Combinatorial control of Arabidopsis proline dehydrogenase transcription by specific heterodimerisation of bZIP transcription factors

Fridtjof Weltmeier, Andrea Ehlert, Caroline S Mayer, Katrin Dietrich, Xuan Wang, Katia Schütze, Rosario Alonso, Klaus Harter, Jesús Vicente‐Carbajosa, Wolfgang Dröge‐Laser

Author Affiliations

  1. Fridtjof Weltmeier1,,
  2. Andrea Ehlert1,,
  3. Caroline S Mayer1,
  4. Katrin Dietrich1,
  5. Xuan Wang1,
  6. Katia Schütze2,
  7. Rosario Alonso3,
  8. Klaus Harter2,
  9. Jesús Vicente‐Carbajosa3 and
  10. Wolfgang Dröge‐Laser*,1
  1. 1 Albrecht‐von‐Haller Institut, Universität Göttingen, Göttingen, Germany
  2. 2 Zentrum für Molekularbiologie der Pflanzen, Pflanzenphysiologie, Tübingen, Germany
  3. 3 Department of Bioquimica y Biologia Molecular, ETSI Agronomos, Universidad Politecnica, Ciudad Universitaria, Madrid, Spain
  1. *Corresponding author. Albrecht‐von‐Haller Institut, University of Götiingen, Untere Karspüle 2, Göttingen 37073, Germany. Tel.: +49 (0)551 39 19816; Fax: +49 (0)551 39 7820; E-mail: wdroege{at}
  1. These authors contributed equally to this work

View Full Text


Proline metabolism has been implicated in plant responses to abiotic stresses. The Arabidopsis thaliana proline dehydrogenase (ProDH) is catalysing the first step in proline degradation. Transcriptional activation of ProDH by hypo‐osmolarity is mediated by an ACTCAT cis element, a typical binding site of basic leucine zipper (bZIP) transcription factors. In this study, we demonstrate by gain‐of‐function and loss‐of‐function approaches, as well as chromatin immunoprecipitation (ChIP), that ProDH is a direct target gene of the group‐S bZIP factor AtbZIP53. Dimerisation studies making use of yeast and Arabidopsis protoplast‐based two‐hybrid systems, as well as bimolecular fluorescence complementation (BiFC) reveal that AtbZIP53 does not preferentially form dimers with group‐S bZIPs but strongly interacts with members of group‐C. In particular, a synergistic interplay of AtbZIP53 and group‐C AtbZIP10 was demonstrated by colocalisation studies, strong enhancement of ACTCAT‐mediated transcription as well as complementation studies in atbzip53 atbzip10 T‐DNA insertion lines. Heterodimer mediated activation of transcription has been found to operate independent of the DNA‐binding properties and is described as a crucial mechanism to modulate transcription factor activity and function.


Osmotic stress, caused by drought, high salinity and cold is a key determinant of plant growth and productivity (Hoekstra et al, 2001). In response to environmental conditions causing osmotic stress, many higher plants accumulate protective compounds, like the compatible osmolyte l‐proline (L‐Pro) (Delauney et al, 1993; Liu and Zhu, 1997; Parvanova et al, 2004). This compound has been proposed to participate in the protection of plasma membrane integrity (Mansour, 1998) and to act as a scavenger of radicals (Smirnoff and Cumbes, 1989). In addition, L‐Pro is discussed to be a source of reducing power (Walton and Boldingh, 1991) and to serve as a transient storage of carbon and nitrogen (Peng et al, 1996).

The L‐Pro homeostasis is tightly regulated by its synthesis (Delauney et al, 1993), degradation (Verbrüggen et al, 1996) and transport (Rentsch et al, 1996). L‐Pro is synthesised in the cytosol from glutamine or ornithin precursors (Figure 1A). In osmotically stressed tissues, L‐Pro mainly results from the glutamine pathway by the activity of the NADPH using enzymes Δ1‐pyrroline‐5‐carboxylate‐synthetase (P5CS) and P5C reductase (P5CR) (Delauney et al, 1993; Roosens et al, 1999). L‐Pro accumulation correlates with osmotic stress tolerance, however, a stringent causal relationship is not yet firmly established (Nanjo et al, 1999). Furthermore, accumulation of L‐Pro is toxic and thus L‐Pro and its metabolites have been discussed to function as signalling molecules in stress‐induced cell death (Hellmann et al, 2000; Nanjo et al, 2003; Deuschle et al, 2004).

Figure 1.

The bZIP transcription factor AtbZIP53 regulates ProDH transcription in the hypo‐osmolarity response. (A) Schematic overview of the metabolic pathways involved in l‐proline (L‐Pro) synthesis and degradation in plant cells. P5CS: Δ1‐pyrroline‐5‐carboxylate‐synthetase; P5CR: reductase; ProDH: L‐Pro dehydrogenase; P5CDH: P5C dehydrogenase (Peng et al, 1996; Verbrüggen et al, 1996; Nakashima et al, 1998; Yoshiba et al, 1999). (B) Gene induction by hypo‐osmotic stress treatment. Arabidopsis plants, 14‐day‐old, were transferred from MS medium (58 mM sucrose) to desalted water (Satoh et al, 2004). RNA of control (0 h) and stress‐treated plants (time points as indicated) was analysed by Northern hybridisation. Depicted are hybridisation experiments with ProDH‐ and AtbZIP53‐specific probes, respectively. (C) Schematic illustration of the T‐DNA insertion mutant atbzip53. The insertion was mapped inside the 5′utr 430 bp upstream of the ATG. (D) Northern analysis of wild‐type (Wt), AtZIP53 T‐DNA insertion lines (atbzip53), and AtbZIP53‐overexpressing plants (Pro35S:HA‐AtbZIP53). Endogenous (e) and transgene encoded (t) AtbZIP53 RNAs show slightly different mobility due to the length of their 5′ leaders. (E) Western Analysis of wild‐type (Wt) and several independent transformants of Pro35S:HA‐AtbZIP53. The transgene‐encoded proteins are detected by a HA‐specific antibody (arrow). (F) Induction of ProDH transcription in response to hypo‐osmolarity treatment. RNA was isolated after the time points indicated and ProDH transcript level of Wt, atbzip53, and Pro35S:HA‐AtbZIP53 plants are compared. Equal loading was verified by ethidium bromide (EtBr) staining and hybridisation with an actin‐specific probe. All experiments have been replicated at least three times with similar results.

Induced by rehydration, L‐Pro is degraded by a pathway localised in the mitochondria, which makes use of the enzymes proline dehydrogenase (ProDH) and P5C dehydrogenase (P5CDH) (Peng et al, 1996; Verbrüggen et al, 1996; Nakashima et al, 1998; Yoshiba et al, 1999). ProDH catalyses the rate‐limiting step in‐Pro degradation. Transgenic Arabidopsis plants expressing a ProDH antisense construct display high L‐Pro levels and enhanced tolerance to freezing conditions and high salinity (Nanjo et al, 1999).

In Arabidopsis thaliana, the ProDH gene (At3g30775) is induced in response to hypo‐osmotic conditions occurring during rehydration after stress recovery (Kiyosue et al, 1996). ProDH transcription is also enhanced by high levels of L‐Pro and reduced by dehydration (Peng et al, 1996; Verbrüggen et al, 1996). Detailed dissection of the 1.4‐kb ProDH promoter revealed an ACTCAT motif which is sufficient for induction of the ProDH gene by L‐Pro or hypo‐osmolarity, respectively (Nakashima et al, 1998; Satoh et al, 2002). Microarray analysis confirmed that among 121 rehydration‐inducible genes, 48% harbour the ACTCAT motif in their promoters (Oono et al, 2003). These findings indicate that this motif is of general relevance with respect to gene regulation during recovery after osmotic stress. Therefore, ACTCAT‐mediated transcription might serve as a fruitful model system for studying transcriptional control of osmotic stress responses, which is of importance for both, basic science as well as genetic engineering of crop plants.

The ACTCAT‐motif is related to the GCN4‐binding site (ATGA(C/G)TCAT) (Ellenberger et al, 1992) recognised by members of the basic leucine zipper (bZIP) transcription factor family (Landschulz et al, 1988). In the model plant Arabidopsis thaliana, 75 bZIP genes have been identified and classified into 10 groups (Jakoby et al, 2002). In vitro DNA‐binding studies and transfection experiments performed in transiently transformed protoplasts revealed that distinct members of Arabidopsis group‐S bZIP proteins are involved in ACTCAT‐mediated transcription (Satoh et al, 2004). Although 17 group‐S bZIPs have been identified, only a subgroup consisting of five highly related members (AtbZIP1, AtbZIP2, AtbZIP11/ATB2, AtbZIP44 and AtbZIP53) seem to be involved in ProDH regulation (Satoh et al, 2004) and will be referred to as group‐S1 bZIPs.

bZIP proteins are phylogenetically widely distributed transcriptional regulators, characterised by a basic DNA‐binding domain (b) (Landschulz et al, 1988). Although DNA binding of bZIP monomers has been described, they normally interact with DNA as dimers (Ellenberger et al, 1992; Metallo and Schepartz, 1997; Cranz et al, 2004). Dimerisation is mediated by the so‐called leucine zipper domain, a heptad repeat of leucine or other bulky hydrophobic amino acids creating an amphipathic helix (Landschulz et al, 1988; Baxevanis and Vinson, 1993). The formation of bZIP homo‐ or heterodimers offers a huge combinatorial flexibility to regulatory transcription systems. By heterodimerisation, DNA‐binding specificity and affinity, transactivation properties and ultimately, cell physiology might be altered (Naar et al, 2001). For instance, in the animal system, heterodimers of the bZIP proteins Jun and Fos have been shown to recognise the AP1 motif, which is not targeted by Jun homodimers (Kouzarides and Ziff, 1988). As bZIP proteins do not heterodimerise promiscuously but specifically (Newman and Keating, 2003), an important function for gene regulation is anticipated. In planta, bZIP heterodimerisation has been shown for closely related bZIPs, for example, the GBFs (Schindler et al, 1992). In tobacco (Strathmann et al, 2001) and parsley (Rügner et al, 2001), specific heterodimerisation was found between members of two groups of bZIPs that are related to the Arabidopsis groups S and C. Interestingly, heterodimerisation between these groups is strongly preferred in comparison to homotypic dimerisation.

In this work, we demonstrate by means of transgenic and ChIP approaches that ProDH is a direct target gene of the group‐S1 bZIP transcription factor AtbZIP53. Using yeast two‐hybrid, BiFC and interaction assays in protoplasts, we show that AtbZIP53 does not function as a homodimer, but preferentially heterodimerises with group‐C bZIP transcription factors. AtbZIP53 and the group‐C heterodimerisation partner AtbZIP10 were found to be colocalised and to control ACTCAT‐mediated ProDH transcription in a synergistic manner. We describe bZIP heterodimerisation as an essential mechanism to modulate transactivation properties of transcriptional regulators in planta. Furthermore, we provide in vivo evidence that ProDH is not regulated by a single bZIP but by a complex heterodimerisation network of group‐S1/C bZIP factors.


AtbZIP53 directly participates in the hypo‐osmolarity‐induced transcription of the ProDH gene

Using in vitro DNA‐binding studies and transfection experiments in Arabidopsis protoplasts, Satoh et al (2004) demonstrated that the group‐S1 bZIP transcription factors bind to the ACTCAT cis element located in the ProDH promoter. However, expression data of the correspondent bZIP genes, obtained from the RARGE database ( or by Satoh et al (2004), suggest that AtbZIP53 constitutes as the main regulator of hypo‐osmolarity‐induced ProDH transcription. Whereas AtbZIP53 is strongly induced by hypo‐osmolarity treatment (Figure 1B), AtbZIP11 or AtbZIP44 are not. We therefore focussed on studying the function of AtbZIP53 in the regulation of ProDH transcription. Expression of an HA‐tagged AtbZIP53 gene driven by the 35S promoter (Pro35S:HA‐AtbZIP53), as monitored by Northern and Western analyses (Figure 1D and E), resulted in a constitutive activation of the ProDH gene (Figure 1F). Comparable results were obtained with several transgenic lines expressing an untagged construct indicating that this response is not due to the HA‐tag (data not shown). Complementarily, a partial loss‐of‐function mutant of AtbZIP53 (atbzip53) (Figure 1C and D) displayed a significantly reduced and delayed induction of ProDH transcription after hypo‐osmolarity treatment (approximately by 40%) (Figure 1F). Owing to redundancy, AtbZIP53 and other group‐S1 bZIP transcription factors (Jakoby et al, 2002) might show an overlapping function which results in the residual activation of ProDH expression in the AtbZIP53 plant.

To verify direct binding of AtbZIP53 to the ProDH promoter in vivo, ChIP was performed applying chromatin obtained from HA‐AtbZIP53‐expressing plants (Pro35S:HA‐AtbZIP53) using an HA‐tag‐specific antibody. A significant enrichment of ProDH promoter fragments was detected by PCR amplification, using ProDH promoter‐specific primers (Figure 2). Control primers, unrelated to the AtbZIP53 target gene did not result in an enrichment following immunoprecipitation (Figure 2). In summary, the in vitro and in vivo evidence support the hypothesis that the ProDH promoter is a direct target of the AtbZIP53 transcription factor.

Figure 2.

In vivo binding of AtbZIP53 to the ProDH promoter. ChIP analysis of chromatin obtained from Pro35S:HA‐AtbZIP53 (+) and Wt (−) plants using a HA‐specific antibody for immunoprecipitation. ProDH primers were used to amplify a 191 bp fragment of the ProDH promoter. Given are the PCR products obtained after 35, 38, or 41 cycles, respectively. As input control, PCR amplicons are shown which have been obtained after 25 cycles with material before immunoprecipitation. The specificity was verified by using primers amplifying the promoter of a nontarget gene (cwINV2, At3g52600). Given are amplicons obtained after 38, 41 or 44 cycles, respectively. The PCR products were stained by ethidium bromide. The experiments have been replicated three times with similar results.

AtbZIP53 heterodimerises with group‐C bZIP transcription factors in yeast and plant cells

bZIP factors usually bind DNA as homo‐ or heterodimers, consequently, dimer formation allows the establishment of complex regulatory networks based on combinatorial interactions. It is tempting to speculate that heterodimers of AtbZIP53 formed with the related group‐S1 bZIP factors are regulating ProDH transcription. However, when applying a yeast two‐hybrid approach, no AtbZIP53 homodimers or AtbZIP53 containing heterodimers were formed with any of the group‐S1 bZIP proteins or other less related group‐S bZIP factors (Figure 3A). A quantitative assay for protein–protein interactions confirmed the weak heterodimerisation capacity of AtbZIP53 with other group‐S1 members (data not shown). We further analysed the potential of AtbZIP53 to interact with other bZIP factors. As described for tobacco (Strathmann et al, 2001) and parsley (Rügner et al, 2001), group‐S1‐like bZIP proteins specifically heterodimerise with group‐C‐like bZIPs. Strong heterodimerisation of AtbZIP53 was detected with all Arabidopsis group‐C bZIP proteins, namely AtbZIP9, AtbZIP10, AtbZIP25 and AtbZIP63 (Figure 3A). As no heterodimerisation with members of the bZIP groups A (e.g. AtbZIP39) and D (e.g. AtbZIP22) was found, this heterodimerisation is assumed to be specific. In order to analyse whether the protein–protein interaction is mediated by the leucine zipper domain, two conserved residues (leucine51 and alanine72) were changed to prolines which should disrupt the α‐helical structure of the ZIP domain (AtbZIP53PP). As shown in Figure 3B, no interaction between AtbZIP53PP and group‐C members, such as AtbZIP10 and AtbZIP9 was observed.

Figure 3.

Yeast two‐hybrid analysis studying heterodimerisation of AtbZIP53 with group‐S and ‐C bZIP factors. (A) Interaction of BD‐AtbZIP53 and the AD‐bZIP fusion proteins indicated lead to prototrophic growth of the yeast strain MAV203 (ProQuest®, Invitrogen) on medium lacking uracil. Group‐C bZIP (AtbZIP63, AtbZIP9, AtbZIP25, AtbZIP10) and group‐S bZIP proteins (AtbZIP53, AtbZIP2, AtbZIP44, AtbZIP11, AtbZIP42, AtbZIP7, AtbZIP6, AtbZIP4, AtbZIP8) are labelled by a grey or white background, respectively. (B) Interaction of AtbZIP53 and group‐C bZIP proteins is mediated by the ZIP domain. ZIP‐specific exchanges of the amino acids leucine51 and alanine72 by Pro (AtbZIP53PP) disrupt the α‐helical structure and results in loss of interaction (right) in comparison to the wild‐type AtbZIP53 (left).

To verify bZIP heterodimerisation in planta, BiFC assays were performed in Agrobacterium‐infiltrated tobacco (Nicotiana benthamiana) leaves (Walter et al, 2004) using the strongly interacting AtbZIP53 and AtbZIP10 as a representative example. Both bZIP factors localise to the nucleus when expressed as YFP fusion proteins (Figure 4A and B). When fused to the N‐ and C‐terminal YFP fragment and coexpressed in tobacco leaves, a strong BiFC signal was detected in the nuclear compartment of the transformed cells (Figure 4C), confirming the interaction of AtbZIP53 and AtbZIP10 in planta. Control experiments in which C‐YFP‐AtbZIP53 and unfused N‐YFP protein are expressed do not show any fluorescence (Figure 4D). In order to demonstrate the specificity of BiFC system, the C‐YFP‐AtbZIP53PP mutation was coexpressed with N‐YFP‐AtbZIP10. Specific mutation of the ZIP domain (see Figure 3B) interferes with heterodimerisation and in consequence does not result in positive BiFC signals (Figure 4E). In order to confirm expression of the BiFC constructs, Western analyses have been performed (data not shown). Finally, homodimerisation of AtbZIP53 and AtbZIP10 was analysed in this system, but no or only significantly weaker homotypic interactions could be detected (data not shown). In summary, these experiments provide conclusive evidence for the formation of group‐S1/C heterodimers in planta.

Figure 4.

BiFC analysis of AtbZIP53/AtbZIP10 heterodimerisation in planta. Localisation of YFP‐AtbZIP53 (A) and YFP‐AtbZIP10 (B) fusion proteins in Nicotiana benthamiana leaf epidermal cells after Agrobacterium‐mediated transient transformation. (C) Fluorescence observed by confocal microscopy which results from complementation of the C‐terminal (C‐YFP) and N‐terminal (N‐YFP) part of the YFP protein fused to AtbZIP53 and AtbZIP10, respectively. (D) Negative control making use of C‐YFP‐AtbZIP53 and an unfused N‐YFP protein. (E) Coexpression of the C‐YFP‐AtbZIP53pp mutant (see Figure 3B) and N‐YFP‐AtbZIP10. YFP‐Epifluorescence (left) and bright field overlay images (right) are depicted.

Heterodimerisation with group‐C bZIPs enhances the transactivation properties of AtbZIP53 in planta

In order to quantify bZIP heterodimerisation in planta, a two‐hybrid approach was established in Arabidopsis protoplasts as described by Ehlert et al (2006). AtbZIP53 and the related group‐S1 and group‐C bZIP factors were fused to the GAL4 DNA‐binding domain (BD) (Giniger et al, 1985) and expressed under control of the 35S promoter. Cotransfection with a GUS reporter, which is driven by GAL4 upstream activating sequence (GAL‐UAS4:GUS), allows quantification of bZIP activation properties independent of their DNA binding (Figure 5A). Whereas group‐S1 members AtbZIP11, AtbZIP44 and AtbZIP2 showed significant activation capacity; AtbZIP53, AtbZIP1 and the group‐C members displayed almost no intrinsic activation (Figure 5B).

Figure 5.

Analysis of AtbZIP53 heterodimerisation in transiently transformed Arabidopsis protoplasts. Experimental set‐up (A) to analyse bZIP activation properties. The activation of a GUS reporter driven by multimeric GAL4‐UAS DNA‐binding sites (GAL‐UAS4:GUS) was measured after expression of GAL4‐BD fusions of group‐S1 and C bZIP proteins (BD‐AtbZIP) (B). Experimental set‐up (C) to analyse AtbZIP53 heterodimerisation. Activation of a GAL‐UAS4:GUS reporter was measured after coexpression of BD‐AtbZIP53 and bZIP proteins fused to an AD domain of GAL4 (AD, black bars) or an HA‐tag (grey bars) (D). In single transfection experiments, 9 μg of reporter and 28 μg of effector plasmid DNA were applied. In cotransfections, 14 μg of each AtbZIP effector plasmid were used. Variations in transformation efficiency are normalised by using 3 μg of a Pro35S:NAN plasmid construct per transfection (Kirby and Kavanagh, 2002). Given are mean values and s.d.'s of four independent transfections as relative GUS/NAN units.

In order to study heterodimerisation of AtbZIP53, BD‐AtbZIP53 was individually coexpressed with GAL4 activation domain (AD) fusions of the bZIP heterodimerisation partners (Figure 5C). Similar to the results obtained in yeast, almost no AtbZIP53 homodimerisation and only weak interaction with the group‐S1 members AtbZIP2, AtbZIP11 and AtbZIP44 was observed (Figure 5D). A strong heterodimerisation of AtbZIP53 with the group‐C members AtbZIP10, AtbZIP25 and AtbZIP9 was detected. In contrast, interaction between AtbZIP53 and group‐C AtbZIP63 was less pronounced.

Heterodimerisation might modulate the transactivation capacity of bZIP transcription factors. We therefore coexpressed BD‐AtbZIP53 with bZIP partners which have not been fused to the AD domain (Figure 5D). Strong reporter gene activation was obtained for AtbZIP53/AtbZIP10 and AtbZIP53/AtbZIP25 heterodimers. As neither AtbZIP53 nor group‐C bZIPs exhibit strong activation on their own (Figure 5B), these data indicate that specific heterodimerisation may act as a key mechanism to modulate transactivation properties in planta. This mechanism will be referred to as heterodimer‐induced transactivation (HIT).

In order to analyse the HIT mechanism, the following control has been included into the experimental set‐up. Fusing a BD domain to the N‐terminus of the bZIP proteins might mask endogenous activation properties of the bZIP proteins analysed in Figure 5B. We therefore coexpressed BD‐AtbZIP53 with the other BD‐bZIP proteins which also lead to the synergistic reporter gene activation in the case of AtbZIP53/AtbZIP10 and AtbZIP53/AtbZIP25, respectively (data not shown).

The HIT mechanism is not restricted to AtbZIP53/AtbZIP10 and AtbZIP53/AtbZIP25 heterodimers. A complete analysis of all group‐S1/C interactions, according to the set‐up described in Figure 5C, revealed that the HIT mechanism is a typical feature of distinct heterodimers (Supplementary Table S1). It is particularly observed between AtbZIP25, AtbZIP10 and most group‐S1 members. Moreover, HIT is also observed inside group‐C, as demonstrated for AtbZIP10/AtbZIP63. These data suggest that HIT is of general importance for the function of the group‐C and S bZIP factors.

Group‐S1 and C bZIPs bind specifically the ACTCAT‐regulatory element present in the ProDH promoter

ProDH promoter analysis revealed an ACTCAT motif to be necessary and sufficient for hypo‐osmolarity‐induced gene activation (Satoh et al, 2002). To assess the significance of AtbZIP53 homo‐ or heterodimer participation in ACTCAT‐mediated transcriptional control, we transiently expressed AD‐bZIP proteins in Arabidopsis protoplasts in conjunction with a reporter vector containing this regulatory element. In this experimental system, activation of a (ACTCAT)2:GUS reporter should solely rely on the binding properties of the bZIP proteins and be independent of their intrinsic activation capacity. The results in Figure 6 show that all bZIP proteins tested are able to interact with this element, and demonstrate that AtbZIP53 binds more effectively the ACTCAT motif than group‐C bZIPs. The specificity of this interaction was studied by introducing mutations into the ACTCAT element. The use of a (ACTtta)2:GUS reporter resulted in a loss of activation for all bZIPs, confirming that the ACTCAT motif is a specific binding site for these proteins in planta (Figure 6).

Figure 6.

ACTCAT‐specific DNA binding of group‐S1 and C bZIP factors. Arabidopsis protoplasts were transiently transformed with a GUS reporter construct driven by the ACTCAT motif ((ACTCAT)2:GUS) (black and grey bars) or a mutated version ((ACTttg)2:GUS) (white bars). Activation of the reporters was measured after expression of AtbZIP proteins (grey and white bars) or AD‐AtbZIP proteins (black bars). In single transfection experiments, 9 μg of reporter and 28 μg of effector plasmid DNA was applied. Variations in transformation efficiency are normalised by using 3 μg of a Pro35S:NAN plasmid construct per transfection (Kirby and Kavanagh, 2002). Given are mean values and s.d.'s of four independent transfections as relative GUS/NAN units.

AtbZIP53‐containing heterodimers bind the ProDH promoter ACTCAT‐cis‐element and regulate ACTCAT‐mediated transcription

The results of the heterodimerisation assays prompted us to analyse whether ProDH is regulated by AtbZIP53‐containing heterodimers. We therefore coexpressed AtbZIP53 in combination with AD‐bZIP fusions of group‐S1 and C bZIPs (Figure 7). As we observed for the GAL‐UAS supported transcription, synergistic enhancement of the ACTCAT‐mediated transcription was obtained by coexpression of AtbZIP53 and its group‐C heterodimerisation partners AtbZIP9, AtbZIP10, AtbZIP25 and AtbZIP63, respectively. In order to analyse the endogenous activation properties, the bZIP proteins were expressed without the AD domain (Figure 7). Equal expression of the proteins was monitored by Western blot using an HA‐tag‐specific antibody (data not shown). Whereas expression of single bZIP proteins resulted in weak reporter activation, a significant synergistic activation was obtained with all AtbZIP53/group‐C heterodimers. Based on these data, a number of heterodimers seem to fulfil a partly redundant function in ACTCAT‐mediated transcription.

Figure 7.

The ACTCAT‐mediated transcription is regulated by AtbZIP53 heterodimers. Regulation of the (ACTCAT)2:GUS reporter by heterodimers was analysed by coexpression of AtbZIP53 and group‐S1 and C bZIP (grey bars) or AD‐AtbZIP proteins (black bars). In single transfection experiments, 9 μg of reporter and 28 μg of effector plasmid DNA was applied. In cotransfections, 14 μg of each bZIP effector plasmid were used. Variations in transformation efficiency are normalised by using 3 μg of a Pro35S:NAN plasmid construct per transfection experiment (Kirby and Kavanagh, 2002). Given are mean values and s.d.'s of four independent transfections as relative GUS/NAN units.

AtbZIP53 and AtbZIP10 cooperate in regulating hypo‐osmolarity induced ProDH transcription

The data described above suggest that a complex network of heterodimers is involved in regulation of ProDH. To get further insight in this regulatory mechanism, we focused on AtbZIP53 and AtbZIP10, the two partners that show the highest degree of ACTCAT‐mediated reporter gene activation. By means of ChIP analysis we could demonstrate that ProDH is a direct target of both, AtbZIP53 (Figure 2) as well as AtbZIP10 (Supplementary Figure S1). Furthermore, electrophoretic mobility shift assays (EMSA) confirmed that heterodimers can be formed on the ACTCAT motif in vitro (Supplementary Figure S2). A detailed experimental series was initiated in which the ratio of bZIP effector plasmids was altered. Figure 8A shows that the synergistic effect of AtbZIP53 and AtbZIP10 clearly depends on the ratio of the two heterodimerisation partners. Furthermore, the synergistic effect depends on the ZIP‐mediated heterodimerisation function. Disruption of the α‐helical structure in the AtbZIP53PP construct impairs activation of the reporter by AtbZIP53/AtbZIP10 (Figure 8B).

Figure 8.

The ACTCAT‐mediated transcription in Arabidopsis protoplasts depends on AtbZIP53/AtbZIP10 heterodimers. (A) Reporter activation depends on the ratio of AtbZIP53 and AtbZIP10. Activation of (ACTCAT)2:GUS reporter was measured using 14 μg of AtbZIP10 and increasing amounts of AtbZIP53 encoding plasmid (squares) or 14 μg of AtbZIP53 and increasing amounts of AtbZIP10 encoding plasmid (circles), respectively. Given are reporter gene induction values in %. Reporter gene activation after applying 14 μg of each effector is defined as 100%. (B) Activation of the (ACTCAT)2:GUS reporter by AtbZIP53/AtbZIP10 depends on a ZIP‐mediated heterodimerisation. ZIP‐specific exchanges of the amino acids leucine51 and alanine72 by proline (AtbZIP53PP) disrupts the α‐helical ZIP structure, bZIP heterodimerisation (see Figure 3B) and reporter gene activation. (C) Graphical description of the T‐DNA insertion mutant atbzip10. Exons are depicted as black bars, introns as white bars. The insertion was mapped inside the third exon. (D) Activation of (ACTCAT)2:GUS reporter by AtbZIP53, AtbZIP10 or both was compared in protoplasts derived from wild‐type (black bars) and atbzip53 atbzip10 double mutant plants (grey bars), respectively. For each bZIP encoding construct, 14 μg of DNA have been used. Given are relative GUS/NAN units.

To verify the impact of heterodimerisation on reporter gene expression, similar experiments as described above were performed in protoplasts obtained from atbzip53 atbzip10 plants. For efficient reporter gene expression, both AtbZIP53 and AtbZIP10 are required (Figure 8D). All together, these data strongly suggest that AtbZIP53/AtbZIP10 heterodimers are regulating ACTCAT‐mediated ProDH transcription.

Colocalisation of the bZIP partners in the same tissue is a prerequisite for heterodimerisation. Whereas transcription of AtbZIP53 is strongly induced by hypo‐osmolarity (Figure 1B), only slight induction can be obtained for AtbZIP10 (Supplementary Figure S3). Analysis of AtbZIP53 and AtbZIP10 promoter GUS reporter lines (ProAtbZIP53:GUS and ProAtbZIP10:GUS) revealed a partial overlap of expression patterns primarily in the meristematic tissues of the Arabidopsis plantlet. Strong enhancement of GUS activity after transfer from MS media to distilled water was found for both, ProAtbZIP53:GUS and ProAtbZIP10:GUS reporter lines (Figure 9). In contrast, ProAtbZIP25:GUS lines do not show this induction, indicating that this bZIP transcription factor is not involved in this response. Coexpression of AtbZIP53 and AtbZIP10 was also observed in parts of the flower and seeds (data not shown) in which a ProProDH:GUS construct is strongly expressed (Nakashima et al, 1998).

Figure 9.

Coexpression studies of AtbZIP53 and AtbZIP10. Activation of ProAtbZIP53:GUS (left), ProAtbZIP10:GUS (middle) and ProAtbZIP25:GUS (right) lines in 14 day‐old plantlets grown on MS media (upper panel) and after transfer to distilled water (hypo‐osmolarity treatment) for 12 h (lower panel).

In conclusion, these data demonstrate that AtbZIP53 and AtbZIP10 colocalise and interact with high affinity. Moreover, protein availability, which is regulated by environmental and developmental cues, restricts heterodimer formation and ultimately controls target gene transcription (Figure 10).

Figure 10.

Model describing the combinatorial control of ProDH transcription by the network of group‐S1/C heterodimers. (1) Developmental and tissue‐specific cues define expression of AtbZIP genes and consequently, the cellular amount of transcription factors. In particular, AtbZIP53, AtbZIP1, and AtbZIP2 are transcriptionally induced by environmental signals, for example, rehydration after osmotic stress (hypo‐osmolarity). As a consequence, colocalisation defines which heterodimers can be formed. (2) Specific group‐S1 and C heterodimers are formed with high affinity before DNA binding. Therefore, a pool of specific heterodimers can be postulated to exist in the nucleus. (3) Heterodimers select target genes, such as ProDH. Moreover, group‐S1/C heterodimers exhibit enhanced transactivation capacity (heterodimer‐induced transactivation, HIT), which is reflected by efficient recruitment of transcriptional cofactors and subunits of the polymerase II‐dependent transcription initiation complex (PolII). We therefore postulate that bZIP heterodimerisation acts as a crucial mechanism for signal integration on the level of transcription.


The amino acid L‐Pro is essential for basic metabolism as well as for plant development. Moreover, under stress conditions, it has been proposed to serve as a protective compound or as a signalling molecule (Delauney et al, 1993; Liu and Zhu, 1997). Consequently, final effects derived from its availability depend on L‐Pro homeostasis, a process tightly regulated by synthesis, degradation and transport. In this work we focus on the transcriptional control of the ProDH gene, encoding the rate‐limiting enzyme of L‐Pro degradation and found that it is directly regulated by the transcription factor AtbZIP53. We disclosed that AtbZIP53 action is modulated by other bZIP heterodimerising partners that integrate a fine regulatory network supported by combinatorial interactions.

The ProDH promoter ACTCAT motif is a direct target of AtbZIP53 transcription factors in vivo

Previously reported in vitro DNA‐binding studies (Satoh et al, 2004), alongside with the different experimental results presented in this work, including overexpression in transgenic plants, ChIP analysis, and activation of ACTCAT reporter constructs in protoplasts, provide conclusive evidence that the ProDH gene is a direct target of the bZIP transcription factor AtbZIP53. Loss‐of‐function approaches show, however, that although AtbZIP53 clearly mediates hypo‐osmolarity‐induced ProDH transcription, it is not the only factor involved in this process, as the elimination of AtbZIP53 activity does not result in complete lack of ProDH activation. In this respect, additional data obtained in transient expression assays in Arabidopsis protoplasts suggest that the closely related group‐S1 bZIP proteins might also participate in ProDH regulation (Satoh et al, 2004). In conclusion, a subset of five group‐S1 bZIP factors including AtbZIP53 seems to have overlapping, partly redundant functions. As single and double knockout lines (FW, KD and WDL, unpublished results) only partly reduce ProDH target gene expression, multiple ko lines are needed to assess their function.

HIT—a central mechanism to modulate activation properties of bZIP transcription factors in plants

By means of yeast two‐hybrid analysis, interaction studies in protoplasts and BiFC, we could establish that AtbZIP53 homodimerisation and heterodimerisation with other group‐S members is not favoured, but instead AtbZIP53 preferentially interacts with group‐C bZIP proteins. A similar heterodimerisation behaviour has already been described for homologous proteins from tobacco (Strathmann et al, 2001) and parsley (Rügner et al, 2001). We have also determined that this interaction relies on the ZIP structure, as it could be abolished by introducing amino‐acid exchanges that disrupt the α‐helical structure of the ZIP domain. Moreover, BiFC assays confirmed the occurrence of this preferred heterodimerisation in planta, as revealed by fluorescence due to complementation of two unfunctional parts of the YFP protein fused to AtbZIP53 and AtbZIP10. The cellular localisation of YFP fusions of AtbZIP53 and AtbZIP10 could also be assigned to the nucleus, in support of a role in transcription.

While the BiFC method provides information about the cellular localisation of an interaction, it does not give quantitative information about affinity of the protein partners. To address this question, we reproduced the principle of the yeast two‐hybrid system to Arabidopsis protoplasts. Basically, results obtained in the yeast system could be confirmed in planta. Moreover, this assay allows discriminating weak interactions inside the group‐S1 from strong interactions between groups S1 and C.

In the model plant Arabidopsis thaliana, 75 bZIP genes have been identified (Jakoby et al, 2002) and heterodimerisation has been experimentally shown for closely related bZIP proteins, like the GBFs (Schindler et al, 1992) or Hy5/HyH (Holm et al, 2002). Based on data obtained for dimerisation specificity of human bZIP proteins, a model for prediction of the Arabidopsis bZIP dimerisation has been proposed (Deppmann et al, 2004). However, group‐S1 and group‐C interactions, which have been confirmed experimentally, have not been predicted. Dimerisation of human bZIPs has been studied experimentally by means of protein‐CHIP techniques (Newman and Keating, 2003); however, bioinformatic studies using these experimental data conclude that no easy rules can be inferred for predictions (Fong et al, 2004). Hence, further experimental studies are needed to assess the complex heterodimerisation patterns in Arabidopsis. In general, group‐S1 and ‐C have been found to form specific heterodimers (Ehlert et al, 2006). As group‐S1 and ‐C share an unusual long ZIP domain harbouring eight repeats, we assume that the observed high‐affinity interaction is due to the intrinsic structural properties of this ZIP domain.

In vitro DNA‐binding experiments show that group‐S1 bZIP proteins can bind DNA without a group‐C partner. This finding could be explained by monomeric DNA binding which has been described before (Metallo and Schepartz, 1997; Cranz et al, 2004). Alternatively, structural analyses of bZIP proteins have demonstrated that by DNA binding the conformation of the basic domain renders into a coiled structure which is linear with the ZIP domain (Pu and Struhl, 1991). Hence, DNA–bZIP protein complex formation might alter dimerisation properties. Our experimental system makes use of the GAL4 DNA‐binding domain and hence, measured protein–protein interactions are independent of DNA‐binding activities. We therefore conclude that under in vivo conditions, group‐S1 and C ZIP domains show a strong affinity without need to be associated to its target promoter DNA. It is therefore tempting to speculate that pools of group‐S1/C heterodimers are formed in the nucleus, which has implications on transcriptional control of target genes (Figure 10).

The functional impact of bZIP heterodimerisation in vivo has rarely been addressed in planta. A recent publication deals with the rice bZIP factor LIP19/OBF1 in regulation of low‐temperature‐induced genes (Shimizu et al, 2005). In this case, heterodimerisation is postulated to modulate DNA‐binding specificity. This mechanism is related to the well‐described prototype in animals in which Fos homodimers cannot bind DNA but by heterodimerisation with Jun, another bZIP protein, the AP‐1‐binding site is recognised (Kouzarides and Ziff, 1988). Depending on the recruitment of bZIP heterodimerisation partners, members of the small MAF proteins can activate or repress transcription of their cognate target genes (Igarashi et al, 1994; Dlakic et al, 2001). MAF proteins are characterised by an extended DNA‐binding domain and therefore guide their heterodimerisation partners to promoter binding sites which are not recognised by the bZIP on its own. In contrast to these examples, specific heterodimers of AtbZIP53 lead to strong transactivation, although neither of the single transcription factors displays significant activation potential by itself. As we have observed a comparable synergistic enhancement of transcription in a system based on ACTCAT binding and a GAL‐UAS binding, this mechanism is independent of protein–DNA complex formation. These data indicate that formation of defined heterodimers generates new properties, which are relevant for gene activation under in vivo conditions. bZIP heterodimerisation might lead to the interaction with cofactors which cannot be recruited by homodimers. Alternatively, heterodimerisation might result in structural changes that might alter protein‐interaction surfaces and recruitment properties. In animal systems, Jun/Fos heterodimers have been found to promote interaction with general transcription factors (GTFs) like TFIIE‐34, TFIIE‐30 and TFIIF‐74 in vitro (Martin et al, 1996). Here, we provide conclusive evidence for the impact of a HIT mechanism in vivo. The isolation of protein‐interaction partners of the group‐S1/C heterodimers will give future insight into transcriptional control. In a comprehensive study, all group‐C and ‐S1 interactions have been analysed, indicating that HIT is not restricted to AtbZIP53/AtbZIP10 heterodimers but is a general feature of several group‐S1/C heterodimers. We therefore postulate that HIT is a general and important mechanism for fine tuning of transcription in plants.

The C/S1 network of bZIP transcription factors—signal integration on the level of transcription

A fine‐tuned mechanism can be anticipated by the disclosed participation of different bZIP factors in the regulation of common target genes. In particular, here we have studied the important contribution of distinct heterodimer combinations, whose concerted action will be ultimately dependent on common localisation, protein expression levels and functional capacity to interact.

Northern analysis described in this work and web‐based expression resources demonstrate that members of the C/S1 network are transcriptionally regulated by various developmental and environmental cues. As L‐Pro homeostasis has to be tightly regulated, the C/S1 network might serve to adjust gene transcription in response to a variety of endogenous and exogenous signals. In particular, AtbZIP53, AtbZIP1, and AtbZIP2 have been shown to be activated by hypo‐osmolarity (RARGE database: and promoter:GUS assays reveal that AtbZIP53 and AtbZIP10 are upregulated by hypo‐osmolarity. As heterodimerisation has been shown in vivo in the nucleus independent of DNA binding to target sequences, we assume that depending on the concentration of particular bZIP proteins, pools of homo‐ or heterodimers can be formed. These dimers compete for promoter target sites such as the ACTCAT motif in order to activate or repress transcription (Figure 10).

In the case of AtbZIP53 and AtbZIP10, it has been shown that the formation of heterodimers depends on the concentration of both partners. In addition, the amount of functionally active bZIP protein in the nucleus is regulated by various means. For instance, post‐transcriptional mechanisms, like sucrose‐induced repression of translation (SIRT) (Wiese et al, 2004) have been described for bZIP11/ATB2 and other S1‐like bZIP genes. Nuclear translocation of AtbZIP10 has been shown to be controlled (KH, unpublished results) by the protein‐interaction partner LSD1 (Dietrich et al, 1997). In summary, availability of bZIP heterodimerisation partners is regulated in a highly complex manner by transcriptional and post‐transcriptional mechanisms.

The biological function of the members of the C/S1 network is not well established. Recent findings provide evidence that AtbZIP10 is involved in oxidative stress response and LSD1‐mediated programmed cell death (KH, unpublished results). Accordingly, proline metabolism has been postulated to play a crucial role in cell death signalling (Deuschle et al, 2004) and very recently, a function in pathogen‐induced hypersensitive response has been demonstrated (Fabro et al, 2004). We therefore assume that the C/S1 network is not only involved in recovery after osmotic stress but also represents a general regulatory circuit in protection against oxidative stress. Owing to the well‐characterised promoter elements, the growing knowledge on trans acting factors and the physiological role of proline in cell‐death, ProDH will develop as a valuable marker gene for oxidative stress responses.

Group‐C bZIP factors are highly related to well‐characterised bZIP proteins involved in seed‐storage protein synthesis, such as O2 in maize (Schmidt et al, 1992). In Arabidopsis, AtbZIP10 and AtbZIP25 have been implicated in regulation of seed‐storage protein genes (Lara et al, 2003). ProDH transcription is also highly controlled during seed development (Nakashima et al, 1998). Hence, we assume that besides gene regulation in oxidative stress response, the C/S1 network is involved in developmentally regulated gene control.

A characteristic feature of the C/S1 network is overlapping functional redundancy. Single ko in group‐S1 or ‐C could not completely abolish ProDH gene activation. Heterodimerisation of group‐S1 and ‐C bZIP factors provide an efficient mechanism to integrate signals and to adjust target gene transcription. A functional study of the bZIP proteins, the identification of target genes and a detailed expression and heterodimerisation analysis is needed to further elucidate the impact of the C/S1 network on gene regulation in higher plants.

Materials and methods

Plant material, plant treatment, plant transformation

For protein isolation, protoplast preparation and ChIP, A. thaliana ecotype Columbia (Col‐O) was grown on soil under controlled environmental conditions at 16 h light/8 h dark cycles. To improve germination uniformity, plants were pretreated at 4°C for 2–4 days. Hypo‐osmolarity treatment was performed according to Satoh et al (2002). Floral dip transformations were performed by using Agrobacterium tumefaciens strains GVG3101 and wild‐type A. thaliana (ecotype Col‐0) (Weigel and Glazebrook, 2002).

Identification of T‐DNA mutants

Seeds of the T‐DNA insertion mutants atbzip10 (NASC ID: N606031) and atbzip53 (NASC ID: N569883) were obtained from the Nottingham Arabidopsis Stock Centre. Homozygous mutants were identified by PCR as described in ( AtbZIP53‐specific primers were 5′‐CTGCAGCTTAGGACAGCTCATCACCA‐3′ (AtbZIP53pGUSf) and 5′‐GTCGACCTCGTTGACTTTTTGACTTC‐3′ (AtbZIP53pGUSr), AtbZIP10‐specific primers were 5′‐TCTTCTCCATTGACGATTTCTC‐3′ (AtbZIP10for) and 5′‐TGATCTTCCGAGAAGCATCGG‐3′ (AtbZIP10rev), the T‐DNA‐specific primer was 5′‐TGGTTCACGTAGTGGGCCATCG‐3′ (Lba1). Homozygous atbzip10 and atbzip53 T‐DNA insertion mutants were crossed and in the f2‐generation double mutants were identified by PCR.

Molecular biological techniques

Standard DNA techniques have been described in Sambrook et al (1989). Plant RNA was isolated using TRIZOL Reagent (Life Technologies, Rockville, MD, USA), Northern analysis has been described in Heinekamp et al (2002). Hybridisation probes were produced by PCR amplification from cDNA using specifically designed primers. ProDH: 5′‐TGGTCGGTCTTTGTCAGCAT‐3′ and 5′‐ATGGCAACCCGTCTTCTCCG‐3′; AtbZIP53: 5′‐CATACAAAGACACGAACAAATTGAA‐3′ and 5′‐GTGCTTGAGATGGTTGAAGAAATTA‐3′; actin8: 5′‐GCTGGATTCGCTGGAGATGA‐3′ and 5′‐AGGTCTCCATCTCTTGCTCG‐3′, respectively. A hybridisation probe for AtbZIP10 was obtained by digestion of pDONR201‐AtbZIP10 with Bsp1407I. Western analysis was described in Heinekamp et al (2002) using a polyclonal HA‐antibody (Santa Cruz, Santa Cruz, CA, USA). Vector constructions are described in detail in Supplementary methods S1. ChIP was performed according to Johnson et al (2001) with modifications described in Supplementary methods S1.

Yeast two‐hybrid analysis

ProQuest® Y2H vectors pDEST22 and pDEST32 containing the different AtbZIP genes were transformed into the yeast strain MaV203 (Invitrogen, Karlsruhe, Germany) according to the protocol of Dohmen et al (1991). Handling of yeast cultures, plate growth assays and quantitative β‐galactosidase assays were performed as described in the Clontech Yeast Protocols Handbook (1996).

Protoplast transformation and GUS assays and BIFC

Protoplast isolation and transformation was performed as described by Ehlert et al (2006). BiFC was performed in 6‐week‐old Nicotiana benthamiana plants after Agrobacterium‐mediated transient transformation according to Walter et al (2004). For analysis of the BiFC signal, a confocal microscope (Leica TCS SP2 AOBS) was applied.

Electrophoretic mobility shift assays

EMSAs were performed as described by Lara et al (2003). The oligonucleotides PRODHS (5′‐ATTCATCATCC ACTCAT CCT‐3′) and PRODHAS (5′‐AGATATGAAGG ATGAGT GGA‐3′) were used as dsDNA probe.

Supplementary data

Supplementary data are available at The EMBO Journal Online.

Supplementary Information

Supplementary Table S1 [emboj7601206-sup-0001.pdf]

Supplementary Figure S1 [emboj7601206-sup-0002.pdf]

Supplementary Figure S2 [emboj7601206-sup-0003.pdf]

Supplementary Figure S3 [emboj7601206-sup-0004.pdf]

Supplementary methods S1 [emboj7601206-sup-0005.doc]


We thank C Gatz (University of Göttingen, Germany) for constant support and for critical reading of the manuscript and T Butterbrodt (University of Göttingen, Germany) for assistance during the development of ChIP. We are grateful to B Weisshaar (University of Bielefeld, Germany), S Smeekens (University of Utrecht, The Netherlands) and T Kavanagh (University of Dublin, Ireland) for supplying plasmid constructs. This research was supported, in part, by grants from the DFG to FW (GRK242) and to KH (SFB446) and from European Community to XW (REGIA).


View Abstract