Light-dependent, plastome-wide association of the plastid-encoded RNA polymerase with chloroplast DNA

Authors


Summary

Plastid genes are transcribed by two types of RNA polymerases: a plastid-encoded eubacterial-type RNA polymerase (PEP) and nuclear-encoded phage-type RNA polymerases (NEPs). To investigate the spatio-temporal expression of PEP, we tagged its α–subunit with a hemagglutinin epitope (HA). Transplastomic tobacco plants were generated and analyzed for the distribution of the tagged polymerase in plastid sub-fractions, and associated genes were identified under various light conditions. RpoA:HA was detected as early as the 3rd day after imbibition, and was constitutively expressed in green tissue over 60 days of plant development. We found that the tagged polymerase subunit preferentially associated with the plastid membranes, and was less abundant in the soluble stroma fraction. Attachment of RpoA:HA to the membrane fraction during early seedling development was independent of DNA, but at later stages of development, DNA appears to facilitate attachment of the polymerase to membranes. To survey PEP-dependent transcription units, we probed for nucleic acids enriched in RpoA:HA precipitates using a tobacco chloroplast whole-genome tiling array. The most strongly co-enriched DNA fragments represent photosynthesis genes (e.g. psbA, psbC, psbD and rbcL), whose expression is known to be driven by PEP promoters, while NEP-dependent genes were less abundant in RpoA:HA precipitates. Additionally, we demonstrate that the association of PEP with photosynthesis-related genes was reduced during the dark period, indicating that plastome-wide PEP–DNA association is a light-dependent process.

Introduction

Transcription of higher-plant plastid genomes is performed by two types of RNA polymerases (Weihe et al., 2012). One is a plastid-encoded eubacterial-type RNA polymerase resembling the Escherichia coli enzyme. It comprises four core subunits (α, β, β′ and β″, encoded by rpoA, rpoB, rpoC1 and rpoC2 within the plastid genome) plus one of several nuclear-encoded sigma factors for promotor recognition (Hu and Bogorad, 1990). The holoenzyme was designated plastid-encoded plastid RNA polymerase (PEP; Hess and Börner, 1999). The other chloroplast polymerase type is represented by nuclear-encoded phage-type RNA polymerases (NEPs), of which there is one in monocots or two in dicots (Weihe et al., 2012). The enzyme types recognize distinct promoters. While PEP recognizes σ70-type promoters, most NEP transcription start sites are characterized by a conserved YRTA motif and share homology with certain mitochondrial transcription start sites. In expressing the chloroplast genome, a division of labor exists between the two transcription systems: plastid genes may be classified as PEP-transcribed genes (mostly photosynthesis genes), NEP-transcribed genes (mostly housekeeping genes, including the genes for the core subunits of PEP), and genes that possess both PEP and NEP promoters (Weihe et al., 2012).

PEP has been shown to be located in plastids in two forms: as a soluble enzyme, and as an insoluble complex, bound to DNA, in a so-called ‘transcriptionally active chromosome’ (TAC) (Briat et al., 1979; Greenberg et al., 1984; Little and Hallick, 1988; Suck et al., 1996). The TAC is a high-molecular-weight DNA/RNA–protein complex that is capable of in vitro transcription (Hallick et al., 1976; Reiss and Link, 1985; Krause and Krupinska, 2000; Pfalz et al., 2006). The soluble PEP isolated from photosynthetically inactive etioplasts essentially consists of the bacterial-type core subunits (Pfannschmidt and Link, 1997). Preparations of soluble PEP from photosynthetically active chloroplasts contain additional proteins (Pfannschmidt and Link, 1997). In total, more than 50 proteins have been identified as components of the soluble PEP complex and/or the TAC (Pfannschmidt and Link, 1994; Pfannschmidt et al., 2000; Loschelder et al., 2004; Suzuki et al., 2004; Pfalz et al., 2006; Schroter et al., 2010; Steiner et al., 2011). Many of them are believed to be transcription factors. In fact, using in silico approaches, 78 putative plastid transcription factors were proposed to be present in the Arabidopsis genome (Wagner and Pfannschmidt, 2006; Schwacke et al., 2007). Not surprisingly, light plays a key role in the activation of plastid transcription in plants and thus in PEP assembly (Steiner et al., 2011). To create the photosynthesis apparatus, light induction leads to an increase in transcription of most plastid genes, and light-dependent transcription of certain plastid genes occurs during greening of the leaves as well as in mature leaves (Liere et al., 2011).

Chromatin immunoprecipitation is a well-established method for investigating the target regions of DNA-binding proteins. Changes in the interaction of the RNA polymerase, transcription factors and accessory proteins with DNA may be directly monitored in vivo (Aparicio et al., 2004). By cross-linking DNA–protein complexes with formaldehyde, in vivo complexes are preserved, and immunoprecipitation with specific antibodies may be used to enrich the protein together with its target DNA. In many eukaryotic systems, co-precipitated DNA was analyzed using microarrays to allow unbiased identification of DNA ligands (ChIP-on-chip assay). While the technique has been used successfully to analyze eukaryotic chromatin (Ren et al., 2000; Wu et al., 2006; Guenther et al., 2007) and to characterize the activity of various RNA polymerases (Sims et al., 2004; Wade et al., 2007), ChIP-on-chip analysis in chloroplasts is still in its infancy. The only study using this technique reported that the nucleic acid-binding protein WHY1 binds non-specifically to the entire maize chloroplast genome (Prikryl et al., 2008), but the function of this interaction is unclear. A few other studies on chloroplast DNA-binding proteins have used ChIP, but analyzed only a few selected target genes. Most of these studies focused on subunits of PEP, including two sigma factors (Hanaoka et al., 2012; Noordally et al., 2013), the core subunits RpoA and RpoB, and two peripheral subunits named pTAC3 and pTAC5 (Yagi et al., 2012; Zhong et al., 2013). These studies demonstrated that PEP associates with several chloroplast DNA regions. Typical PEP-dependent genes such as psbA and psaA were found to be attached to PEP (Hanaoka et al., 2012; Yagi et al., 2012; Noordally et al., 2013), and association increased in a light-dependent manner (Yagi et al., 2012). A comprehensive analysis of target preferences for any transcription factor or any RNA polymerase across the entire chloroplast genome is lacking.

Here, we used ChIP-on-chip to study the association of PEP with its target genes under various light conditions. We demonstrate that PEP association is restricted to a set of photosynthetic genes as well as tRNA and rRNA genes. Moreover, we found that binding of PEP is strictly light-regulated.

Results

Construction of transplastomic tobacco lines expressing an HA-tagged PEP subunit, and expression analysis of RpoA:HA

Detection and enrichment of plastid proteins may be facilitated by adding a hemagglutinin (HA) tag to the native protein sequence (Zoschke et al., 2010). Sensitive and highly specific antibodies against the HA tag are then used for diverse immunological experiments. To tag PEP, we targeted the C–terminus of the rpoA gene as rpoA is more easily manipulated than genes encoding other PEP subunits: rpoA is the last gene in a large chloroplast operon and is followed by an intergenic spacer, suitable for inserting the selectable marker gene aadA.

The general design of the plasmid construct used for transformation of the plants is shown in Figure 1(a) (see 'Experimental procedures' for details of the constructs used for transformation). To facilitate integration of the tag into the plastid genome by homologous recombination, the modified HA-containing rpoA sequence was flanked by large stretches of sequences homologous to the corresponding parts of the plastome (the borders of the insert are indicated by asterisks in Figure 1a). After plastid transformation, several transplastomic lines were obtained. Primary transformants were subjected to additional cycles of selection on antibiotic-containing medium to eliminate any residual wild-type copies of the plastome. Seeds generated by self-pollination of these plants produced exclusively spectinomycin-resistant plants, suggesting that these lines were indeed homoplastomic (Figure S1). Additionally, the transplastomic lines were checked by Southern hybridization to detect the relative quantities of the wild-type and transplastomic alleles (Figure 1b). Four independent lines (L3, L24A, L25A and L26B) were found to be homoplastomic for rpoA:HA. The macroscopic phenotypes of the transplastomic tobacco plants grown on soil were indistinguishable from that of the wild-type (Figure 1c). As knockout of rpoA and of PEP genes in general leads to albino plants (De Santis-Maciossek et al., 'Plant material and growth conditions'), and the RpoA:HA lines do not exhibit any phenotypic deviation from WT, we conclude that the tag does not interfere with RpoA function.

Figure 1.

Tagging of the tobacco chloroplast rpoA gene.

(a) Map of the rpoA genomic region. Exons are shown as black boxes and inserts as open boxes. Introns are represented by thin dotted lines connecting the exons. Black arrows indicate gene orientation. Insert borders are at 79155 and 81128. aadA, selective marker; HA, hemagglutinin tag; AflIII, Eco147I and BglII, restriction sites relevant to the cloning strategy. The EcoRV restriction sites are required for Southern analysis. The thick line indicates the position of the probe used for Southern hybridization; dotted lines show expected restriction fragments of EcoRV digests in Southern analysis.

(b) Test for homoplastomy of transformed tobacco lines by restriction fragment length polymorphism analysis. The identified EcoRV restriction fragments correlate with calculated sizes. Four independent lines (3A,C,D,E, 24A, 25A, 26B; letters refer to individual clones from a single transplastomic line) are homoplastomic for both the selective marker and the HA tag. Line 24B is homoplastomic only for the selective marker but not the HA tag. Line 26A is wild-type.

(c) Phenotypes of tagged RpoA and wild-type control lines. Transplastomic tobacco plants grown on soil are phenotypically indistinguishable from wild-type. RpoA:HA, C-terminal HA-tagged RpoA plant; WT, wild-type.

(d) Expression of recombinant RpoA:HA during tobacco development. From top to bottom: immunological detection of RpoA:HA with anti-HA serum; a section of the Ponceau S-stained blot showing the large subunit of RuBisCo; photographs of the plant material harvested at various ages. The graph shows quantification of chemoluminescence signals from three independent immunological analyses. The signals were normalized to the maximum signal for each experiment, and mean values are plotted against plant age. Vertical bars indicate standard deviations.

The accumulation of RpoA:HA was examined over the course of 60 days of development. Immunoblots of total shoot protein revealed that the tagged RpoA was already present at the 3rd day after imbibition. As shown in Figure 1(d), RpoA:HA accumulated gradually at later stages of development, and reached its highest expression between 7 and 25 days of plant development. A decrease of expression was observed in older material (38 and 59 days). No signal was detected in dry seeds (0 days). The specificity of RpoA:HA detection was demonstrated by the absence of any signals in lanes containing wild-type protein extracts (Figure 1d).

The association of HA-tagged RpoA with membranes depends on DNA in old tissue, but is independent of DNA in younger tissue

We were interested in the sub-organellar distribution of RpoA:HA under various conditions, as both soluble and membrane-associated PEP enzymes have been described (Little and Hallick, 1988; Suck et al., 1996; Sato, 2001; Sato et al., 2003; Schweer et al., 2010). We separated chloroplast membranes from stroma, and assayed RpoA:HA distribution in samples harvested during the light and dark phases of a 16 h day/8 h night cycle (Figure 2). Analysis of the thylakoid protein PetD verified that stroma fractions were free of membrane contaminations (Figure 2a). Membrane fractions were depleted of stromal contaminants by multiple washing steps (see RbcL in Figure 2a). On average, plants contained 23 times more RpoA:HA in the membrane than in the stroma fractions, and no effect of light on the RpoA:HA distribution was detected (Figure 2b).

Figure 2.

Localization of HA-tagged RpoA.

(a) Localization of RpoA:HA within tobacco chloroplasts during the light and dark periods. Upper panel: immunological detection of RpoA:HA. Middle panel: immunological detection of PetD on the same blot. Lower panel: Ponceau S staining of the same blot. TC, total chloroplast; Str, stroma; M, membrane; 1st/2nd/3rd/4th w, 1st, 2nd, 3rd and 4th washes of membranes.

(b) Immunoblots as shown in (a) were quantified. The results shown are based on two biological replicate experiments. The signals were normalized to total chloroplast extracts for each experiment. Values are means ± standard deviations.

As the plastid transcriptional apparatus as well as the plastid DNA are membrane-associated (Sato, 2001; Sato et al., 2003; Karcher et al., 2009; Schweer et al., 2010), we hypothesized that PEP is retained at membranes by a DNA tether. To test this, we purified membrane fractions from young tobacco plants (7-day-old seedlings) and older tobacco plants (60 days old), and treated buffer-suspended fractions with DNase I. After pelleting DNAse-treated membranes by centrifugation, RpoA:HA will be found in the supernatant if it was attached to the membranes via DNA. In young seedlings, DNase I treatment did not lead to any significant release of RpoA:HA from the membrane (Figure 3; left panel), whereas a considerable amount of the membrane-bound tagged protein was solubilized after treatment with DNase I in 60-day-old plants (Figure 3, right panel). It was only possible to extract DNA from the untreated membrane fractions, indicating successful digestion of the DNA by DNase I (Figure S2a). To further verify the efficacy of DNase treatment, a plastid gene fragment (psbA) was amplified from DNA extractions performed on the two fractions (membrane and supernatant) before and after DNase treatment. The fragment was amplified only from the untreated membrane samples, thus confirming the successful elimination of DNA by our treatment (Figure S2b). These data suggest a development-dependent attachment of RpoA:HA and probably also of the PEP holoenzyme to chloroplast membranes via DNA.

Figure 3.

Analysis of the role of DNA in membrane attachment of RpoA:HA. Detection of RpoA:HA in soluble and membrane fractions of isolated chloroplasts after DNase I treatment by immunoblot analysis. The upper panel shows the detection of tagged RpoA; the lower panel shows Ponceau S staining of the same membrane as a loading control. Equal aliquots of the preparation of the soluble as well as of the insoluble fractions were loaded. S, supernatant after centrifugation; P, insoluble membrane pellet; +, after DNase I treatment, –, control without DNase I treatment.

Co-enrichment of a PEP-like complex with RpoA:HA

Although our transplastomic approach should in theory not interfere with any sequence elements important for rpoA expression, it is still possible that most of the RpoA:HA protein is monomeric and thus not functional. To test whether RpoA:HA is part of a larger PEP-like complex, we immunoprecipitated the protein and analyzed precipitates by silver staining after SDS–PAGE. This revealed 13 polypeptides specifically enriched in to immunoprecipitations with the anti-HA antibody on plants expressing RpoA:HA, which were lacking in similar precipitations from wild-type plants (Figure 4). One of the polypeptides had a molecular mass of 41 kDa, which corresponds to the calculated molecular mass of tobacco RpoA:HA. Further bands with molecular masses corresponding to those of other PEP subunits [RpoB (121 kDa), RpoC1 (79 kDA) and RpoC2 (157 kDa)] were also detected. The slightly stronger signal observed for RpoA relative to the other core components of PEP may be explained by the stoichiometry of the enzyme (2:1:1:1, A:B:C1:C2) (Weihe et al., 2012). Additional proteins may represent known PEP-associated factors (Pfalz et al., 2006). These data suggest that we did indeed co-precipitate a complex resembling the PEP, similar to what was observed with alternative enrichment methods (Steiner et al., 2011), which is a prerequisite for identification of PEP-associated nucleic acids.

Figure 4.

Proteins co-precipitating with HA-tagged RpoA. Protein was precipitated from membrane chloroplast extracts using the specific HA-antibody, separated by SDS–PAGE and then silver stained. Proteins occurring only in the precipitates of the RpoA:HA line are indicated by asterisks, and their molecular masses are shown on the right. The HA-tagged RpoA has a calculated molecular mass of 41 kDa. The expected molecular masses of the PEP subunits are 121 kDa (RpoB), 79 kDa (RpoC1) and 157 kDa (RpoC2). HC and LC, heavy and light chains of the anti-HA antibody.

Identification of PEP-associated plastid DNA regions by ChIP-on-chip analysis of RpoA:HA

To examine DNA regions associated with PEP in vivo, we performed tobacco chloroplast ChIP-on-chip analysis, essentially following a recently established protocol (Newell and Gray, 2010). The plants were harvested 1 h after the onset of light. Chloroplasts were extracted and treated with formaldehyde to preserve the association of the DNA with the tagged RNA polymerase. Solubilized membrane fractions of cross-linked chloroplasts from 7-day-old plants were used for immunoprecipitation. Subsequent immunoblotting demonstrated that RpoA:HA was present in the solubilized membrane fraction, and that the tagged protein appeared in the immunoprecipitate. A small amount of protein remained in the supernatant (Figure S3). No RpoA:HA-specific signals were detected in the wild-type fractions, validating the specificity of the antiserum in in vivo extracts.

DNA from the precipitate and supernatant fractions from each immunoprecipitation experiment was then isolated and treated with RNase A to remove RNA contamination. DNA from supernatant and pellet fractions of immunoprecipitations were labeled with either red or green fluorescent dyes, and competitively hybridized onto a tobacco chloroplast genome tiling array. The array was designed to cover the entire tobacco chloroplast genome using overlapping PCR products with a mean length of 716 bp (see 'The association of HA-tagged RpoA with membranes depends on DNA in old tissue, but is independent of DNA in younger tissue'). Control experiments were performed using the anti-HA antiserum applied to solubilized membranes from wild-type chloroplasts. Ratios of enrichment in tagged versus non-tagged plants were calculated. These microarray results are illustrated in Figure 5(a) (see also Table S1).

Figure 5.

Identification of RpoA:HA-associated DNA sequences.

(a) Identification of DNA fragments associated with HA-tagged RpoA in membranes of chloroplasts under light and dark conditions by DNA co-immunoprecipitation and microarray analysis. Enrichment ratios of precipitated versus supernatant DNA from the RpoA:HA immunoprecipitation were compared to the corresponding ratios of the wild-type immunoprecipitation by division, and plotted against their genome position. The ratios for light conditions were normalized between two assays with RpoA:HA lines and two control experiments with wild-type. The ratios of the dark conditions were normalized between two assays with RpoA:HA lines and one wild-type control experiment. To compare light and dark conditions, the signals were normalized to signals for probes corresponding to the rpoB operon (a non-PEP target). The relevant data are shown in Table S1.

(b) Validation of ChIP-on-chip data by dot-blot analysis. DNA from the precipitate and supernatant from immunoprecipitations with the anti-HA antibody was isolated, spotted onto a nitrocellulose membrane and hybridized to radiolabeled PCR probes as indicated above each blot. Extracts from RpoA:HA and wild-type plants were probed in parallel. PCR probes were identical to probes used on the microarray. S, supernatant; P, pellet.

(c) Quantification of dot blots. Quantification of dot blots using a phosphorimager and the Quantity-One Software (BioRad). The ratio of pellet and supernatant signals of RpoA:HA lines was calculated and compared to the corresponding ratios derived from experiments with wild-type lines by division. Two biological replicates were combined. Vertical bars indicate standard deviations.

The ChIP assay showed enrichment of plastid genes known to be driven by PEP promoters, and these were mostly, but not exclusively, photosynthesis genes. The genes psbA, psbC/D and rbcL showed the strongest enrichment of all protein-encoding genes with RpoA:HA. Among the top 25 hits on the microarray are 12 probes for photosynthetic genes (see Table S2). In addition to photosynthesis-related genes, a strong association of RpoA:HA was also found with the rrn operon and certain tRNA genes, e.g. the trnV (UAC) to trnM (CAU) region. None of the genes known to be controlled mainly by NEP promoters, such as the rpoB operon, clpP, the large operon for ribosomal proteins or accD, were among the most enriched DNAs. Thus, PEP shows a clear preference for photosynthetic genes, rRNA genes and some tRNA genes.

Although several genes expected to be transcribed by PEP were among the top hits, some known PEP-dependent genes did not show enrichment. This group of genes included the psbB operon. There is genetic proof that the psbB operon is transcribed predominantly by PEP (Hajdukiewicz et al., 1997; Legen et al., 2002), but these genes were not more enriched than well-established NEP-dependent genes such as clpP or accD.

Selective reduction of DNA association with PEP in the dark

For certain genes, PEP–DNA association has been demonstrated to be light-dependent (Yagi et al., 2012). To analyze the DNA association with PEP at a plastome-wide scale in the dark, we performed ChIP-on-chip experiments using plants that were kept in the dark for 16 h before harvesting. The age and harvesting time of the plants was identical to the experiment described above in which the plants were irradiated for 1 h prior to the preparation of chloroplasts. There was no difference in the amount of precipitated RpoA:HA between plants grown under normal light conditions and those adapted to darkness (Figure S3).

To compare ChIP-on-chip experiments for material harvested in the light versus material harvested in the dark, we normalized two ‘dark’ and two ‘light’ experiments based on the sum of the signal for all probes representing the rpoB operon (20 PCR products). The rpoB operon seemed best suited for normalization purposes as this operon is transcribed by NEP (Legen et al., 2002) and was shown not to be associated with PEP in ChIP experiments (Yagi et al., 2012). We assume that minor signals obtained for the rpoB operon represent background DNA precipitation in these experiments.

Our comparison revealed that many regions of the plastid chromosome are preferentially associated with PEP in the light (Figure 5a and Figure S4). Differential enrichment in ‘light’ versus ‘dark’ experiments is most pronounced for the regions encompassing the psbC/D operon, the rrn operon, psbA, the psaA/B operon, and rbcL. Among tRNA genes, the intron-containing tRNA genes trnI-(GAU) and trnA-(UGC) show light-dependent enrichment with PEP, which is not entirely surprising as these are part of the rrn operon. In addition, the trnD/Y/E region, which is transcribed as a single precursor RNA (Tsunoyama et al., 2004), showed strong association with PEP in the light. All of the above discussed regions are represented on the microarray by one or more probes with a more than twofold increase in PEP association in the light (altogether 17 probes with at least twofold enrichment in ‘light’ versus ‘dark’ experiments; Figure 5a and Figure S4).

Overall, 227 of 302 probes on the array showed a differential enrichment in light over dark experiments (Figure 5a and Figure S4). This roughly corresponds to 75% of the entire plastome. Thus, PEP association with the plastid chromosome is globally increased in the light, and is most pronounced for genes that are known to be light-regulated (e.g. psbA and psbC/D; Gamble and Mullet, 1989; Greenberg et al., 1989; Christopher and Mullet, 1994). Of the 75 probes showing stronger enrichment of RpoA:HA co-precipitated DNA in the dark, only one shows a more than twofold increase in association compared with experiments in the light (Figure 5a and Figure S4). This corresponds to the petL–petG–trnW/P region. Further minor peaks of DNA-enrichment in dark experiments relative to light experiments were found for genes in the ndh operon. On the whole, few genes appear to be preferentially associated with PEP in the dark.

To verify the ChIP data by an independent method, dot-blot DNA hybridization with selected plastid genes was performed. DNA extracted from the immunoprecipitation fractions (precipitate and supernatant) was spotted onto nylon membranes and hybridized with plastid gene-specific probes generated by PCR (Figure 5b,c). In the case of rbcL, rrn23, psbD and psbA (all exhibiting strong binding to the tagged PEP in precipitations from material harvested in the light), a specific increase of the hybridization signal was observed in the RpoA:HA immunoprecipitate fraction versus the wild-type control fraction, indicating binding of the tagged PEP and thus confirming the ChIP-on-chip data. With rpoC2 and ndhF probes, representing genes that showed no association with RpoA:HA in the Chip-on-chip array, very low signals were observed in the dot-blot experiment (Figure 5b,c).

Discussion

PEP associates with large regions of the chloroplast chromosome

At least three RNA polymerases are active in chloroplasts of dicotyledonous plants: PEP and two versions of NEP (the latter encoded by the nuclear genes RpoTp and RpoTmp). Attempts have been made to dissect the targets of these genes genetically: tests of RNA accumulation in polymerase-deficient mutants identified target genes of PEP and NEP (Hajdukiewicz et al., 1997; Kuhn et al., 2009). However, the results of such analyses must be treated with caution as knockouts of PEP as well as of the two nuclear-encoded polymerases have strong phenotypic effects, i.e. albinism in the case of PEP (Allison et al., 1996), and leaf deformations, bleaching and slower growth in the case of NEP (Baba et al., 2004; Hricova et al., 2006; Kuhn et al., 2009). Such massive secondary effects may interfere with transcription by the remaining RNA polymerase activities. For example, loss of chloroplast gene expression in general is known to generate so-called retrograde signals that lead to a massive reprogramming of nuclear gene expression (Woodson and Chory, 2008), including the NEP genes (Emanuel et al., 2004, 2006). Thus, modifying PEP activity will alter NEP activity as well (Zhelyazkova et al., 2012). Similarly, as chloroplast PEP genes are transcribed by NEP, it is difficult to obtain clean NEP phenotypes even in double knockouts of the two NEP genes RPOTmp and RPOTp. Finally, knockouts of the chloroplast- and mitochondrial-targeted RPOTmp protein have been shown to lead to compromised mitochondrial respiratory chain activity by obliterating complex I expression (Kuhn et al., 2009). As defective mitochondria are known to affect chloroplast gene expression (Leister, 2005; Van Aken and Whelan, 2012), this may also lead to problems identifying the true targets of the plastid RNA polymerases. These complex inter-relationships complicate the analysis of chloroplast transcription mutants.

A more direct approach to characterize the genes served by a specific polymerase involves ChIP experiments. Co-precipitation of DNA together with a particular RNA polymerase may provide strong evidence for a functional interaction, although it does not deliver proof of polymerase activity at the respective site. This technique has been widely used to characterize bacterial RNA polymerases (Wade et al., 2007) as well as eukaryotic polymerases, particularly RNA polymerase II (Sims et al., 2004). For the bacterial RNA polymerase, a large number of studies performed under various conditions and genetic backgrounds have uncovered regulatory roles for this enzyme (Wade et al., 2007).

For the chloroplast transcription machinery, first attempts to characterize bound DNA consisted of visualizing restriction-digested DNA isolated from TAC preparations. This demonstrated that the entire plastid chromosome is present in the TAC, but did not lead to a more refined view of transcribed DNA (Reiss and Link, 1985). Recently, immunoprecipitations of chloroplast RpoA using an antibody raised against almost the entire RpoA protein were performed, and bound DNA was detected by quantitative PCR for 14 probes (Yagi et al., 2012). Enrichment was found for probes detecting psbA, rbcL, psbD, the rrn operon, trnDY and psaA. Similarly, enrichment of psbA, psaA, rbcL and the rrn operon was found in immunoprecipitations of RpoB (Zhong et al., 2013). Likewise, ChIP/quantitative PCR of sigma factor 5 identified psbA, psbC/D, psbA/B and psbB/T as target regions (Noordally et al., 2013), and ChIP/quantitative PCR of sigma factor 1 identified rbcL, psbB/T, psaA/B, clpP and psbEFLJ as target regions (Hanaoka et al., 2012). This suggests that PEP is associated with several photosynthetic genes.

Our ChIP-on-chip approach confirms these targets and puts these genes in the context of the entire plastome, thus pinpointing the main targets of PEP. A technical benefit of our approach is use of the HA epitope tag, which enables highly specific immunoprecipitations that pull-down exclusively RpoA:HA; no cross-reactions were seen in protein gel-blot analysis of pelleted proteins or total protein extracts from wild-type plants. We also provide evidence that the precipitated PEP is intact, since the amount of precipitated RpoA protein appears to be similar to the other PEP subunits identified, which would reflect approximately the stoichiometry of PEP (A2:B:C1:C2). We conclude that most of the RpoA precipiated is probably within a PEP complex. Finally, as we are performing direct chemical labeling of DNA, our technique lacks any enzymatic reaction, unlike ChIP/quantitative PCR, which uses target amplification with the risk of introducing enzymatic biases.

The top target regions of PEP correspond to the rRNA operon and several photosynthetic genes. In fact, two of the genes identified as preferred targets for PEP were identified by all studies mentioned above: psbA and psaA/B. In addition, the psbC/D operon, rbcL and trnV–trnM are within the top five peaks in Figure 5(a). These results mirror the most abundant RNAs in tobacco, and therefore further confirm that PEP is the most important RNA polymerase for plastid transcription at least in green tissue (Legen et al., 2002; Zhelyazkova et al., 2012). The results also show that, in green tissue, the efficacy of PEP–DNA co-enrichment (Figure 5) reflects transcriptional activity in most cases (Legen et al., 2002).

Light-dependent PEP–DNA association

As we have investigated binding to DNA in the light and the dark, we are able to identify which are the main light-regulated genes at the transcriptional level. On a global scale, we determined how much of the chloroplast genetic information is subject to PEP-dependent light activation. We found that approximately 75% of the chloroplast chromosome is more strongly associated with PEP in the light than in the dark. Such global changes may be brought about by light-dependent activation of PEP mediated by changes in the phosphorylation state of core PEP subunits as well as sigma factors (Baginsky et al., 1997; Isono et al., 1997; Kanamaru et al., 1999; Tan and Troxler, 1999).

What are the main genes associated with PEP in the light? Light regulation of chloroplast genes at the level of transcription of individual genes has been known for a long time, including for the genes rbcL, psbA, rrn16 (DuBell and Mullet, 1995; Chun et al., 2001), psbD (Mochizuki et al., 2001; Thum et al., 2001) and psaB (DuBell and Mullet, 1995). Our analysis provides complementary and direct evidence that PEP is important for light regulation of these key genes in photosynthesis. In addition, our analysis reveals that genes for ribosomal RNA and tRNAs are among the main PEP targets in the light (including the rrn operon, the trnV (UAC) to trnM (CAU) region, the psbZ–trnG–trnfM region, and the trnR (ACG) trnN (GUU) orf75 region; see Table S2). This suggests that transcription of RNA components of the chloroplast translational apparatus is switched on in the light relative to the dark. Indeed, moderate increases in transcriptional activity of rRNA genes upon increasing light have been described in Sinapis alba (Baena-Gonzalez et al., 2001). The reason for light-induced PEP association with rRNA and tRNA genes in green tissue may be to replenish ribosome and tRNA pools that turnover between day and night. The recent finding that the PEP is in particular important for tRNA expression further supports this idea (Williams-Carrier et al., 2013). A time-resolved ChIP-on-chip analysis performed at the onset of daylight is necessary to clarify this issue.

However, these results are in contrast to those of studies on de-etiolation, which showed no increase in transcription or even a decrease in transcriptional activity for rDNA (Rodermel and Bogorad, 1985; Zhu et al., 1985; Deng and Gruissem, 1987; Klein and Mullet, 1990). Possibly, harvesting light-grown seedlings after a slightly extended dark period (as performed here) does not compare directly to the situation of etiolated tissue starting to green. In fact, several light regulatory mechanisms only occur in green, mature leaves (Aro et al., 1993), and action spectra as well as photoreceptors are different between etioplasts and chloroplasts (e.g. Fluhr et al., 1986; Chory et al., 1989; Quail, 1994). In addition, preparations of RNA polymerizing activities from chloroplasts and etioplasts have different protein compositions (Reiss and Link, 1985; Pfannschmidt and Link, 1994; Suck et al., 1996; Majeran et al., 2011). Thus, our data and data obtained from de-etiolation systems are not directly comparable.

Are there membrane anchors for PEP?

Two Spinacia oleracea NEP enzymes have been reported to be membrane-associated in mature chloroplasts in a non-DNA-mediated manner (Azevedo et al., 2006). Furthermore, in Arabidopsis, the NEP protein encoded by RpoTmp has been shown to be anchored to the thylakoid membrane by a RING-finger protein in response to light signals (Azevedo et al., 2008). No similar light-regulated attachment factor is known for PEP, despite the fact that membrane association has been known for a long time (Reiss and Link, 1985). PEP subunits have been detected in preparations of nucleoids and the TAC (Suck et al., 1996; Pfalz et al., 2006; Majeran et al., 2011), both of which are known to be localized to membranes. We quantified membrane association and demonstrate that more than 95% of PEP is membrane-bound and that this does not change in a light-dependent fashion. It will be interesting to investigate the mechanisms for PEP attachment to membranes in comparison to RING-finger protein-based attachment of NEP (RPOTmp; Azevedo et al., 2008).

We suspected that PEP may be linked to membranes via DNA, as chloroplast DNA is membrane-attached (Miyamura et al., 1986; Liu and Rose, 1992; Sato et al., 1993, 1999). However, digestion of DNA did not change the localization of PEP, at least in young tissue, suggesting that there is a permanent non-DNA anchor that retains PEP at the membrane irrespective of light conditions. The situation is further complicated by the finding that, in old tissue, PEP may be partially solubilized by digesting the DNA. Whether this is simply due to a change in transcriptional activity or is a property of the unknown membrane anchor remains to be seen. Candidates for attachment factors include the PEND protein and MFP1. PEND, a member of the bZIP protein family, is associated with the inner envelope membrane and is expressed at early stages of plastid development, i.e. at the same time as the nucleoids are found to co-localize with the plastid envelope (Sato et al., 1993, 1998). MFP1 is a coiled-coil protein with a C–terminal DNA-binding domain and is associated with nucleoids in vivo (Jeong et al., 2003). It will be interesting to test whether any of these general DNA-binding proteins directly interact with PEP.

Experimental procedures

Plant material and growth conditions

Wild-type and transplastomic Nicotiana tabacum (var. Petite Havana) RpoA:HA lines were grown on soil with a 16 h light/8 h dark cycle at 27°C. For chloroplast isolation, plants were grown on water-saturated vermiculite with fertilizer (Osmocote Pro 5–6M, Scotts, http://www.scotts.com/smg/home/home4.jsp) under a 16 h light/8 h dark cycle at 27°C for 7 days. For dark treatments, plants were grown under standard conditions for 6 days, and then adapted to darkness 16 h prior to isolation of plastids. The harvesting time was always 9:00 am, i.e. 1 h after the onset of light.

Cloning and generation of transplastomic lines

The target region of petD–poA (Figure 1) for homologous recombination was amplified from N. tabacum chloroplast DNA using primers EEpetDfor and EErpoAfor (see Table S3 for the list of primers) and cloned into pBluescript II SK+ (Stratagene, http://www.genomics.agilent.com/en/home.jsp) via Eco32I and PstI linkers (an AflIII site was removed in the vector for cloning reasons). A DNA fragment encoding the hemagglutinin (HA) epitope (GeneArt, http://www.lifetechnologies.com) with AflIII and BglII linkers was introduced into the construct at the 3′ end of rpoA. The aadA selective marker was PCR-amplified from plasmid pRZN+ (Zoschke et al., 2010), and blunt end-ligated into the above-described vector using an Eco147I site in the petD–rpoA intergenic region. Orientation of the aadA cassette was verified by PCR using primers aadAoutfor and M13-rev (Table S3). As a control for effects of the aadA insertion, a vector lacking the HA epitope was also constructed. Chloroplast transformation was performed using a biolistic gun as described previously (Svab and Maliga, 1993). The presence of the HA tag in plastid transformants was verified by Southern blot, PCR and immunologically.

Southern blot analysis of transplastomic plants

Total DNA was isolated from 7-day-old seedlings of wild-type and transplastomic tobacco plants by the cetyl trimethylammonium bromide method (Murray and Thompson, 1980), and analyzed by Southern hybridization (Southern, 1975). DNA (7 μg) was digested using EcoRV. A 32P-radiolabeled rpoA-specific probe was generated by PCR using primers SF_rpoArevS and SF_rpoAforS (Table S3) and hybridized to size-separated and immobilized DNA.

Immunoblot analyses

Total protein from tobacco leaves or chloroplasts was separated by SDS–PAGE and blotted onto Hybond–C Extra nitrocellulose (GE Healthcare, http://www.gelifesciences.com). Loading and transfer efficiency of proteins was tested by Ponceau S staining. The blots were incubated for 1 h with anti-HA antiserum (clone HA–7, purified immunoglobulin; Sigma, http://www.sigmaaldrich.com) in 2% skim milk powder in TBST (10 mm Tris/HCl, pH 7.5, 150 mm NaCl and 0.1% Tween–20), followed by a 30 min incubation with horseradish peroxidase-coupled antibody (GE Healthcare). Immunoblots were quantified using a Chemidoc XRS+ imager and Quantity One software (Bio–Rad, http://www.bio-rad.com/).

Isolation of chloroplasts and preparation of membrane fractions

Intact chloroplasts were isolated from 7-day-old seedlings of wild-type and RpoA:HA lines of N. tabacum (Voelker and Barkan, 1995), with the following modifications. After homogenization of the tissue, chloroplasts were pelleted at 1000 g for 6 min at 4°C, the pellet was washed in resuspension buffer (50 mm HEPES/KOH pH 8, 330 mm sorbitol) by gentle agitation at 4°C and re-centrifuged at 1000 g for 6 min at 4°C. The washed pellet was resuspended in 1 ml resuspension buffer. Then, a formaldehyde cross-linking step was performed as described previously (Newell and Gray, 2010) to cross-link DNA to associated proteins. Lysis of chloroplasts was performed in 300 μl extraction buffer [2 mm dithiothreitol, 200 mm KOAc, 30 mm HEPES/KOH, pH 8, 10 mm MgOAc, 1× proteinase inhibitor cocktail (Roche, http://www.roche.de)] per 40 g leave material, and the lysate was then passed 40 times through a 0.5 mm × 25 mm syringe (Kupsch et al., 2012). Membranes and stroma were separated by centrifugation at 21 000 g at 4°C for 30 min. Membranes were washed three times in extraction buffer, and finally resuspended in 500 μl extraction buffer.

DNase treatment of purified chloroplast membranes

Chloroplast membrane fractions were purified from 7-day-old seedlings and 60-day-old plants as described above. Aliqouts (100 μl) were incubated at room temperature with 200 units DNase I (Roche) or in control reactions without DNase I as described previously (Thelen and Ohlrogge, 2002). After incubation, the supernatant was separated from the insoluble membranes by centrifugation for 10 min at 21 000 g at 4°C. Membrane pellets were resuspended in 50 μl extraction buffer, and a one-tenth volume for each fraction was used for immunoblot analysis. To test the success of the DNase I digest, the DNA from insoluble membranes and supernatant fractions was extracted with phenol/chloroform/isoamyl alcohol (12:12:1) and analyzed by agarose gel electrophoresis. PCR analysis was performed using primers psbA fw and psbA rev (Table S3).

Immunoprecipitation of RpoA:HA from chloroplast membrane fractions

Aliquots (50 μl) of resuspended membranes were diluted with 100 μl extraction buffer, NP-40 was added to a final concentration of 1%, and the samples were incubated on ice for 15 min to solubilize all membrane-bound proteins. At this stage, the DNA was sheared by sonication with four short bursts of 2 sec using a Bandelin Sonopuls HD60 (Bandelin, http://www.bandelin.com), allowing the suspension to cool down on ice after every sonication step, yielding fragments of mean length 400 nt (Figure S5). The sample was centrifuged at 21 000 g for 10 min at 4°C. The supernatant was used for subsequent co-immunoprecipitations (Kupsch et al., 2012) with 10 μl of anti-HA antiserum. Control reactions contained wild-type chloroplast membranes. The magnetic beads carrying the antibody–protein–DNA complexes were collected and the supernatant was retained. The precipitate was washed three times in co-immunoprecipitation buffer (0.15 m NaCl, 20 mm Tris-HCl (pH 7.5), 2 mm MgCl2, 0.5 Vol.-% NP-40.5 μg/ml Aprotinin).

DNA extraction and RNase digestion

To reverse the initial cross-linking, the immunoprecipitation samples were incubated at 70°C for 30 min. The DNA in the precipitate and supernatant was extracted using phenol/chloroform/isoamyl alcohol. RNase A was added to a final concentration of 0.5 mg ml−1 to digest RNA contaminants, and the DNA was again extracted using phenol/chloroform/isoamyl alcohol.

Tobacco chloroplast genome tiling array design

Overlapping DNA fragments covering the entire tobacco chloroplast genome (see Table S3 for the list of primers) were amplified from tobacco chloroplast DNA using Phusion DNA polymerase (Finnzymes, http://www.thermoscientificbio.com/finnzymes/). A total of 500 ng of each PCR product was spotted onto glass slides as described previously (Kupsch et al., 2012).

DNA labeling and microarray hybridization

Equal-volume aliquots of DNA from precipitate and supernatant were labeled by adding 2 μl of 10× labeling buffer and 1 μl of Cy5 or Cy3 fluorescent dye (aRNA labeling kit; Kreatech Diagnostics, http://www.kreatech.com/). The labeling reaction, microarray hybridization, scanning and data evaluation was performed as described previously (Kupsch et al., 2012).

DNA dot-blot hybridization

DNA from the co-immunoprecipitation pellet and supernatant fractions was used for dot-blot analyses. Six replicate dots per sample were prepared from one immunoprecipitation. Aliquots of 20 μl DNA each from pellet and supernatant fractions were mixed with 480 μl sample buffer (66% deionized formamid, 7.7% formaldehyde, 1.3× MOPS buffer, pH 7). The mixture was denatured at 75°C for 15 min, 180 μl of 20× SSC were added, and the samples were blotted to a Hybond N nylon membrane (GE Healthcare) using a Bio–Dot SF microfiltration apparatus (Bio–Rad). The membranes were then washed with 5× SSC, UV cross-linked at 150 mJ cm−2, and hybridized to PCR probes that were radioactively labeled with [α–32P]dCTP using the DecaLabel DNA labeling kit (Fermentas, http://www.thermoscientificbio.com/fermentas/) according to the manufacturer's instructions.

Acknowledgements

We thank Karsten Liere und Thomas Börner for critical discussion of the data. C.S.L. wishes to acknowledge support by the Deutsche Forschungsgemeinschaft (SCHM1698/4-1). R.Z. is supported by a postdoctoral fellowship from the Deutsche Forschungsgemeinschaft (grant number ZO 302/1–1). The antisera against PetD were kindly provided by Alice Barkan (Institute of Molecular Biology, University of Oregon, Eugene, OR).

Ancillary