183 bp product, while EF1αF and EF1αR (Hily and Liu 2009) were utilized to amplify a 630 bp EF1α-specific product as an internal control. In the case of those lines that also contained the 35S::GFP cassette (35S/AGI::GUS and 35S/AGIP::GUS), primers eGFPF1 and eGFPR2 were additionally used to amplify a 245 bp eGFP-specific product. The conditions for PCR amplification of GUS-specific cDNA were 95°C for 2 min, followed by 28 cycles of 95°C for 30 s, 60°C for 30 s, and 72°C for 30 s, with a final elongation step of 72°C for 5 min. The same general parameters were used for amplification of the EF1α- and eGFP-specific fragments except that an extension time of 1 min and 22 cycles were utilized in the case of EF1α, and an annealing temperature of 55°C, extension time of 1 min, and 30 cycles were utilized in the case of eGFP. PCR products were resolved on 1–2% agarose gels and visualized with ethidium bromide. Primer sequences and annealing sites can be found in Table 1.
Table 1

Primer sequences

Primer name

Sequence (5′ to 3′)

Annealing sitea











































aPrimer annealing sites indicate the 5′ annealing site with respect to the start codon of each respective gene except in the case of those denoted by an asterisk, in which case the site specifies the 5′ annealing site of the primer with respect to the 5’ splice site of the AGI

5′ rapid amplification of cDNA ends

The 5′ RACE System for Rapid Amplification of cDNA Ends (Invitrogen) was utilized to map the 5′ ends of GUS and eGFP transcripts from two independent lines bearing each construct. For GUS transcripts, first-strand cDNA synthesis reactions were carried out on 2–5 μg DNase-treated total RNA (prepared as described above) with primer GUSRTR1. The resulting cDNAs were C-tailed and 5 μl purified reaction were utilized in subsequent PCRs using HotStart GoTaq and primers GUSR6 and AAP (Invitrogen), which anneals to the 5′ polyC tail. In the case of AGI::GUS and AGIP::GUS lines, which expressed GUS only in stamens and carpels, nested PCRs were carried out to improve the specificity and increase the amount of product generated using 1 μl PCR product as template and primers GUSR2 and AUAP (Invitrogen). 5′ RACE was also employed to determine the transcriptional start site of the eGFP transcript driven by the 35S promoter in 35S/AGI::GUS and 35S/AGIP::GUS lines. First-strand cDNA synthesis was conducted using primer eGFPR2 and PCR amplification was carried out with primers eGFPR3 and AAP. In each case, PCR products from unadulterated PCR reactions, as well as gel-purified products in the case of the AGI::GUS lines, were cloned into pGEM-T easy (Promega) and a number of separate clones were sequenced in each instance.

Northern blot analysis

Total RNA was isolated from floral tissue using TRI Reagent according to the manufacturer’s instructions (Ambion). Twenty μg of 35S/AGI::GUS, 35S/AGIP::GUS and ::GUS RNA were resolved through a 1% agarose gel, blotted, hybridized and washed using the NorthernMax-Gly kit (Ambion). Probe template was generated using PCR with primers GUSprobeF2 and GUSprobeR1 to generate an 800 bp GUS-specific fragment (anneals to nucleotides +558 to +1,357 of the GUS transcript, where +1 is the ATG) and AGI-IIF1 and AGI-IIR12 to generate a 1,712 bp product specific to the 3′ half of the AGI (anneals to +1,274 to +2,985, where +1 is the first nucleotide of the AGI), as well as TUB4F2 and TUB4R2 to generate an 801 bp β-tubulin4-specific product (+196 to +996) for use as an internal control. PCR products were gel-purified and utilized with the random primers labeling system (Invitrogen) to generate 32P-labeled probes. Hybridizations were carried out for 2 h to overnight at 42°C and bands were visualized with a Typhoon PhosphorImager.

Bioinformatic analyses

Sequences were scanned for putative promoter regions and transcription initiation sites using the Berkeley Drosophila Genome Project’s Neuronal Network Promoter Prediction program ( Possible TATA-boxes were identified using the PlantCARE database (; Lescot et al. 2002) and in silico translation of nucleotide sequences was carried out using the Translate tool (


The AGI enhancer drives stamen- and carpel-specific GUS expression in the absence of a linked minimal promoter

Previous studies have shown that the AGI enhancer fused to the 60 bp minimal 35S promoter, which we have termed here the AGIP promoter, specifies carpel- and stamen-specific expression (Deyholos and Sieburth 2000; Liu and Liu 2008; Sieburth and Meyerowitz 1997). To test whether the AGI enhancer alone can drive a similar pattern of GUS expression, we fused the 3,048 bp AGI with very small regions of associated flanking exonic sequence directly to the GUS coding region to create an AGI::GUS fusion (Fig. 1b). This construct, along with a negative control enhancerless/promoterless GUS vector and the positive control AGIP::GUS vector (Fig. 1a, b) were introduced into Arabidopsis, respectively. Histochemical analyses of GUS staining indicated that 80.0% of AGIP::GUS lines (n = 20) exhibited GUS activity specifically in carpels and stamens with no observable staining in vegetative tissues, which is consistent with findings from earlier studies (Liu and Liu 2008), while none of the enhancerless/promoterless ::GUS lines (n = 11) displayed GUS activity in any tissue type. Interestingly, 47.1% (n = 17) of AGI::GUS lines exhibited floral organ-specific GUS staining that was similar in pattern and strength to that observed in AGIP::GUS plants (Fig. 2a). In fact, the only difference noted between AGIP::GUS and AGI::GUS lines was the increased efficiency with which the AGIP was able to drive GUS expression in a transgenic population compared to the AGI alone. The ability of the AGI enhancer to activate flower-specific GUS expression was further confirmed by semi-quantitative RT-PCR analyses (Fig. 2b), which indicated that GUS transcripts were present in flower, but not leaf, tissues from both AGIP::GUS and AGI::GUS lines.
Fig. 2

Analyses of reporter gene expression in various transgenic lines. Histochemical assays of floral and leaf tissues from transgenic lines demonstrated that while no background GUS expression was present in negative control ::GUS lines, both the AGI alone and AGI fused to the 35S minimal promoter were able to drive downstream reporter gene expression in a stamen- and carpel-specific pattern in AGI::GUS and AGIP::GUS lines, respectively (a). Conversely, when the 35S promoter was situated upstream and in the reverse orientation of the AGI::GUS and AGIP::GUS cassettes (as in 35S/AGI::GUS and 35S/AGIP::GUS constructs), the flower-specificity of AGI-driven reporter gene expression was superseded by the ubiquitous 35S enhancer, resulting in GUS expression in every tissue analyzed (c). In addition to driving upstream GUS expression in 35S/AGI::GUS and 35S/AGIP::GUS lines, the 35S promoter also enabled the constitutive expression of its downstream reporter gene, GFP (d). RT-PCR analyses with GUS- and GFP-specific primers confirmed the flower-specific or ubiquitous expression patterns in the various lines (b, e). EF1α was utilized as an internal standard. Representative lines are displayed in each case. AGI: AGI::GUS line; AGIP: AGIP::GUS line; 35S/AGI: 35S/AGI::GUS lines; 35S/AGIP: 35S/AGIP::GUS lines; GUS: β-glucuronidase; EF1α: elongation factor 1 α; GFP: green fluorescent protein

Detection of multiple GUS transcription initiation sites in AGI::GUS lines

To better understand the mechanism underlying AGI enhancer-mediated flower-specific activation of GUS expression in the absence of a target promoter, we performed 5′ RACE analyses to characterize the transcription initiation sites of GUS transcripts in AGI::GUS and AGIP::GUS lines (two independent lines of each) shown by histochemical analyses to express GUS exclusively in stamens and carpels. Interestingly, in one of the AGI::GUS lines tested, the majority of GUS transcripts initiated 78 nucleotides downstream of the GUS translational start codon. This truncated GUS transcript variant could potentially generate a partial GUS protein lacking 104 amino acids at its N-terminus and while very unlikely, we cannot rule out the possibility that it retained its function and contributed to the observed GUS enzymatic activity of this particular line at this point. Another, weak, gel-purified band yielded an additional initiation site 25 bp upstream of the GUS coding sequence within the short segment of vector DNA separating the AGI and GUS sequences. The second AGI::GUS line tested exhibited a single transcription start site 87 bp upstream of the GUS translational start site within the short segment of exonic I-box flanking the 3′ end of the AGI (Fig. 3a; Table 2). In contrast, all GUS transcripts from the AGIP::GUS lines initiated within 10 bp of each other in the minimal 35S promoter downstream of the AGI (Fig. 3a; Table 3).
Fig. 3

5′ RACE analyses of enhancer-activated reporter gene transcripts. 5′ RACE results of AGI- (a) and 35S-mediated (b) reporter gene expression are depicted in graph format. Vertical bars represent the number of clones from total PCR reactions exhibiting transcript initiation sites within each 50 bp along the length of the various transgenic constructs. Gray and white bars represent 5′ start sites of GUS transcripts from two independent lines, respectively, while black bars represent the 5′ start sites of the GFP transcript. The asterisk indicates the 5′ terminus of GUS transcripts derived from gel-purified products. The relative positions of each of these sites can be extrapolated from the transcript diagrams shown beneath each graph. Horizontal, dashed arrows indicate the direction of transcription, with start sites depicted by short vertical lines. GUS: β-glucuronidase; GFP: green fluorescent protein; 35S: partially duplicated 35S promoter; T: transcriptional terminator; M: 84 bp 3′ portion of the MADS-box; AGI: 3,048 bp AG second intron; I: 11 bp 5′ portion of the I-box; P: minimal 35S promoter

Table 2

5′ termini of GUS transcripts in two independent lines bearing AGI::GUS or 35S/AGI::GUS constructs, respectively








−954 to −349a

−348 to −181

−180 to −96

−95 to −85

−84 to −1

+1 to 1,812



−25 (6)b*

+78 (9)



−87 (8)*



−421 (1)

−268 (1)


−47 (1)


−386 (3)


−16 (1)


−384 (1)


−372 (1)



−386 (1)

−280 (1)


−10 (1)


−384 (1)


35S: partially duplicated 35S promoter; M: 84 bp exonic region of the AGAMOUS MADS-box flanking the 5′ end of the AGI; I: 11 bp exonic region of the AGAMOUS I-box flanking the 3′ end of the AGI; GUS: β-glucuronidase coding sequence

* Transcription start sites indicated by an asterisk denote clones that were derived from a gel-purified PCR product

aNumbers refer to nucleotides of each particular region of the construct relative to the ATG (+1) of the GUS coding sequence

bNumbers outside of parentheses indicate 5′ start sites of GUS transcript relative to the ATG; numbers within parentheses indicate the number of clones identified for each start site. Blank boxes denote that no transcript start sites were detected in the respective construct region

Table 3

5′ termini of GUS transcripts in two independent lines bearing AGIP::GUS and 35S/AGIP::GUS constructs, respectively









−1,003 to −399a

−398 to −232

−231 to −148

−147 to −136

−136 to −38

−37 to −1

+1 to 1,812



−76 (10)b




−66 (5)




−238 (1)


−74 (3)


−66 (5)


−54 (1)


−48 (1)




−239 (1)


−144 (1)

−103 (1)


−81 (1)


−66 (1)


35S: partially duplicated 35S promoter; M: 84 bp exonic region of the AGAMOUS MADS-box flanking the 5′ end of the AGI; I: 11 bp exonic region of the AGAMOUS I-box flanking the 3′ end of the AGI; P: minimal 35S promoter; GUS: β-glucuronidase coding sequence

aNumbers refer to nucleotides of each particular region of the construct relative to the ATG (+1) of the GUS coding sequence

bNumbers outside of parentheses indicate 5′ start sites of GUS transcript relative to the ATG; numbers within parentheses indicate the number of clones identified for each start site. Blank boxes denote that no transcript start sites were detected in the respective construct region

The constitutive 35S enhancer overrides AGI-conferred tissue specificity regardless of the presence of a cognate promoter

The 35S enhancer is known to constitutively activate adjacent tissue-specific promoters in transgenic plants (Hily et al. 2009; Jagannath et al. 2001; Zheng et al. 2007). To test whether this enhancer is able to override both AGI- and AGIP-driven floral organ-specific GUS expression in vegetative tissues, we tested several transgenic constructs in which the 35S::eGFP cassette was fused in a head-to-head orientation with the AGI::GUS and AGIP::GUS fragments, respectively (Fig. 1c). This arrangement positions the duplicated enhancers of the 35S promoter (Kay et al. 1987) 167 bp upstream of the AGI, which allows both 35S promoter activity and its enhancer activation function to be evaluated by assaying GFP and GUS expression, respectively, in the same transgenic line. Histochemical assays for GUS staining revealed that 92.9% of 35S/AGI::GUS lines (n = 14) and 92.3% of 35S/AGIP::GUS lines (n = 13) exhibited strong, constitutive GUS expression in both flowers and leaves (Fig. 2c), resembling the typical expression pattern induced by the 35S promoter. Furthermore, GFP activity driven by the 35S promoter was also detected ubiquitously in leaves and flowers (Fig. 2d) of both 35S/AGI::GUS and 35S/AGIP::GUS lines, indicating that the 35S promoter and its enhancer activate identical patterns of gene expression. Semi-quantitative RT-PCR assays revealed that both GUS and GFP were expressed at comparable levels in flower and leaf tissues of 35S/AGI::GUS and 35S/AGIP::GUS lines (Fig. 2e), confirming the enzymatic results.

The 35S enhancer initiates transcription at multiple sites in 35S/AGI::GUS and 35S/AGIP::GUS lines

5′ RACE was utilized to determine the 5′ termini of GUS transcripts in 35S/AGI::GUS and 35S/AGIP::GUS lines (two independent lines of each). In 35S/AGI::GUS lines, transcription of GUS initiated at a range of sites within each line, the majority of which (8 of 13 clones) were clustered within the first of the two 35S enhancers (proximal to the AGI) of the partially duplicated 35S promoter fused to eGFP. The remaining transcript variants possessed 5′ termini within vector sequence separating either the 35S promoter and AGI or the AGI and GUS coding sequence (Fig. 3b; Table 2). Conversely, the vast preponderance of GUS transcripts (13 of 16 clones) initiated downstream of the AGI within the minimal 35S promoter in 35S/AGIP::GUS lines and in only a single case in each line was a transcript found to initiate upstream of the AGI (Fig. 3b; Table 3). Sequence analyses of 5′ RACE products indicated that in every instance in which GUS transcripts were initiated 5′ of the AGI in 35S enhancer-containing lines, this intronic sequence was accurately spliced out of the resulting RNA. When transcription was driven by an intact 35S promoter (containing both the enhancers and their cognate promoter), as was the case for the 35S::eGFP cassette in 35S/AGI::GUS and 35S/AGIP::GUS lines, transcription initiated within a 22 bp region of the 35S promoter in every instance (Fig. 3b).

Activation of GUS transcription by the 35S enhancer results in the generation of two major transcript variants in 35S/AGI::GUS and 35S/AGIP::GUS lines

The detection of multiple GUS transcription start sites by 5′ RACE assays in both 35S/AGI::GUS and 35S/AGIP::GUS lines suggests the existence of several GUS transcript variants. To validate this speculation, we conducted Northern blot analyses using total floral RNA from 35S/AGI::GUS, 35S/AGIP::GUS and enhancerless/promoterless ::GUS lines with GUS-specific and AGI-specific (lacking any flanking exonic sequence) probes, respectively, as well as a TUB4-specific probe as an internal standard. Relative probe annealing sites are indicated in Fig. 4a.
Fig. 4

Characterization of long-range action of the 35S enhancer. Northern blot analyses of GUS- and AGI-specific transcripts were performed to determine the relative lengths of transcript variants within 35S/AGI::GUS and 35S/AGIP::GUS lines. Relative probe annealing sites are depicted in (a). Black bars indicate probe binding sites and the lengths of sequences are indicated in nucleotides. Northern blot results following hybridization of total floral RNA with a GUS-specific probe and AGI-specific probe are shown in (b). Both 35S/AGI::GUS and 35S/AGIP::GUS lines exhibited two clusters of GUS transcripts, while hybridization of the same membrane with an AGI-specific probe revealed an approximately 5.3 kb band that corresponded in size to GUS transcript cluster 1. In all cases, the promoterless GUS construct did not yield any transgene-specific transcripts. Hybridization of the same blots with a TUB4-specific probe was utilized as an internal control to ensure the intactness of RNA. Schematic diagrams of GUS transcripts corresponding to the two major clusters are depicted in (c). T: transcriptional terminator; GFP: green fluorescent protein; 35S: partially duplicated 35S promoter; M: 84 bp 3′ portion of the MADS-box; AGI: 3,048 bp AG second intron; I: 11 bp 5′ portion of the I-box; GUS: β-glucuronidase; AGIP: AG second intron and associated partial exonic regions fused to a downstream minimal 35S promoter; M: molecular weight marker

Interestingly, we identified two major GUS transcript variants at approximately 5.3 kb (cluster 1) and 2.0–2.3 kb (cluster 2) in 35S/AGI::GUS and 35S/AGIP::GUS lines, but not the promoterless GUS control (Fig. 4b, c). The smaller of the two major clusters (cluster 2) corresponded in size to full-length GUS transcripts initiating upstream of the GUS coding sequence. However, because only 167 nucleotides separate the 35S enhancer and the 5′ end of the AGI sequence, it is also possible that these transcripts were variants in which transcription initiated within or immediately downstream of the 35S enhancer and the AGI was spliced out. The larger of the two major transcript clusters (cluster 1) included an additional 3 kb of sequence, which hinted at the possibility that it initiated upstream of an unspliced AGI, possibly within the 35S enhancer itself. Interestingly, this transcript variant was more pronounced in lines bearing the AGI fused to the minimal promoter than in lines containing the AGI alone. Weak, diffuse bands approximately 1.5–1.7 kb in size were also noted in both lines and probably corresponded to transcripts that initiated downstream of the GUS translational start codon and/or transcripts that terminated prematurely. In addition, a large band (>5.3 kb) of low intensity was also observed exclusively in the 35S/AGIP::GUS line, the identity of which is not known (Fig. 4b).

To ascertain that the 5.3 kb GUS transcripts were indeed the result of an unspliced AGI, we re-hybridized the membrane with an AGI-specific probe. As expected, bands that were virtually identical in size and intensity to the 5.3 kb GUS transcripts were detected in both 35S/AGI::GUS and 35S/AGIP::GUS lines, indicating that GUS transcripts bearing an unspliced AGI were in fact generated. No such band was observed in the promoterless GUS line, confirming that the production of these transcripts was linked to GUS expression.

Bioinformatic analyses of transgenic elements

To determine whether transcripts initiated by the AGI and 35S enhancers resulted from promiscuous activation of transcription or if cryptic promoter regions including typical plant signals existed within the various transgenic constructs, sequences were scanned for putative promoter elements and transcription start sites using computational tools. As expected, promoters were identified at the correct locations within the partially duplicated 35S promoter in the case of 35S/AGI::GUS and 35S/AGIP::GUS constructs and within the minimal promoter in cassettes containing the AGIP sequence. Furthermore, a cryptic promoter was also predicted in the forward orientation near the 5′ end of the GUS coding region (position +108 to +158 relative to the translation start codon). In addition, several cryptic promoters were predicted within the AGI (positions +198 to +248, +1,530 to +1,580 and +2,303 to +2,353 relative to the 5′ splice site). No potential promoters could be detected within any of the sequences in which promoter-independent enhancer-activated transcription initiation was found to occur, including intervening vector sequence, exonic fragments flanking the AGI, or the 35S enhancer region.

To predict whether enhancer-activated GUS transcripts were capable of generating functional GUS protein, transcript sequences were translated in silico and analyzed. In the case of all putative GUS transcripts from AGI::GUS and AGIP::GUS lines, as well as those transcripts initiating downstream of the AGI in 35S/AGI::GUS and 35S/AGIP::GUS lines, no in frame start or stop codons were identified upstream of the GUS coding sequence, which suggests that the transcripts would likely be translated correctly. Conversely, GUS transcripts initiating within or upstream of the associated exonic regions of the AGI in 35S/AGI::GUS lines possessed an in frame stop codon 54 nucleotides upstream of the GUS start codon in vector-derived sequence immediately 3′ of the AGI and associated exon. Likewise, the single GUS transcript initiating upstream of the AGIP in 35S/AGIP::GUS lines contained an in frame stop codon 156 nucleotides upstream of the GUS start codon within MADS-box derived exonic sequence 5′ of the AGIP. In neither case did these GUS transcripts possess in frame ATGs between the identified stop codons and GUS translational start site, suggesting that only translation initiating at the true start codon would result in the production of GUS protein. Similarly, spliced 35S-initiated transcripts did not contain any open reading frames of significant length in the short intervening sequence between the 35S enhancer and GUS coding region, and would likely be capable of generating a functional transcript. However, unspliced transcripts initiated by the 35S enhancer contained an open reading frame corresponding to 100 amino acids within the AGI, along with several other potentially disruptive regions, which indicates that the generation of functional GUS protein from such a transcript is not probable.


In this study, we analyzed the function and characterized the derived transcripts of two distinct transcriptional enhancer elements in Arabidopsis to gain insight concerning the mechanism behind their ability to activate transcription in a plant system; a subject about which very little is currently known. We chose to investigate the constitutive virus-derived 35S promoter and tissue-specific plant-derived AGI, which both contain well-known enhancer elements that adhere to the formal definition of an enhancer in that they exhibit very little or no dependence on either their orientation or their position relative to the transcription start site (Kay et al. 1987; Sieburth and Meyerowitz 1997). In addition, the 35S enhancer has been shown to override nearby enhancers/promoters, even at a distance of 4 kb (Hily et al. 2009; Singer et al. 2010); a capacity that we took advantage of in the design of our transgenic constructs. These constructs were also devised to enable the comparison of enhancer performance in the presence and absence of a target promoter since the activation function of transcriptional enhancers is typically described as involving a competent promoter upon which the enhancer exerts its function (Fig. 1).

It has been known for some time that the second intron of the Arabidopsis AGAMOUS gene and the CaMV35S promoter contain enhancer elements that promote the flower-specific and constitutive expression, respectively, of a downstream transgene when linked to a minimal promoter (Busch et al. 1999; Deyholos and Sieburth 2000; Kay et al. 1987). However, to the best of our knowledge, there have been no reports to date in which either of these enhancers has been found to drive the expression of a reporter gene in the absence of a cis-linked promoter. In this study, we found that both the AGI and 35S enhancers were able to activate reporter gene expression directly without any requirement for a cognate promoter (Fig. 2) and in both cases, this promoterless activation of gene expression generated the same spatiotemporal pattern of transcription as in the presence of an associated promoter: while the AGI instigated the tightly regulated stamen- and carpel-specific expression of a closely linked downstream GUS gene (Fig. 2a, b), the 35S enhancer exerted its constitutive activation function to override the AGI-specific expression of the GUS reporter gene (Fig. 2c–e).

5′ RACE assays of AGI::GUS and 35S/AGI::GUS lines, which lack a cognate promoter, exhibited transcription initiation at various sites within and/or downstream of the respective enhancer elements in each line (Fig. 3; Table 2). In the case of GUS transcript variants that initiated upstream of the AGI in 35S/AGI::GUS lines, the AGI was faithfully spliced out in every instance. Northern blot data provided further evidence that the majority of GUS transcripts in 35S/AGI::GUS lines lacked the intron (cluster 2; Fig. 4), while only a minority included an unspliced AGI (cluster 1; Fig. 4). The fact that the two AGI::GUS lines and two 35S/AGI::GUS lines analyzed by 5′ RACE, respectively, exhibited similar positions for transcription initiation, but the exact locations of the sites were different (Fig. 3; Table 2), suggests that these transcriptional initiation events were promiscuous in nature and were not the result of the unanticipated presence of cryptic promoters. Indeed, bioinformatic analyses of the 35S enhancer and AGI themselves, as well as the intervening vector sequence, suggest that this enhancer-driven transcription in the absence of an associated promoter was independent of elements containing typical plant signals since the transcription start sites were not found to occur within any predicted promoter regions.

Earlier reports using animal and yeast systems have also demonstrated that enhancer elements are able to initiate transcription autonomously in the absence of a nearby promoter (Dobi and Winston 2007; Kim et al. 2007; Ling et al. 2005). Moreover, enhancer-directed expression of a reporter gene without a cis-linked promoter has been described previously for the metazoan HS2 and HS3 enhancers of the globin genes (Routledge and Proudfoot 2002; Tuan et al. 1992) and is likely attributable to the fact that the intervening sequence between enhancer and target gene was so small that any additional untranslated 5′ regions generated by autonomous enhancer-mediated transcription would not interfere with reporter gene translation. This was almost certainly also the case in our AGI::GUS and 35S/AGI::GUS lines, since less than 100 nucleotides separated the AGI and GUS coding regions in AGI::GUS lines and the splicing out of the AGI enhancer in 35S/AGI::GUS lines would bring the 35S enhancer and GUS coding region within 348 nucleotides of each other. While it is possible that the transcription events initiated by the 35S enhancer in 35S/AGI::GUS lines were dependent upon the presence of the AGI sequence, this does not appear to be the case as preliminary results indicate that the 35S enhancer is able to activate constitutive GUS expression even in transgenic plants bearing constructs comprising the 35S::GFP and promoterless/enhancerless GUS cassettes in a head-to-head orientation (i.e., lacking a transgenic copy of the AGI altogether; data not shown). Taken together, these results suggest that promiscuous transcription initiation at proximal sites may be an inherent property of transcriptional enhancer elements, especially in artificial systems which lack the mechanisms that are in place in a genomic context to prevent unwanted interactions between enhancers and non-target genes.

In contrast, the presence of a target minimal promoter, as in AGIP::GUS and 35S/AGIP::GUS lines, apparently caused the enhancers to adopt different modes of action for transcriptional activation. In the case of AGIP::GUS lines, transcription of the GUS transgene initiated exclusively at the minimal promoter downstream of the AGI (Fig. 3a; Table 3). Since the AGI and its linked minimal promoter were in such close proximity in these lines, it is very probable that they utilized a short-range mode of enhancer action whereby the enhancer and promoter communicated directly without the need for any facilitating mechanisms (Bondarenko et al. 2003).

35S/AGIP::GUS lines, in which the 35S enhancer and its target minimal promoter linked to the GUS coding region were separated by over 3 kb (due to the presence of the intervening AGI sequence), evidently instigated an entirely different means of transcriptional activation by the 35S enhancer. Northern blot analyses (Fig. 4) of 35S/AGIP::GUS lines displaying 35S enhancer-mediated constitutive expression of the GUS coding sequence indicated that two major clusters of GUS transcripts were generated: the larger of the two clusters (cluster 1) resulted from transcription initiation near the 35S enhancer and included an unspliced AGI while the smaller of the two transcripts (cluster 2) appeared to either initiate in the same region as the long transcripts but splice out the AGI, or to initiate downstream of the AGI. 5′ RACE results (Fig. 3b; Table 3) revealed that the vast majority of GUS transcripts in these lines initiated within the minimal promoter fused to the AGI, with only a single transcript with a start site 5′ of the spliced AGI detected in each line tested. It is not surprising that unspliced transcript variants were not identified using the RACE method because of its tendency to preferentially amplify small fragments. However, further 5′ RACE assays utilizing primers that annealed near the 5′ terminus of the AGI indicated that transcripts including an unspliced AGI that initiated within the 35S enhancer were indeed generated in 35S/AGIP::GUS lines (data not shown).

Taken together, these results suggest that two distinct transcriptional initiation regions were present in 35S/AGIP::GUS lines. The first transcript variant (cluster 1; Fig. 4) was initiated by the 35S enhancer and was transcribed through the intervening AGI, which was generally not spliced out, and at least a large portion of the GUS coding sequence. This transcript was undoubtedly not translated into a functional GUS protein because of the length of its untranslated 5′ region and its inclusion of a variety of alternative open reading frames of significant length. A second transcript variant (cluster 2; Fig. 4) was initiated at the minimal promoter fused to the AGI, generating constitutively expressed, functional GUS transcripts. Intriguingly, these results are very similar to those observed during long-range transcriptional activation by the HS2 enhancer from the human globin locus in an artificial vector (Ling et al. 2005) where in the presence of a cis-linked promoter, non-coding intergenic RNAs were initiated from multiple sites within the HS2, which elongated through the intervening DNA and target promoter and into the reporter gene, while functional reporter gene mRNA were initiated at a single site within the promoter.

In fact, several studies in animal systems have revealed that in addition to the generation of functional gene transcripts, non-coding RNA derived from sequences between distantly situated enhancers and promoters are also produced (Dobi and Winston 2007; Ho et al. 2006; Masternak et al. 2003; Rogan et al. 2004; Tchurikov et al. 2009; Zhu et al. 2007). These intergenic transcripts have been suggested to be the by-product of a facilitated tracking mechanism in which RNA polymerase II and the attached enhancer migrate along the DNA from enhancer to target promoter to ultimately form a loop (Zhu et al. 2007). This tracking of RNA polymerase II along the intervening DNA to synthesize non-coding intergenic RNAs is thought to deliver enhancer-bound proteins to the cis-linked promoter and/or, through the association of histone acetyltransferases, ‘open’ the nucleosomal structure of the gene domain (Zhu et al. 2007).

The characteristics of the derived intergenic RNAs have been found to be quite diverse depending on the system analyzed. For example, a series of overlapping, polyadenylated, intergenic RNAs of less than 3 kb were found to be generated by the HS2 enhancer towards its linked globin promoter at its endogenous loci (Ling et al. 2005; Zhu et al. 2007), whereas in an artificial vector the same enhancer synthesized intergenic transcripts that were initiated from multiple sites within the enhancer and elongated through the cis-linked promoter and into the reporter gene (Ling et al. 2005). In contrast, the chicken α-globin gene locus has been found to generate enhancer-initiated intergenic RNAs of longer than 20 kb that span the sequence separating enhancer and promoter (Broders and Scherrer 1987; Razin et al. 2004), while at the human prostrate specific antigen locus, the scanning of polymerase II from enhancer to promoter has not been found to generate intergenic RNAs at all (Wang et al. 2005). Therefore, while scanning/facilitated tracking seems to be a rather widespread means of establishing communication between regulatory elements, the mechanism by which this is accomplished appears to vary and a debate still exists whether long, enhancer-initiated transcripts are required for the proper expression of the target gene (for example Müller et al. 1990). However, the vast majority of studies provide evidence for the importance of these intergenic transcripts to enhancer–promoter communication and the initiation of genic transcription (Dobi and Winston 2007; Drewell et al. 2002; Ho et al. 2006; Ling et al. 2004).

It is not definitively known whether the long 35S enhancer-initiated intergenic transcripts in 35S/AGIP::GUS lines were required for the activation of the short, functional GUS transcripts initiated from the minimal promoter. Nevertheless, our results provide evidence that the 35S enhancer assembled an RNA polymerase transcription complex, which mediated long-range enhancer function by scanning through intervening DNA to generate long, non-coding intergenic RNAs to reach a distant promoter and activate constitutive mRNA synthesis of the GUS reporter gene. It has been proposed that the generation of non-coding intergenic and coding genic transcripts in long-range enhancer/promoter systems require independent transcription complexes that possibly differ from one another (Kim et al. 2007; Ling et al. 2005), which might also be the case in our system and may explain the prevalence of unspliced variants initiated from the 35S enhancer in 35S/AGIP::GUS lines.

At this point we cannot say whether the scanning of RNA polymerase along the intervening DNA sequence in the long-range activation of a target promoter ultimately results in the formation of a direct link between enhancer and promoter via looping of the chromatin fiber in our system as would be expected for a facilitated tracking mechanism. However, the development of chromatin conformation capture technology in plants may allow further insight into long-range enhancer function; information which will likely be crucial for the development of transgenic technology in the future.


The authors wish to thank Mr. Dennis Bennett (USDA-ARS, Kearneysville, WV) for his technical assistance. This study was funded by the USDA-ARS Headquarter 2007 class of postdoctoral grants and a USDA CSREES BRAG grant (2006-03701).

Copyright information

© U.S. Government 2010

Authors and Affiliations

  • Stacy D. Singer
    • 1
    • 2
  • Kerik D. Cox
    • 2
  • Zongrang Liu
    • 1
  1. 1.USDA-ARS Appalachian Fruit Research StationKearneysvilleUSA
  2. 2.Department of Plant Pathology, New York State Agricultural Experiment StationCornell UniversityGenevaUSA

Personalised recommendations