The Gene for the P-subunit of Glycine Decarboxylase from the C 4 Species Flaveria trinervia : Analysis of Transcriptional Control in Transgenic Flaveria bidentis (C 4 ) and Arabidopsis thaliana (C 3 )

Glycine decarboxylase (GDC) plays an important role in the photorespiratory metabolism of plants. GDC is composed of four subunits (P, H, L and T) with the P-subunit (GLDP) serving as the actual decarboxylating unit. In C 3 plants, GDC can be found in all photosynthetic cells whereas in leaves of C 3 -C 4 intermediate and C 4 species its occurrence is restricted to bundle-sheath cells. The specific expression of GLDP in the bundle-sheath cells might have constituted a biochemical starting point for the evolution of C 4 photosynthesis. To understand the molecular mechanisms responsible for restricting GLDP expression to bundle-sheath cells, we performed a functional analysis of the GLDPA promoter from the C 4 species Flaveria trinervia . Expression of a promoter-reporter gene fusion in transgenic plants of the transformable C 4 species Flaveria bidentis (C 4 ) showed that 1571 bp of the GLDPA 5´ flanking region contain all the necessary information for the specific expression in bundle-sheath cells and vascular bundles. Interestingly, we found that the GLDPA promoter of F. trinervia exhibits a C 4 -like spatial activity also in the C 3 plant Arabidopsis thaliana , indicating that a mechanism for bundle-sheath-specific expression is also present in this C 3 species. Using transgenic Arabidopsis, promoter deletion studies identified two regions in the GLDPA promoter, one conferring repression of gene expression in mesophyll cells and one functioning as a general transcriptional enhancer. Subsequent analyses in transgenic F. bidentis confirmed that these two segments fulfill the same function also in the C 4 context.


Introduction
Net photosynthetic CO 2 assimilation rates in C 3 plants are reduced by photorespiration, a process that results from the oxygenase activity of ribulose-1,5-bisphosphate carboxylase/oxygenase (Rubisco). C 4 plants usually show no apparent photorespiration, and this is achieved by splitting the photosynthetic reactions between two morphologically and biochemically distinct cell types, the mesophyll and the bundle-sheath cells. Initial CO 2 fixation in C 4 plants occurs exclusively in the mesophyll cells and is performed by the enzyme phosphoenolpyruvate carboxylase (PEPC) to form a C 4 acid, oxaloacetate. Depending on the C 4 subtype, oxaloacetate is converted to either malate or aspartate, which subsequently move to the bundle-sheath and become decarboxylated, resulting in significant elevation of the CO 2 concentration in these cells. Final refixation of CO 2 is achieved by Rubisco, which, in C 4 plants, is only present in bundle-sheath cells. The enrichment of CO 2 in the vicinity of Rubisco effectively inhibits the enzyme´s oxygenase activity (Hatch, 1987).
In C 4 plants, the CO 2 assimilatory enzymes are compartmentalized into either mesophyll or bundle-sheath cells, and this is governed by differential gene expression. It has been shown that mesophyll-specific expression of C 4 cycle genes is mainly regulated at the transcriptional level (Schaffner and Sheen, 1992;Stockhaus et al., 1994;Rosche et al., 1998;Sheen, 1999), whereas bundle-sheath-specific expression is controlled at both transcriptional and post-transcriptional levels (Long and Berry, 1996;Marshall et al., 1997;Sheen, 1999;Patel et al., 2006).
There are indications that photorespiration also exists in C 4 plants, albeit at a much lower level than in C 3 plants (Osmond and Harris, 1971;Furbank and Badger, 1983;de Veau and Burris, 1989). This is likely due to the fact that photorespiration in C 4 species is strictly confined to the bundle-sheath cells in the leaves, while in C 3 plants photorespiration occurs in all photosynthetically active mesophyll cells (Ohnishi and Kanai, 1983).
The mitochondrial multienzyme complex glycine decarboxylase (GDC) plays a key role in the photorespiratory pathway. GDC is composed of four different subunits (P, H, T and L) and catalyzes, in cooperation with serine hydroxymethyltransferase (SHMT), the oxidative decarboxylation of glycine that originates from the breakdown of photorespiratory phosphoglycolate. In the course of these reactions, two molecules of glycine are converted to one molecule each of serine, NH 3 and CO 2 (Neuburger et al., 1986). Consistent with the compartmentation of the photorespiratory cycle, GDC is present in all photosynthetic cells of C 3 plant leaves, but strictly confined to the bundle-sheath cells in C 4 species (Ohnishi and  Kanai, 1983). None of the GDC subunits have been detected in the mesophyll cells of C 4 plants ( Morgan et al., 1993).
Plant species possessing a C 3 -C 4 intermediate type of photosynthesis are of special interest for studying the evolution of C 4 -characteristic traits. Some C 3 -C 4 plants are to some extent able to fix CO 2 into malate and aspartate (Monson et al., 1986), but most of these species do not have a C 4 metabolism. They can be characterized as "intermediate", for example, by their CO 2 compensation points, which are lower than those of C 3 plants, but higher than those of C 4 plants (Edwards and Ku, 1987;Rawsthorne, 1992). As in C 4 plants, functional GDC occurs only in the bundle-sheath cells of the leaves of C 3 -C 4 intermediate plants, as it was first demonstrated for Moricandia arvensis (Rawsthorne et al., 1988a(Rawsthorne et al., , 1988b. Later on, it was discovered that the loss of GDC activity in the mesophyll is due to a lack of the P-subunit (GLDP). Because of the absence of GDC activity in mesophyll cells of M. arvensis leaves, photorespiratory glycine moves to the bundle-sheath cells to be processed by GDC. The bundle-sheath cells of M. arvensis contain a large number of mitochondria which are arranged at the centripetal cell wall adjacent to the vascular tissue, whereas the chloroplasts are located at the cell periphery. This special distribution of organelles and the restriction of glycine oxidation to the bundle-sheath compartment result in an efficient recapture of released photorespiratory CO 2 , thereby lowering the CO 2 compensation point when compared to a typical C 3 plant (Rawsthorne et al., 1998). C 3 -C 4 intermediate species are thought to represent a stage in the evolutionary transition from C 3 to C 4 photosynthesis (Edwards and Ku, 1987). It was therefore tempting to speculate that the confinement of GDC to the bundle-sheath cells has been one of the biochemical starting points for the evolution of C 4 plants ( Morgan et al., 1993;Bauwe and Kolukisaoglu, 2003;Sage, 2004). However, the possible effects of this relocation for C 4 evolution are discussed controversially (Edwards et al., 2001). The loss of the P-subunit seems to be the initial step to inactivate GDC in the mesophyll, and the absence of all GDC subunits in the leaf mesophyll of other C 3 -C 4 intermediate species suggests that they have developed further towards C 4 photosynthesis than M. arvensis (Morgan et al., 1993).
A well established experimental system for investigating the evolution of C 4 characteristic traits is the genus Flaveria of the Asteraceae (Powell, 1978). This small genus comprising 23 known species includes both C 3 and C 4 species, but also a large number of C 3 -C 4 intermediate species (Edwards and Ku, 1987). In this study we examined the promoter of the gene encoding the P-subunit of GDC from the C 4 species Flaveria trinervia to gain insight into the molecular basis of bundle-sheath-specific gene expression. Two genes encoding the P-subunit of GDC have been identified in F. trinervia, GLDPA and the pseudogene GLDPB (Cossu, 1997). The GLDPA promoter was fused to a β-glucuronidase (GUS) reporter gene and promoter activity was analyzed in transgenic Flaveria bidentis (C 4 ) and Arabidopsis thaliana (C 3 ). Similar expression patterns were observed in these two species, which allowed the use of A. thaliana as a heterologous system for testing a series of promoter deletions to identify C 4 -characteristic regulatory elements within the GLDPA promoter. These analyses resulted in the identification of regions within the GLDPA promoter that contribute mainly to the regulation of expression quantity or to the spatial expression pattern of the GLDPA gene, respectively.

In Situ RNA Hybridization
Immunolabeling studies have shown that, in C 4 plants, the P-subunit of GDC (GLDP) accumulates exclusively in bundle-sheath cells of leaves (Hylton et al., 1988;Morgan et al., 1993;Yoshimura et al., 2004). In order to examine whether this C 4 -characteristic localization of the P-protein is due to specific accumulation of GLDPA mRNA in this compartment we analyzed the expression pattern of the GLDPA gene in leaves of the C 4 species F. trinervia by in situ hybridization. As a probe we used a 2,4 kb fragment of the GLDPA cDNA from F. trinervia, and control hybridizations were performed with the corresponding sense probe.
In leaves of F. trinervia, transcripts of the GLDPA gene could only be detected in bundle-sheath and not in mesophyll cells (Fig. 1A). The GLDPA mRNA accumulated near the centripetal cell walls of the bundle-sheath cells, which was due to the concentration of cytoplasm in this region. The confinement of the P-protein to the bundle-sheath cells therefore is controlled by the specific accumulation of GLDPA mRNA in this compartment. The same result was obtained by in situ hybridization of the GLDPA probe to leaf cross-sections of the C 4 -species F. bidentis (Fig. 1C).

Expression of a GUS Reporter Gene under the Control of the GLDPA Promoter from F. trinervia in transgenic F. bidentis
The in situ RNA hybridization analysis showed that the occurrence of GLDPA transcripts is restricted to the bundle-sheath cells in F. trinervia and F. bidentis (Fig. 1). To test whether the available 1571 bp of the 5´ flanking region of the GLDPA gene (including the 5´ untranslated region upstream of the AUG start codon) harbour all the necessary information for this 8 bundle-sheath-specific expression pattern, we fused this region to a β-glucuronidase (GUS) reporter gene (construct GLDPA-Ft, Fig. 2A) and examined its expression behaviour in transgenic F. bidentis plants. The C 4 -species F. bidentis is a close relative to F. trinervia, but unlike F. trinervia it is suitable for transformation by Agrobacterium tumefaciens mediated gene transfer (Chitty et al., 1994).
Histochemical analysis of the expression of the GLDPA-Ft promoter-GUS construct revealed an intense blue staining in bundle-sheath cells, but not in the mesophyll cells (Fig.   2B). GUS activity could also be observed in most vascular bundles, with the degree of GUS expression varying with the size of the veins. The small minor veins usually exhibited a strong blue staining, while higher-order vascular bundles showed only moderate GUS activity. Additional weak GLDPA promoter activity was also detected in the guard cells of the stomatal complexes (Fig. 2C).

F. bidentis
Bundle-sheath cells are not a unique feature of C 4 plants. They are also present in many C 3 plants, but compared to the situation in C 4 species these cells exhibit fewer chloroplasts and mitochondria (Kinsman and Pyke, 1998;Leegood, 2002). To examine whether the GLDPA promoter of F. trinervia shows a cell-specific activity in a C 3 background we introduced the GLDPA-GUS construct into A. thaliana.
The histochemical analysis revealed GUS expression in the vascular tissue and in the surrounding bundle-sheath cells (Fig. 3C, D). Notably, very similar to the expression pattern in F. bidentis, no GUS activity could be detected in the mesophyll cells of transgenic Arabidopsis plants. The quantification of GUS levels showed that the median activity of the reporter protein was comparable in Arabidopsis and F. bidentis leaves (Fig. 2D, 3B).
To verify the results obtained from the histochemical GUS analysis, the GLDPA promoter was also fused to the green fluorescent protein (GFP) reporter gene mgfp5-ER (Siemering et al., 1996;Haseloff et al., 1997) and a Histone 2B/ yellow fluorescent protein (H2B-YFP) fusion gene (Boisnard-Lorig et al., 2001), which are targeted to the endoplasmic reticulum and the nucleus, respectively. These reporter proteins allow a non-destructive analysis by fluorescence or confocal laser microscopy, thereby avoiding any potential diffusion of the reporter protein which might occur during a histochemical staining procedure.
In both cases, the reporter proteins could be detected in bundle-sheath cells and vascular bundles, but not in mesophyll cells ( Fig. 3E-L, supplemental data S1, S2). 9

Deletion Analysis of the GLDPA Promoter
The "C 4 -like" expression pattern of GLDPA-GUS in transgenic Arabidopsis provided the opportunity to functionally dissect the GLDPA promoter by using this C 3 model organism as an experimental system. To identify cis-regulatory determinants that are responsible for the activity of the GLDPA promoter in bundle-sheath cells and the vascular bundle, we produced a set of 5´ deletions and analyzed their expression specificity and level in transgenic Arabidopsis. The GLDPA promoter was subdivided into seven fragments that are referred to as region 1 to region 7 in the following (Fig. 4A).
The removal of a 182 bp segment (region 1) from the 5´ end of the GLDPA promoter, of region 3 results in a further reduction of promoter activity which impeded further analysis of cell type-specific expression within the leaf.

Regions 1 and 2 of the GLDPA Promoter Together Function as a General Enhancer of Transcription in the Arabidopsis Leaf
To investigate whether the GLDPA promoter fragment reaching from -1571 to -1139 (regions 1 and 2) was able to act as a transcriptional enhancer we combined this segment of the promoter with region 7 of the GLDPA promoter (-298 to -1). Region 7 harbours a putative TATA-box and the starting point of transcription, but -as reported above -this part of the promoter alone is not sufficient to drive GUS expression in the Arabidopsis leaf ( To test whether regions 1 and 2 of the GLDPA promoter function also in a heterologous promoter context we fused this segment in front of the proximal promoter region of the ppcA1 gene of F. trinervia (Fig. 5A). The ppcA1 gene encodes the C 4 isoform of phosphoenolpyruvate carboxylase (Hermans and Westhoff, 1992), and its complete 2188 bp promoter directs high and mesophyll-specific GUS expression in transgenic F. bidentis (Stockhaus et al., 1997). In contrast, the activity of the 570 bp long proximal ppcA promoter part (ppcA-PR Ft ) is extremely low and can hardly be visualized in histochemical GUS assays (Gowik et al., 2004). In plants in which the low activity permits a histological analysis, the ppcA-PR Ft promoter fragment directs a uniform expression in all cells of the leaves of F. bidentis, including the vascular bundles (Akyildiz et al., 2007).
The fusion of regions 1 and 2 of the GLDPA promoter with the ppcA-PR Ft promoter fragment resulted in strong GUS expression in leaves of Arabidopsis (Fig. 5B). The GUS reporter gene was active in both mesophyll and bundle-sheath cells as well as in the vascular bundles (Fig. 5D), and the expression profile of this chimeric promoter is thus indistinguishable from that of GLDPA-Ft-1-2-7. We conclude from these experiments that regions 1 and 2 of the GLDPA promoter constitute a general transcriptional enhancer module that, in combination with a basal promoter, stimulates the expression of a linked reporter gene in all types of interior leaf cells of Arabidopsis.

Region 3 of the GLDPA Promoter Acts as a Mesophyll-Specific Repressor of Gene Expression
The role of region 3 (-1138 to -927) in regulating GLDPA promoter activity was investigated by introducing the relevant promoter fragment into construct GLDPA-Ft-1-2-7, resulting in the production of the chimeric promoter GLDPA-Ft-1-2-3-7 (Fig. 6A). The addition of region 3 to GLDPA-Ft-1-2-7 caused a significant change in the spatial expression pattern of the GUS reporter gene. While GLDPA-Ft-1-2-7 plants expressed the GUS reporter gene in mesophyll and bundle-sheath cells as well as in the vascular tissue (Fig. 5C), GUS activity of GLDPA-Ft-1-2-3-7 plants was strictly confined to the bundle-sheath cells and the vascular compartment (Fig. 6B). These observations indicate that region 3 of the GLDPA promoter from F. trinervia functions as a mesophyll-specific repressor of gene expression in the Arabidopsis leaf.

Analysis of GLDPA Promoter Regions 4, 5 and 6
We have shown that the GLDPA promoter fragment comprising base pairs -1571 to -927 (regions 1 to 3) in combination with the most proximal promoter part (region 7) is sufficient to direct GUS expression in the bundle-sheath cells and vascular bundles of transgenic A. thaliana plants. Nevertheless, additional cis-regulatory determinants that could be involved in the spatial regulation of transcriptional activity might also be present in promoter regions 4, 5 and 6. In order to investigate the occurrence of cis-regulatory elements within these promoter regions it was necessary to raise the GUS expression levels of constructs GLDPA-Ft-∆3 to GLDPA-Ft-∆6 to a level that allowed a histochemical analysis. Since regions 1 and 2 of the GLDPA promoter contain a general transcriptional enhancer with no apparent leaf cell specificity we attached this GLDPA transcriptional enhancer module to the 5´ borders of the truncated promoters (Fig. 7A).
As expected, the transcriptional activity of these constructs was dramatically higher than that of their "enhancerless" counterparts and was therefore suitable for performing GUSstainings in situ (Fig. 4D, 7B). In construct GLDPA-Ft-1-2-4-5-6-7, only region 3 was which regions 4 and 5 were further deleted (Fig. 7D, E, G, H). In contrast, as already reported above, the additional deletion of region 6 in construct GLDPA-Ft-1-2-7 resulted in a uniform expression pattern in the leaf (Fig. 5C). These findings suggest that additional cis-regulatory elements conferring repression of gene expression in the mesophyll are located in region 6 of the GLDPA promoter. However, when compared to the highly effective repressor elements located in region 3, these additional elements in region 6 do not provide robust repression. GLDPA-Ft-1-2-3-7 in Transgenic F. bidentis A truncated promoter containing the transcription enhancing regions 1 and 2, the mesophyll repressor region 3 and the basal expression segment 7 generated the same spatial expression profile in the leaf of the C 3 plant Arabidopsis as the complete GLDPA promoter, i.e. the promoter regions 4, 5 and 6 (-926 to -299) were not essential for creating the C 4characteristic spatial expression pattern of a reporter gene. We now wished to examine whether this chimeric GLDPA-Ft-1-2-3-7 promoter is capable of providing this C4 expression profile also in the C 4 background of F. bidentis. This chimeric promoter construct was therefore transformed into F. bidentis, and its expression was examined in the leaves of the transgenic plants (Fig. 8).

Analysis of Promoter Construct
No differences between the spatial expression patterns of GLDPA-Ft-1-2-3-7 and the full-length promoter construct GLDPA-Ft were observed (compare Figs. 2 and 8). In both cases, GUS expression was found exclusively in the bundle-sheath cells and -with variable intensities -in the vascular strands. While GUS staining was strong in some minor veins, it was absent from other minor and all major vascular strands (Fig. 8B, C). These results indicate that regions 1 to 3 in combination with the basal TATA box-containing segment 7 of the GLDPA promoter are sufficient to direct reporter gene expression in bundle-sheath cells and the vascular bundles of both the homologous C 4 species F. bidentis and the heterologous C 3 plant A. thaliana.

Discussion
The correct functioning of the C 4 photosynthetic cycle requires strict compartmentalization of C 4 enzymes in either mesophyll or bundle-sheath cells of the leaf. This cell type-specific accumulation of proteins is governed by differential gene expression (Sheen, 1999). To broaden our knowledge on the molecular basis of bundle-sheath specific gene expression in 13 gene from F. trinervia (C 4 ). The GLDPA gene encodes the P-subunit (GLDP) of glycine decarboxylase (GDC), which is specifically located in the bundle-sheath cells of C 4 species (Morgan et al., 1993). To determine whether the bundle-sheath-specific accumulation of the GLDP protein in the C 4 leaf is paralleled by the accumulation of the corresponding mRNA we studied the occurrence of GLDPA transcripts within the leaf of F. trinervia and F. bidentis by in situ hybridizations. GLDPA RNA was exclusively found in the bundle-sheath cells of both C 4 species, indicating that the presence of this protein in bundle-sheath cells, and its absence in mesophyll cells, is caused by differential GLDPA mRNA accumulation.
We then investigated whether the available 1571 bp of the 5´ flanking region of the Alternatively, the absence of GLDPA mRNA in the vascular tissue might be caused by low GLDPA RNA stability in this tissue (Parker and Song, 2004;Moore, 2005). While the 5´ untranslated region of the GLDPA mRNA is included in the GLDPA promoter construct, the 3´ untranslated segment is not. If this latter RNA segment or the coding region contain stability determinants leading to degradation of the GLDPA mRNA in the vascular tissue, the GUS mRNA lacking these segments would accumulate.
Interestingly, expression of the GLDPA promoter-GUS construct in transgenic Arabidopsis showed a spatial activity of the GLDPA promoter that was very similar to that observed in transgenic F. bidentis (Fig. 3). As found for the C 4 species, the GLDPA promoter of F. trinervia was inactive in the leaf mesophyll of the C 3 plant A. thaliana, but active in the 14 bundle-sheath cells and the vascular bundle. The level of reporter gene expression was similar in both species. We therefore conclude that the C 4 -characteristic cis-regulatory transcriptional determinants are recognized in the same spatial manner also in the C 3 context. This indicates that the transcription factors necessary for the correct interpretation of these cis-regulatory sequences are already present in this C 3 species and are operating in the same spatial pattern.
Moreover, the similar spatial expression profiles in a C 4 and C 3 leaf allow to conclude that the gene regulatory networks operating in mesophyll and bundle-sheath cells of dicot C 3 and C 4 species share common elements, as it was previously proposed by Matsuoka et al. (1993).
The exact physiological and biochemical functions of bundle-sheath cells in C 3 species are poorly understood. They are involved in phloem loading and unloading (van Bel, 1993), and for tobacco it was shown that the bundle-sheath cells of stems and petioles exhibit high activities of enzymes characteristic of C 4 photosynthesis, thus allowing the decarboxylation of four-carbon organic acids derived from the xylem and phloem (Hibberd and Quick, 2002). Additionally, a class of Arabidopsis mutants termed dov (differential development of vascular associated cells) demonstrates that differential chloroplast development occurs between bundle-sheath and mesophyll cells in the Arabidopsis leaf (Kinsman and Pyke, 1998).
These observations from tobacco and Arabidopsis provide some evidence that bundlesheath cells in C 3 plants are somehow predetermined to evolve C 4 characteristic features. The special physiology of bundle-sheath cells in A. thaliana and the fact that pre-existing transcription factors in this C 3 species are able to recognize heterologous C 4 characteristic cisregulatory elements in the correct fashion provide further evidence for the view that the evolution of C 4 plants must have been relatively simple in genetic terms (Westhoff and Gowik, 2004). C 4 -like spatial activities of C 4 promoters in transgenic C 3 plants have also been reported for the C 4 isoform of PEPC of maize (Matsuoka et al., 1994;Nomura et al., 2000), the pyruvate orthophosphate dikinase gene of maize (Matsuoka et al., 1993) and the phosphoenolpyruvate carboxykinase of Zoysia japonica (Nomura et al., 2005). On the other hand, the C 4 -PEPC promoter of F. trinervia loses mesophyll specificity when it is introduced in Arabidopsis ( Akyildiz et al., 2007). Similarly, the NADP-dependent malic enzyme promoter from maize loses its bundle-sheath specificity in rice (Nomura et al., 2005). This shows that the functionality of C 4 -specific regulatory cis-elements in C 3 plants cannot be generalized.
The series of GLDPA promoter deletion and recombination constructs were analyzed and two major functional modules were identified and localized, a non-cell-type-specific transcriptional enhancer and a segment that represses gene expression in mesophyll cells.
The transcriptional enhancer is located within the outermost distal regions 1 and 2 of the GLDPA promoter comprising base pairs -1571 to -1139. The enhancer functioned in all interior leaf tissues of Arabidopsis, i.e. in mesophyll and bundle-sheath cells as well as in the vascular bundle. The transcription-enhancing activity of these regions was not restricted to the context of the GLDPA promoter but was also functional when combined with the proximal part of the ppcA1 promoter of F. trinervia. The enhancer is thus not GLDPA gene-specific but functions as a general enhancer module.
The quantitative analysis of promoter activities (Fig. 4D) indicates that region 1 has a higher potential for transcriptional enhancement than region 2. A search for known cisregulatory elements (Prestridge, 1991;Higo et al., 1999) revealed the presence of a motif with similarity to the simian virus 40 enhancer core (GTGGWWHG) at positions -1455 to -1448 in region 1. This motif is also present in a region of the GLDPA promoter of F. pringlei that is associated with an increase in expression quantity (Bauwe et al., 1995).
Region 3 (-1138 to -927) harbours cis-regulatory elements that confer cell specificity to the GLDPA promoter by repressing its activity in the mesophyll cells of the Arabidopsis leaf. A chimeric promoter consisting of the transcription-enhancing regions 1 and 2, region 3, and the proximal basal expression region 7 is also not active in the mesophyll cells of the C 4 species F. bidentis. This indicates that region 3 can repress expression in mesophyll cells also in the C 4 context, i.e. the mesophyll-repressing function of region 3 is conserved between the C 3 and the C 4 species. The lack of GLDPA expression in the mesophyll is thus caused by transcriptional regulation and not by post-transcriptional regulation as it was reported for the FbRbcS1 gene of F. bidentis. FbRbcS1 encodes the small subunit of Rubisco, and its bundlesheath-specific expression is entirely established by selective rbcS transcript stabilization in the bundle-sheath cells (Patel et al., 2006).
Additional mesophyll-repressing cis-regulatory sequences are located in region 6 (-521 to -299). They can partially compensate for the lack of the mesophyll-repressing cisregulatory sequences in region 3, when this segment is not present in the promoter construct.
However, these cis-regulatory elements are not able to establish a robust repression of reporter gene activity in the mesophyll cells of Arabidopsis and appear to be of minor importance. This is documented by the cell type-specific expression of construct GLDPA-Ft-1-2-3-7 that consists of the transcription-enhancing regions 1 and 2, region 3, and the

16
proximal basal promoter region 7. GLDPA-Ft-1-2-3-7 directs a C 4 -characteristic expression profile in F. bidentis. This demonstrates that the cis-regulatory motives present in region 6 are not necessary for the repression of GLDPA expression in mesophyll cells of this C 4 species.
Moreover, we can infer from the expression profile of this truncated promoter that regions 4 and 5 are also not necessary to achieve a C 4 -typical GUS expression pattern in both Arabidopsis and F. bidentis. Regarding the mechanism of mesophyll repression no predictions can be made at this moment, since searching for known cis-regulatory elements in region 3 did not identify any robust candidate motifs.
The cis-regulatory determinants for the mesophyll-specific repression of GLDPA expression in the leaf have not been determined yet, and no other gene system for bundlesheath specific expression has been investigated in such a detail that its cis-and transregulatory elements are known. However, cis-regulatory elements for mesophyll-specific gene expression, have recently been identified at the nucleotide level (Gowik et al., 2004;Akyildiz et al., 2007) for the C 4 PEPC gene ppcA1 from F. trinervia. It was found that variation at two positions in a 41 bp element, a G to A transition and the presence or absence of the tetranucleotide CACT, determines whether the promoter is active in all interior tissues of the leaf of transgenic F. bidentis or only in the mesophyll cells (Akyildiz et al., 2007).
Hence, for both genes, GLDPA and ppcA1, cell type-specific gene expression is achieved by repressing the activity of the gene in the respective other cell type. Further dissection of repressor regions 3 and 6 within the GLDPA promoter of F. trinervia should allow to precisely map the involved cis-regulatory elements. Such identification of a bundle-sheathspecific expression module consisting of a cis-regulatory sequence and the corresponding transcription factor would mark an important step in our understanding of the evolution of C 4 photosynthesis. Moreover, this information would also be of great value for attempts to convert a C 3 species into a C 4 species, as it has been proposed for rice (Mitchell and Sheehy, 2006).

Construction of Chimeric Promoters
DNA manipulations and cloning were carried out according to Sambrook and Russell (2001).
Construct GLDPA-Ft (Cossu, 1997) served as the basis for the series of GLDPA promoter deletions. In GLDPA-Ft, the 5´ upstream region of the GLDPA gene of F. trinervia from -cDNA in the binary plant transformation vector pBI121 (Clontech Laboratories, Palo Alto, CA). Different 5´-deleted fragments of the GLDPA-promoter were generated by PCR amplification (Tables 1 and 2). The primers added an XbaI restriction site at the 5´ border of the DNA fragments and an XmaI site at the 3´ end. Therefore the deleted promoters could be inserted into XbaI/XmaI-cut pBI121, resulting in the formation of the constructs GLDPA-Ft- XbaI sites were introduced at both ends of the PCR product which allowed the insertion of the DNA fragment into XbaI-cut GLDPA-Ft-∆3, GLDPA-Ft-∆4, GLDPA-Ft-∆5, GLDPA-Ft-∆6 and ppcA-S-Ft (Stockhaus et al., 1994). The resulting plasmid constructs were named  To produce construct GLDPA-Ft::H2B:YFP, the H2B:YFP gene fusion was excised from plasmid pBI121-35S::H2B:YFP (Boisnard-Lorig et al., 2001) with BamHI and SacI.

Plant Transformation
In all transformation experiments the Agrobacterium tumefaciens strain AGL1 was used (Lazo et al., 1991). The promoter-GUS constructs were introduced into AGL1 by electroporation. Arabidopsis plants were transformed via the floral dip method according to Clough and Bent (1998). The transformation of Flaveria bidentis was performed as described by Chitty et al. (1994). The integration of the transgenes into the genome of regenerated F. bidentis or T1 Arabidopsis plants was proved by PCR analyses.

Measurement of GUS Activity and Histochemical Analysis of Reportergene Activity
F. bidentis T0 plants used for GUS analysis were 40 to 50 cm tall and before flower initiation; the Arabidopsis T1 plants were examined around three weeks after germination.
Fluorometrical quantification of GUS activity was performed according to Jefferson et al. (1987) and Kosugi et al. (1990). For histochemical analysis of GUS activity leaves were cut manually with a razorblade and the sections were transferred to incubation buffer (100 mM Na 2 HPO 4 , pH 7.5, 10 mM EDTA, 50 mM K 4 [Fe(CN) 6 ], 50 mM K 3 [Fe(CN) 6 ], 0.1% (v/v) Triton X-100, 2 mM 5-bromo-4-chloro-3-indolyl-β-D-glucuronid acid). After brief vacuum infiltration the sections were incubated at 37°C for 1 to 20 hrs. After incubation, chlorophyll was removed from the tissue by treatment with 70% ethanol. Fluorescence microscopy was performed using a Zeiss Axiophot (Carl Zeiss AG, Jena, Germany) equipped with an Olympus DP50 camera (Olympus Optical Co., LTD., USA) and a Zeiss GFP imaging filter system FT 510,. Bright field and fluorescence images were overlaid with Adobe Photoshop 7.0 (Adobe Systems Inc., San José, CA, USA). For confocal laser microscopy a Zeiss LSM 510 with a Plan-Neofluar 25x objective was used. YFP fluorescence was monitored with a 505 to 550 nm band pass emission filter (488 nm excitation line). Chlorophyll autofluorescence was visualized with a long pass 560 nm emission filter.

In Situ RNA Hybridization
Non-radioactive in situ hybridization experiments were performed according to the protocol described by Simon (2002). Embedded leaves of F. bidentis and F. trinervia were cut into cross-sections of 20 µm thickness using a standard microtome. For the generation of the