PPR8522 encodes a chloroplast-targeted pentatricopeptide repeat protein necessary for maize embryogenesis and vegetative development.

The pentatricopeptide repeat (PPR) domain is an RNA binding domain allowing members of the PPR superfamily to participate in post-transcriptional processing of organellar RNA. Loss of PPR8522 from maize (Zea mays) confers an embryo-specific (emb) phenotype. The emb8522 mutation was isolated in an active Mutator (Mu) population and co-segregation analysis revealed that it was tightly linked to a MuDR insertion in the first exon of PPR8522. Independent evidence that disruption of PPR8522 caused the emb phenotype was provided by fine mapping to a region of 116kb containing no other gene than PPR8522 and complementation of the emb8522 mutant by a PPR8522 cDNA. The deduced PPR8522 amino acid sequence of 832 amino acids contains 10 PPR repeats and a chloroplast target peptide, the function of which was experimentally demonstrated by transient expression in Nicotiana benthamiana. Whereas mutant endosperm is apparently normal, mutant embryos deviate from normal development as early as 3 days after pollination, are reduced in size, exhibit more or less severe morphological aberrations depending on the genetic background, and generally do not germinate. The emb8522 mutation is the first to associate the loss of a PPR gene with an embryo-lethal phenotype in maize. Analyses of mutant plantlets generated by embryo-rescue experiments indicate that emb8522 also affects vegetative plant growth and chloroplast development. The loss of chloroplast transcription dependent on plastid-encoded RNA polymerase is the likely cause for the lack of an organized thylakoid network and an albino, seedling-lethal phenotype.

Introduction

Maize is a crop species that presents a tremendous agronomic and economic interest. Today 60% of the world’s nutrition is provided by the three cereals wheat, maize, and rice, maize being the highest yielding species. Consequently the maize kernel has been at the heart of substantial research efforts aimed at the understanding and manipulation of the molecular mechanisms governing both its development and the accumulation of important reserve substances such as oil in the embryo and starch in the endosperm. The mature embryo consists of the embryo axis comprising the shoot and root primordia, which give rise to the seedling, and the scutellum, a massive storage organ which is functionally equivalent to cotyledons in dicots (Sosso ). The highly differentiated embryo is elaborated from the single-celled zygote in three major developmental events, which are the establishment of an apical-basal polarity, the differentiation of the protoderm, and the formation of the shoot and root meristems (Kaplan and Cooke, 1997).

Maize embryo mutants have been collected and characterized since the beginning of the last century (Demerc, 1923; Mangelsdorf, 1923) and have served as a powerful tool for the identification of genes required for seed development (Vernoud ). Within the large category of defective kernel (dek) mutants, which show defects in either the embryo or the endosperm or both (Demerc, 1923), the class of embryo-specific (emb) mutants, in which only the embryo but not the endosperm is affected, has attracted particular interest (Clark and Sheridan, 1991). In theory, mutations affecting only the embryo and not the endosperm are less likely to concern housekeeping genes involved in basic cellular functions or primary metabolism, because the loss of this type of gene should affect both the embryo and the endosperm. Consequently the focus on emb rather than dek mutants supposedly increases the likelihood of identifying genes involved specifically in embryo development rather than in basic cellular metabolism. Whereas numerous emb mutants have been isolated in maize (Clark and Sheridan, 1991; Sheridan and Clark, 1993), only 17 have been subject of a detailed phenotypic characterization (Sheridan and Thorstenson, 1986; Heckel ; Elster ; Consonni ; Sosso ). The causal mutations have been identified for two emb mutants, emb8516 and lem1. Contrary to the expectation, in both cases the mutations were found to affect housekeeping genes and more precisely nuclear genes encoding subunits of plastid ribosomes (Ma and Dooner, 2004; Magnard ).

Plastids are well known to play an important role in embryo development, since in Arabidopsis approximately 30% of all embryo-defective (emb) mutations concern plastid-targeted proteins (Devic, 2008; Bryant ). Among them, a great number carried lesions in pentatricopeptide repeat (PPR) proteins, which are characterized by the presence of the homonymous motif. The PPR motif is a degenerate 35 amino acids repeat sequence that is found in animal, fungal, and plant proteins (Small and Peeters, 2000). The PPR protein family is particularly large in plants, where the majority of the family members are predicted to be targeted to mitochondria or chloroplasts (Lurin ) and to play a major role in the processing of organellar RNAs (Schmitz-Linneweber and Small, 2008). Chloroplast-targeted PPR proteins have also been functionally characterized in maize, but the reported mutant phenotypes never concerned the seed. The ppr2 mutant shows an albino phenotype and lacks plastid rRNA and translation products (Williams and Barkan, 2003). Likewise, the ppr5 mutant is deficient in chloroplast ribosomes and dies as an albino seedling (Beick ).

Lesions in other classes of chloroplast-targeted proteins have also been shown to cause embryo lethal phenotypes. Among the first ones were the mutations in glycyl-tRNA synthetase and chaperonin-60alpha, which lead to embryo lethality in the embryo-defective development1 (edd1) and schlepperless (slp) mutants, respectively (Uwer ; Apuya ). Recently, a comprehensive compilation of 119 genes coding for chloroplast-targeted proteins required for embryo development suggested that three groups of proteins were frequently encountered: (i) enzymes required for the biosynthesis of amino acids, vitamins, nucleotides, and fatty acids, (ii) proteins required for the import, modification, and localization of essential proteins in plastids and (iii) proteins required for plastid translation (Bryant ).

In maize, embryo lethality has been reported for mutations in the nuclear-encoded plastid ribosomal proteins PRPS9 and PRPL35, which cause an embryo-specific phenotype affecting only the embryo and not the endosperm (Ma and Dooner, 2004; Magnard ). However, not all mutations affecting plastid translation are lethal in maize. For example, the crs2, caf1, or caf2 mutants impaired in the splicing of the plastid-encoded rpl16, rps16, and/or rps12 mRNAs produce viable albeit ivory seedlings (Ostheimer ). Furthermore, all non-photosynthetic maize mutants analysed so far produce viable seedlings and have not been associated with embryo lethality (Stern ). The current report presents the molecular and phenotypic characterization of emb8522, in which a mutation in PPR8522 causes an embryo-specific phenotype. The data show that PPR8522 is a chloroplast-targeted PPR protein that is necessary for the transcription of nearly all plastid-encoded genes.

Materials and methods

Plant material and culture

The isolation of the emb8522 mutation in stock R-scm2 (Clark and Sheridan, 1991) and subsequent backcrosses to inbred line A188 have been described (Heckel ). Further backcrosses involved the inbred line B73 provided by the Maize Genetics Cooperation Stock Center (http://maizecoop.cropsci.uiuc.edu/). To ascertain the presence of the mutation during backcrosses, emb heterozygotes were always used as males and self-pollinated in parallel to the backcross and scored for ears with a segregation ratio of 3:1 for wild-type:emb phenotype kernels. For co-segregation and mapping, plants were grown under field conditions at the ENS-Lyon (Lyon, France) or the SMH INRA field station (St. Martin de Hinx, France). For all other purposes, plants were grown in a S2 greenhouse (authorized for the culture of transgenic plants) with a 16/8 light/dark cycle (100 Wm-2) at 24/19 °C without control of the relative humidity. Seeds were germinated in 0.2 l Favorit MP Godets substrate (Eriterre) and were transferred between 14 and 21 days after sowing to 10 l Favorit Argile substrate (Eriterre). For embryo-rescue experiments, immature embryos were excised from caryopses at 16 days after pollination (DAP) and cultivated in the dark at 22 °C on one of the following media: medium 1 (1% sucrose, 2.2g/l Murashige and Skoog salts, 1% agar-agar), medium 2 (6% sucrose, 2.2g/l Murashige and Skoog salts, 1% agar-agar), and medium 3 (6% sucrose, 2.2g/l Murashige and Skoog salts, 3g/l Phytagel). After reaching a mesocotyl length of 2cm, the rescued embryos were transferred to a different medium (5% sucrose, 1.1g/l Murashige and Skoog salts, 8g/l agar-agar) and grown with a 16/8 light/dark cycle at 25 °C.

Light microscopy of cytological sections

Immature wild-type and mutant kernels were harvested from the ear of a self-pollinated heterozygous plant at 3, 5, 9, 18, and 24 DAP and cut along the longitudinal axis in three equal parts. The central slice containing the embryo was fixed for 1h at room temperature in 100mM Dulbecco’s phosphate buffer (Lonza Walkersville) with 1% paraformaldehyde and 3% gluteraldehyde. The fixed material was dehydrated in a graded ethanol series (10, 30, 50, 70, 90, and 100% ethanol), embedded in Technovit 7100 resin (Kulzer Heraeus) according to manufacturer’s instructions, sectioned at 6–8 µm with a HM3555 microtome (Microm Microtech), and stained with Schiff reagent (Sigma Aldrich). The sections were observed with an Axio Imager M2 microscope coupled with an Axio Cam MRC digital camera (Zeiss).

Transmission electron microscopy

Wild-type and mutant leaf squares or embryos were fixed overnight at room temperature in 0.2M cacodylate buffer, pH 7.1, with 2% glutaraldehyde. The fixed tissue was then washed in cacodylate buffer and treated for 1h in Dulbecco’s phosphate buffer supplemented with 1% osmium tetroxide, followed by dehydration in a graded ethanol series and embedding in Agar Low Viscosity Resin (Agar Scientific). Ultrathin sections of 80nm were cut using a diamond knife and lifted onto 2mm copper grids. Grids were stained in 2% uranylacetate and lead citrate prior to observation with a 1200EX transmission electron microscope (JEOL).

Co-segregation analysis and fine mapping

Genomic DNA was extracted from young leaf pieces (approximately 1cm2) or micro-dissected endosperms (12-DAP caryopses) using the BioSprint Workstation and the DNA Plant Kit (Qiagen), according to the instructions of the supplier. Genomic DNA was diluted 20-fold in water prior to 35 cycles of PCR amplification with the GoTaq Core System mixture (Promega) with annealing temperatures between 58 and 62 °C (optimized for each primer pair) and an extension time of 20 s. After control amplification with primers Isum2E/Isum2F, co-segregation analyses or fine mapping were carried out with the primers listed in Supplementary Table S1 (available at JXB online). PCR products were visualized by conventional agarose gel electrophoresis (Sambrook ) or the Origins Electrophoresis System (Elchrom Scientific, Switzerland). In addition to simple sequence repeat markers from the MaizeGDB database (www.maizegdb.org/probe.php), new indel markers detecting insertions/deletions of at least 2bp were designed after comparative sequencing of the respective region in wild-type and emb8522 mutant endosperm.

Subcellular localization of PPR8522

To generate a translational fusion between PPR8522 and GFP, the full-length open reading frame of PPR8522 lacking its stop codon was amplified by reverse-transcription (RT) PCR with primers attB1-PPRF1 and attB2-PPRR2 (Supplementary Table S1) on 9-DAP kernels from inbred line A188 and introduced into the binary vector pK7FWg2 and pK7m34GW (Karimi ) by GATEWAY site-specific recombination (Invitrogen). These binary constructs, which contained a cauliflower mosaic virus 35S promoter driving constitutive expression of PPR8522:GFP and PPR8522:mCherry, respectively, were introduced into A. tumefaciens strain LBA4404 by electroporation. Similarly fusions of the Arabidopsis PEND (PLASTID ENVELOPE DNA BINDING) gene were obtained starting after initial amplification with primers SP/AthPEND and ASP/AthPEND on published plasmid A8 (Terasawa and Sato, 2005) and cloning into pEntr-D-TOPO. Agrobacterium-mediated transient expression in Nicotiana benthamiana was performed as following standard procedures (Batoko ). Constructs were co-infiltrated with the 35S:P19 plasmid expressing the silencing suppressor p19 of tomato bushy stunt virus (Voinnet ). GFP and mCherry fluorescence and chloroplast autofluorescence were detected 3 days after infiltration by the use of a Leica TCS SP5 confocal microscope with simultaneous acquisition at 522–572nm (GFP) and 667–773nm (autofluorescence) and sequential acquisition at 569–643nm (mCherry).

RNA extraction and RT-PCR

Approximately 250mg fresh tissue was quick frozen in liquid nitrogen and ground to powder with a mortar and pestle. Total RNA was extracted with 1ml TRI reagent (Molecular Research Center), according to the instructions of the supplier. After ethanol precipitation, the RNA was resuspended in 30 µl RNAse free water and treated with RNase free DNase (Ambion), which was then inactivated according to the instructions of the supplier. Approximately 5 µg of total RNA were reverse transcribed using random hexamers (Amersham Biosciences) and reverse transcriptase without RNaseH activity (Fermentas) in a final volume of 20 µl. For RT-PCR, template was provided as 2 µl of a 50-fold dilution of cDNA in water in a total reaction volume of 20 µl.

Plastid transcriptome analysis

The hybridization of a microarray with 248 oligonucleotides covering the entire maize plastid genome was performed as described previously (Schmitz-Linneweber ). Data were imported into GenePix Pro 6.0 (Axon Instruments) and filtered. The median of the relative enrichment between two samples was log2-transformed and plotted on the Y-axis for every single chloroplast gene. The log2-transformed enrichment ratios were normalized between the two samples in the comparison.

Quantitative reverse-transcription PCR

cDNA was diluted 50 times and 2 µl used in a volume of 20 µl containing 10 µl Platinum SYBR Green qPCR SuperMix UDG (Invitrogen), according to the instructions of the supplier, for quantitative RT (qRT) PCR on a Applied Stepone Plus (Applied Biosystem). Dilutions series (2n with n=0–7) of a mixture of all cDNA within a comparison were used to fix the CT (threshold cycle). Each reaction was done in duplicate and the gene expression levels relative to the 18S rRNA reference gene were calculated by the ΔΔCT method (Schmittgen and Livak, 2008). All the primers used are listed in Supplementary Table S1. Error bars in the figures represent the standard deviation calculated on experimental repetitions.

Plant transformation for complementation

The plasmid used for the production of PPR8522-OE plants contained the backbone of vector pSB11 (Ishida ), a Basta resistance cassette (rice Actin promoter and intron, Bar gene, and Nos terminator) next to the right border, and the PPR8522 coding sequence (primers attB1-PPRF1 and attB2-PPRR1STOP, Supplementary Table S1) under the control of the rice Actin promoter next to the left border. Agrobacterium-mediated transformation of maize inbred line A188 was based on a published protocol (Ishida ). For each transformation event, the number of T-DNA insertions was evaluated by qRT-PCR, and the integrity of the transgene was verified by PCR with primers situated in the AtSac66 terminator downstream of PPR8522 next to the left border.

Results

The emb8522 phenotype is genotype dependent

The embryo-specific mutation emb8522 was originally isolated in the F2 generation of a cross between a maize stock carrying the R-scm2 allele (referred to hereafter as ‘R-scm2’) and an active Mutator stock (Clark and Sheridan, 1991). The recessive mutation was backcrossed to R-scm2 over four generations to reduce the copy number of the mutagenic Mutator transposon (Heckel ) and introgressed into inbred line A188 (five generations), from where it was backcrossed to B73 (two generations). In all three genetic backgrounds, mature emb8522 kernels could be macroscopically distinguished from mature wild-type kernels of the same ear by two distinct features (Fig. 1A): mutant caryopses lacked the imprint of the scutellum and the embryo axis and they had a somewhat translucent and collapsed appearance on the adaxial side. They shared with wild-type caryopses a normal starchy endosperm, both in volume and texture.

Fig. 1.

Embryo and seedling phenotype of emb8522. (A) Morphology of emb8522 caryopses in different genetic backgrounds: mature wild-type (top) or emb8522 kernels (bottom) in R-scm2, A188, and B73 genetic background (left to right) were harvested and photographed adaxial side up. (B) Morphology of emb8522 embryos in different genetic backgrounds at 21 DAP: from left to right: emb8522 embryo in R-scm2, A188, B73, wild-type embryo in A188. (C–E) Growth of wild-type (left half of Petri dish) and emb8522 embryos (right half of Petri dish) from A188 rescued at 16 DAP and cultured on MS medium with 1% sucrose (C), 6% sucrose (D), or Phytagel with 6% sucrose. (F) Green wild-type (left) and albino emb8522 seedlings (right) obtained by embryo rescue at 16 DAP and culture on MS medium with 6% sucrose. Bars, 15mm (A), 6mm (B), 2cm (C–E), and 10cm (F). DAP, days after pollination.

A comparison of hand-dissected emb8522 embryos partially explained these differences, as at 21 DAP, emb8522 embryos showed a quite different morphology depending on the genetic background (Fig. 1B). In R-scm2, mutant embryos formed a tubular structure of seemingly undifferentiated cells, which was not observed during normal embryogenesis. This structure possessed apical-basal polarity but did not differentiate into embryo proper and suspensor. In A188, mutant embryos showed most of the characteristics seen in wild-type embryos but were severely retarded in their development. In B73, the embryo phenotype was even less severe and, apart from their reduced size (around 6-times smaller than the wild type), the overall morphology was quite normal with a well-formed scutellum and embryo axis.

To assess the possibility of an effect of the maternal cytoplasm on the severity of the phenotype, reciprocal crosses between the emb8522 mutant (R-scm2 background) and inbred line B73 were carried out. Independently of the direction of the original cross, an intermediate phenotype was observed, arguing against an influence of the origin of the cytoplasm and its organelles on the emb phenotype.

At germination, tangible differences were observed among the three genetic backgrounds. In R-scm2, emb8522 seeds never germinated, in A188, a single ear was found to carry viable emb8522 seed and in B73 80% of ears segregating the mutation gave rise to albino seedlings. Consequently the original classification of the mutation by Clark and Sheridan (1991) was only valid for the R-scm2 and A188, since the definition of embryo-specific mutations in maize not only includes abnormal embryo and normal endosperm development but also embryo lethality; the second criterion was only partially fulfilled in B73.

emb8522 embryos can be rescued in vitro

Embryo rescue is a means of distinguishing metabolic defects, which can frequently be counteracted by the addition of appropriate nutrients to the culture medium, from developmental defects, which are more difficult to overcome by the supply of exogenous factors. This study cultured 16-DAP emb8522 embryos from all three genetic backgrounds on three different media optimized for specific developmental stages. Mutant embryos from R-scm2 could not be rescued on any medium, whereas emb8522 embryos from both A188 and B73 developed into small plantlets both on medium 2 and medium 3 (Fig. 1C–E). More importantly, all the rescued emb8522 embryos from A188 and B73 gave rise to shoots that were retarded in their growth and presented an albino phenotype (Fig. 1F). The albino phenotype suggested that the gene mutated in emb8522 played a role not only in embryogenesis but also in chloroplast biogenesis and development.

Embryo morphology is altered in emb8522 kernels

To complete earlier microscopic descriptions of mutant and wild-type embryos at 9 and 16 DAP in the R-scm2 background (Heckel ), and to determine precisely the point of onset of the morphological aberrations in the emb8522 mutant, the cytological observations were extended to earlier (3, 5, and 9 DAP) and later (18 and 24 DAP) stages, in both the R-scm2 and A188 backgrounds (Fig. 2). At 3 DAP, emb8522 embryo morphology had already diverged from the wild type in R-scm2 with only 2–3 cells observed in the mutant embryo and 4–5 cells in the wild type (Fig. 2A, 2F). At 5 DAP, the difference in cell number (on average, 15 for the mutant versus 25 for the wild type) and embryo length (120 versus 200 µm) became more pronounced (Fig. 2B, 2G). At 9 DAP, the emb8522 phenotype was somewhat more variable among sister kernels than described by Heckel . Indeed the cell size and cytoplasmic content were not always equal between the upper and lower domains of the embryo, suggesting that apical/basal differentiation was taking place at least in some of the embryos. Although not representative of the majority of mutant embryos at 9 DAP, Fig. 2H was chosen to illustrate this onset of a differentiation into embryo proper and suspensor with smaller apical and larger basal cells. Between 9 DAP and 18 DAP no further development occurred in mutant embryos, although they generally showed a slight increase in length and a more marked increase in width at the base of the suspensor (Fig. 2I). At 24 DAP (Fig. 2J) the embryos were morphologically similar to those found at 18 DAP, but the cell walls had started to degrade and the cells had started to undergo necrosis, never achieving a full differentiation into embryo proper and suspensor, the formation of a protoderm, or the elaboration of shoot and root meristems.

Fig. 2.

Developmental profile of emb8522 embryos. Longitudinal sections of maize caryopses were stained using the periodic acid–Schiff procedure and photographed. Wild-type (A–E) and emb8522 embryos (F–N) in R-scm2 (A–J) or A188 (K–N) were observed at 3 (A, F), 5 (B, G, K), 9 (C, H, L), 18 (D, I, M), and 24 DAP (E, J, N). Bars, 200mm.

In A188, the deviation from wild-type development also started at very early developmental stages (5 DAP, Fig. 2K). However, rather than forming the atypical tubular structure observed in the R-scm2 background, mutant embryos resembled more or less wild-type embryos strongly retarded in their development. At maturity, the emb8522 embryos were about 15-times smaller than their wild-type siblings. This study also observed a stronger heterogeneity of the phenotype in A188 than in R-scm2, the micrographs in Fig. 2K–N being representative of approximately three-quarters of the embryos. Despite their delayed development and reduced size, mutant embryos had at 24 DAP all major attributes of wild-type morphology at 15 DAP: a well-formed scutellum, a clearly defined embryo axis, a continuous protoderm, a coleoptile, and a first leaf primordium (Fig. 2N). Although the presence of a functional shoot and root meristem could not be ascertained on the mere basis of cytological data, the embryo rescue data argue in favour of it.

The emb8522 phenotype is caused by an insertion in a PPR gene

The starting point for the molecular cloning of the emb8522 mutation was a DNA gel blot band of 5kb hybridizing to a MuDR probe and co-segregating with the emb8522 phenotype in a segregating population of 79 individuals (Heckel ). The co-segregating fragment was cloned by the screening of a size-fractionated EcoRI library of genomic DNA from heterozygous plants with a MuDR probe. Nucleotide sequence analysis revealed the presence of 3.3kb of MuDR sequence and 1.7kb of maize flanking sequence in the fragment. Alignment of the flanking sequence with the maize genome sequence (www.maizesequence.org) placed the insertion unambiguously in the last BAC of the long arm of chromosome 1, approximately 100kb from the end of the chromosome (Fig. 3A).

Fig. 3.

emb8522 fine mapping and PPR8522 gene structure. (A) Physical map of the chromosome segment harbouring emb8522 mutation, showing the markers used for fine mapping, their distance from the end of chromosome arm 1L, and the number of recombinants in a population of 431 individuals: green box indicates the position of PPR8522; yellow box indicates the position of the closest predicted gene (www.maizesequence.org). (B) Schematic drawing of PPR8522: exons are designated as rectangles, introns as lines, the position of the flanking sequence tag (FST) by a black bar, and the position of the MuDR insertion by an arrow; names and positions of primers for reverse-transcription PCR are also shown. (C) Sequence comparison of the 10 PPR motifs (REP1 to REP10) found in PPR8522. Identical residues are shaded in black and similar residues in grey. (D) Phylogenetic tree of PPR8522 and the nine closest PPR family members from Arabidopsis (At), rice (Os), sorghum (Sb), Ricinus (Rc), and maize (GRMZM). Numbers are bootstrap values (this figure is available in colour at JXB online).

To provide independent proof for the cloning of the emb8522 mutation, this study undertook fine mapping to exploit sequence polymorphisms between the introgressed region around the emb8522 mutation and the A188 background. By the use of a segregating population of 431 plants, in a first instance the closest publicly available polymorphic simple sequence repeat markers ZCT131, UMC2244, and UMC1797 were mapped. To increase marker density, additional markers (Pr6 and Pr9 in Fig. 3A) were developed that detected small insertion/deletions in the region between UMC1797 and the MuDR insertion. Despite some efforts, this study was not able to find polymorphisms between the MuDR insertion and the end of chromosome 1. The marker ZCT131, located at 7.68Mb from the end of chromosome 1, was only loosely linked to emb8522, since 102 recombinants were found. The number of recombinants dropped considerably closer to the end of chromosome 1: 34 were identified for UMC2244 at 2.64Mb, five for UMC1797 at 836kb, four for Pr6 at 376kb, four for Pr9 at 116kb, and none for the MuDR insertion at 100kb from the chromosome end. These data confirmed that emb8522 was located between Pr9 and the end of chromosome 1, in a region of around 116kb, which contained a single predicted gene. Moreover, this predicted gene, GRMZM5G884466, carried the MuDR insertion in the first of nine predicted exons (Fig. 3B). Among three alternative transcripts predicted in the present release 5b.60 of the maize genome, GRMZM5G884466_T03 was confirmed by experimental sequencing of three RT-PCR clones isolated from B73 kernels (GenBank accession JQ763390).

Whereas these data strongly suggested that the MuDR insertion was the cause for the emb8522 phenotype, they did not formally exclude the possibility that the transposon insertion was only closely linked to the causal mutation, which was situated somewhere else in the 116kb mapping interval. To rule out this possibility, this study generated transgenic A188 plants expressing a full-length cDNA of gene model GRMZM5G884466_T03 (corresponding to the gene structure shown in Fig. 3b) under the control of the rice Actin promoter. Out of 18 independent transformation events, 12 were successfully crossed to the emb8522 mutant in the A188 background. Among the offspring were plants that were heterozygous for the emb8522 mutation and hemizygous for the transgene. These plants were self-pollinated to test whether the emb8522 mutation could be genetically complemented with the GRMZM5G884466_T03 full-length cDNA. The frequency of emb kernels (8.3%) was much closer to the 1:16 (6.25%) ratio expected in the case of complementation than to the 1:4 (25%) ratio expected otherwise. Furthermore, after germination of 120 randomly chosen kernels, this study identified 29 plantlets (24.2%) that were homozygous for the emb8522 mutation and hemizygous for the transgene. All 29 plantlets had normal green colour demonstrating that not only the emb phenotype but also the white leaf colour was complemented by the transgene. PPR8522 expression levels were comparable in non-complemented and complemented plants (Supplementary Fig. S1). Taken together, all these results provided convincing evidence that the MuDR insertion in GRMZM5G884466 was indeed responsible for the emb8522 mutant phenotype.

PPR8522 codes for a PPR protein of the P family

The deduced amino acid sequence of the gene model GRMZM5G884466_P03 disrupted by the MuDR insertion showed highest similarity with the Arabidopsis DG1 protein (Chi ) and the rice Os05g0315100 protein, both members of the pentatricopeptide repeat (PPR) superfamily (Supplementary Fig. S2). The gene was consequently called PPR8522. More detailed sequence analysis showed that the region extending from residue 212 to residue 627 contained 10 PPR motifs that exhibited a variable degree of conservation (Fig. 3C). Moreover, among the 10 PPR motifs of PPR8522, five were in tandem, whereas the remaining ones were interrupted by stretches of several amino acids. These structural characteristics together with the absence of other distinctive domains placed PPR8522 into the P family of PPR proteins, most commonly known to be involved in RNA stabilization and splicing (O’Toole ).

To further explore the relationship of PPR8522 with other family members and in particular with DG1, a phylogenetic tree was constructed (www.phylogeny.fr) of the closest amino acid sequences from Arabidopsis, Ricinus communis, rice, sorghum and maize obtained by a BlastP search of the NCBI non-redundant protein database (http://blast.ncbi.nlm.nih.gov/Blast.cgi; Fig. 3C). The tree (Fig. 3D) demonstrated that PPR8522 was the orthologue of DG1 from Arabidopsis, since PPR8522 was phylogenetically closer to DG1 than to any other Arabidopsis protein and vice versa. The phylogenetic link was nicely supported by functional similarities. Just like emb8522 embryos, dg1 embryos (originally named emb1408 and complementing emb246) do not germinate (Tzafrir ; Meinke ) and give rise to pale seedlings after embryo rescue. However, rescued dg1 plantlets eventually turn green rather than to stay pale and die (Chi ).

PPR8522 is universally expressed

Since only wheat and rice but no maize expressed sequence tags corresponding to PPR8522 were available in the databases, this study set out to provide experimental evidence for the predicted gene in maize and to establish an expression pattern in the maize plant. RT-PCR experiments with various combinations of intron spanning primers (Fig. 3B) confirmed the gene model GRMZM5G884466_T03 and, in particular, the inclusion of exon 1, containing the MuDR insertion in the emb8522 mutant, in the mature mRNA (data not shown).

Spatial expression data were obtained by qRT-PCR experiments with gene-specific primers (Fig. 3B) in a total of 15 different vegetative and reproductive tissues (Fig. 4A). PPR8522 was expressed at rather low levels in all vegetative and reproductive tissues tested. On a relative basis, high expression was observed in the aerial parts of seedlings and juvenile leaf blade, with intermediate expression in all other vegetative and reproductive tissues, with the exceptions of developing kernels between 3 and 20 DAP in which expression was weak (Fig. 4A). A comparison of the expression of PPR8522 and the closely related gene GRMZM2G059449 in seedlings of the maize genotypes B73, A188, and R-scm2 did not reveal any major differences (Fig. 4B). In conclusion, PPR8522 is a ubiquitously expressed gene coding for a PPR protein of the P family.

Fig. 4.

Expression pattern and subcellular localization of PPR8522. (A) Relative expression levels of PPR8522 in different maize tissues established by qRT-PCR: a, aerial parts; r, roots; bj, blade juvenile; sj, sheath juvenile; ba, blade adult; sa, sheath adult; im, immature; m, mature; number in parentheses, days after pollination; error bars represent the standard deviation of triplicates. (B) Relative expression level of PPR8522 and the related gene GRMZM2G059449 in different maize genotypes established by qRT-PCR: error bars represent the standard deviation of triplicates. (C–J) Subcellular localization of PPR8522 (C, G) and PEND (D, H) in leaf epidermal cells of Nicotiana benthamiana. Confocal laser scanning microscopy showing chloroplast autofluorescence in blue and transient expression of GFP (C) and VENUS (H) fusions in green and mCherry fusions (D, G) in red. (F) and (J) are overlays of (C, D, E) and (G, H, I), respectively. Bar, 10 µm.

PPR8522 is targeted to the chloroplast

A common feature of most members of the PPR superfamily is their targeting to organelles (Lurin ). Sequence analysis with the two independent prediction tools Predotar (Small ) and TargetP (Emanuelsson ) clearly assigned PPR8522 to the chloroplast, and the SignalP (Emanuelsson ) tool predicted cleavage of a signal peptide after amino acid residue 24. To obtain experimental confirmation of the predicted subcellular localization of PPR8522, translational PPR8522-GFP and PPR8522-mCherry fusion constructs were generated under the control of the 35S promoter, which were used for transient expression studies in N. benthamiana leaves via infiltration with Agrobacterium strains carrying the construct. In agreement with the computer predictions, confocal laser scanning microscopy of the infiltrated leaf samples showed in vivo co-localization of the PPR8522-GFP (green, Fig. 4C) and PPR8522-mCherry (red, Fig. 4G) fluorescent signals with the blue autofluorescence signal of chloroplasts (Fig. 4E, 4I). The PPR8522 signal was not spread all over the chloroplast but localized to discrete but yet unidentified foci. The signal was reminiscent of the one observed with GUN1 (GENOMES UNCOUPLED1) or pTAC2 (PLASTID TRANSCRIPTIONALLY ACTIVE CHROMOSOME PROTEINS2) fusions, which are likely associated with sites of active transcription on plastid DNA (Koussevitzky ), leading to the hypothesis that PPR8522 may be a plastid nucleoid-associated protein. Co-localization experiments with the nucleoid marker PEND (PLASTID ENVELOPE DNA BINDING, Fig. 4D, 4H) from Arabidopsis (Terasawa and Sato, 2005) revealed only partial overlap between PPR8522 and PEND signals (Fig. 4F, 4J) and did not provide a definite answer as to the identity of the PPR8522 expression foci.

PPR8522 is needed for chloroplast gene expression

To investigate whether the emb8522 mutation had any effect on gene expression in chloroplasts, a microarray analysis of chloroplast gene expression was performed in leaves of 15-day-old wild-type and emb8522 seedlings obtained by embryo rescue in the A188 background. The analysis, which covered the entire chloroplast genome, revealed a considerable reduction in gene expression for the vast majority of chloroplast genes in emb8522 (Fig. 5). A control experiment comparing emb8522 with ivory ppr2-2 rather than green wild-type seedlings showed that this profile was specific to emb8522 and not a profile common to albino seedlings in general (Supplementary Fig. S3).

Fig. 5.

Plastid gene expression in the emb8522 mutant. A microarray containing oligonucleotides corresponding to all 248 plastid genes was hybridized with cDNA probes prepared from emb8522 and wild-type seedlings obtained by embryo rescue. The graph depicts the log2 ratio of mutant and wild-type transcript levels. Plastid genes are represented by bars and regrouped in families, which are labelled (number of genes in parentheses) and delimited by alternation between grey or red colour (this figure is available in colour at JXB online).

When the chloroplast genes were grouped into 12 families based on predicted function (Maier ), nine families were consistently under-expressed in mutant plants. These were the psa (13/14 genes), psb (20/20), atp (17/17), ndh (28/30), pet (17/17), ycf (11/11), rrn (10/10), and trn (52/52) families. In contrast, most members of the orf (10/12), rpl (9/14), and rpo (11/14) families showed increased expression in the emb8522 mutant. Finally the rps (27) family and the remaining group of 10 unclassified genes (three rbc, two cem, two clp, infA, IR/psbA, and matK), contained both up- and downregulated genes. Combining these results with RNAseq data on the transcription of plastid genes by either PEP or nuclear-encoded RNA polymerase (NEP) or both in maize and the closely related cereal barley (Zhelyazkova ) led to the conclusion that the lesion in PPR8522 affected primarily genes transcribed exclusively or preferentially by PEP. The data provided additional support for functional similarities between PPR8522 and its orthologue DG1 from Arabidopsis, since an identical conclusion had been reached based on targeted RNA gel blot analysis of the dg1 mutant (Chi ).

To confirm the microarray data, this study performed qRT-PCR experiments for six genes representative of four gene families. The genes psbA, psbB, psbC, and petA, which are all preferentially transcribed by PEP, were very strongly downregulated (50–400-fold), whereas clpP and rpoB, which are mainly transcribed by NEP (Silhavy and Maliga, 1998), showed much weaker expression differences between wild-type and mutant leaves. Like most NEP-transcribed genes, rpoB was upregulated whereas clpP showed downregulation (Fig. 6).

Fig. 6.

Expression of selected plastid and nuclear genes in the emb8522 mutant. QuantitativeRT-PCR analysis of relative expression levels of the chloroplast genes psbA, psbB, psbC, petA, clpB, and rpoB and the nuclear genes Sig1, Sig2, Sig3, Sig6, TOR, PPR8522, and GRMZM2G059449 in leaves of 15-day-old emb8522 or wild-type (WT) seedlings. The genotype of the seedlings was A188 (top 12 panels) or B73 (last two panels). Error bars represent SD (this figure is available in colour at JXB online).

To investigate the possibility of a retrograde signal on nuclear genes, the expression of genes encoding sigma factors necessary for PEP activity was examined. Particular attention was paid to genes encoding sigma factors targeted to the chloroplast (Sig genes) since the PPR8522 orthologue DG1 had been reported to interact with SIG6 in Arabidopsis (Chi ). Most examined Sig genes were weakly but significantly downregulated in the emb8522 mutant (Fig. 6), the effect being somewhat stronger for Sig6 (8-fold). In contrast, no significant expression differences between wild-type and mutant leaves were detected for TARGET OF RAPAMYCIN (TOR), coding for a kinase linking external and internal cues to growth processes (Dobrenel ). The data suggested that lack of PPR8522 in plastids somehow triggered a retrograde signal to the nucleus that selectively acted on the expression of nuclear genes coding for a class of proteins reported to interact with an orthologue of PPR8522.

Finally, possible effects of the MuDR insertion on PPR8522 expression were examined. The levels of PPR8522 transcript in the mutant leaves were 3.2-fold and 1.3 fold lower than in wild-type leaves of genotypes A188 and B73, respectively (Fig. 6). This decrease was likely the consequence of increased transcript degradation triggered by incomplete translation due to the interruption of the PPR8522 open reading frame by the MuDR insertion. It may have been attenuated by transcriptional readout from the Mu element (Barkan and Martienssen, 1991) or alternative splicing (Vernoud ) with translation of the altered transcript. It needs to be emphasized that, in a first instance, the stronger decrease in A188 does not provide a plausible explanation for the stronger phenotype in A188, since in neither genotype a functional protein is likely to be produced. Interestingly, the expression of the closely related gene GRMZM2G059449 was increased in mutant leaves (Fig. 6), possibly compensating the loss of active PPR8522 protein.

Cellular ultrastructure is altered in the emb8522 mutant

To determine if the observed defects in chloroplast RNA metabolism had repercussions on chloroplast structure, the cellular ultrastructure of emb8522 and wild-type seedlings and embryos was compared in the A188 background. In seedlings, structural changes were expected due to the ivory colour of the mutant leaves and the fact that albinism is frequently the result of dramatic alterations in chloroplast biogenesis (Chi ; Kumari ; Zhou ). Transmission electron microscopy demonstrated that neither mesophyll nor bundle sheath cells of the emb8522 mutant contained normal chloroplasts (Fig. 7A–E). Mutant chloroplasts (Fig. 7B) were less numerous and smaller than wild-type chloroplasts (Fig. 7A), exhibited interruptions in the concentric external membranes and lacked internal membrane structures (Fig. 7D). In all the albino leaves scored, only one single chloroplast with some lamellar structure was found (Fig. 7E). These results suggest that PPR8522 is required for plastid biogenesis including thylakoid membrane formation or maintenance.

Fig. 7.

Cellular ultrastructure of wild-type and emb8522 leaves and embryos. (A-E) Transmission electron micrographs of leaves from 16-day-old wild-type (A, C) or emb8522 (B, D, E) seedlings obtained by embryo rescue. (A) Wild-type cell containing abundant chloroplasts. (B) Mutant cell with plastid-like structures. (C) Close-up of wild-type chloroplasts of mesophyll (top) and bundle sheath (bottom) cells. (D, E) Close-up of chloroplast-like organelles in mutant mesophyll (D) and bundle sheath (E) cells.(F, G) Electron micrographs of representative cells of the shoot apex of 15 DAP wild-type (F) or 21 DAP mutant (G) embryo. ob, oil bodies; sg, starch grains. Bars = 2 µm. (H) Histograms indicating the mean number of oil bodies or starch grains per cell of 15 DAP wild-type or 21 DAP emb8522 embryo shoot apices; the differences are statistically significant at the 1% and 5% levels, respectively (this figure is available in colour at JXB online).

In embryos, this study focused on cells in the shoot apex comparing 15-DAP wild-type embryos to 21-DAP emb8522 embryos in order to analyse developmentally equivalent stages. In these non-photosynthetic cells, only proplastids were present, and the proplastid number and size detected by transmission electron microscopy was very similar between wild-type and mutant embryos (Fig. 7F, 7G). Although the images did not reveal any obvious structural differences between wild-type and mutant proplastids, the small size and the lack of clearly defined structural features for proplastids made it impossible to draw a definite conclusion as to the absence of structural differences between wild-type and mutant proplastids. Nevertheless, two differences were observed between wild-type and mutant cells. The number of starch grains, a differentiated form of proplastids, was increased in the mutant whereas the number of oil bodies was reduced (Fig. 7H).

Discussion

Embryogenesis is a finely regulated process in which a pre-programmed morphology is acquired by asymmetric cell division and elongation and in which tissues and organs are generated by the differentiation of undifferentiated cells. At least in Arabidopsis, the presence of functional plastids is indispensible for this developmental pathway (Tzafrir ; Devic, 2008; Meinke ), possibly because certain enzymatic steps in vital metabolic pathways such as the biosynthesis of fatty acids or phytohormones (gibberellic acid, abscisic acid, and jasmonate) are exclusively accomplished in this organelle (Bryant ). Nucleus-encoded factors play essential roles in the regulation of plastid development, which requires the coordinated expression of both nucleus-encoded and plastid-encoded genes. One of these factors is the nucleus-encoded, chloroplast-targeted pentatricopeptide repeat protein PPR8522 of maize, since its loss leads not only to a strong downregulation of chloroplast genes transcribed by PEP, lack of chlorophyll, and compromised chloroplast ultrastructure in leaves, but also to a pronounced delay or even deviation of embryo development, impairing germination in most genotypes.

Embryo lethality and plastid-targeted proteins in maize

After emb8516 (Magnard ) and lem1 (Ma and Dooner, 2004), emb8522 is the third embryo-specific mutation for which the molecular lesion has been identified in maize. As in the two previous studies, the gene product underlying the emb8522 mutation is a protein targeted to plastids. These results seem to suggest that the viability of the maize embryo strictly depends on the correct functioning of plastids.

However, this conclusion contrasts sharply with data on the maize loss-of-function mutants crs2 (chloroplast RNA splicing2), caf1 (CRS2-associated factor1), caf2, rnc1 (RNAseIII1), wtf1 (what’s this factor1), ppr2, ppr4, ppr5, or why1 (whirly1), which are all affected in the biogenesis of the plastid translation machinery and yet germinate as viable albino seedlings (Ostheimer ; Williams and Barkan, 2003; Schmitz-Linneweber ; Beick ; Prikryl ; Kroeger ). In these mutants, the effect on plastid translation is generally indirect since CRS2, CAF1, CAF2 RNC1, and WTF1 are required for the splicing of group II introns in chloroplasts, PPR4 facilitates the trans-splicing of the chloroplast rps12 pre-mRNA, PPR5 is bound in vivo to the unspliced precursor of trnG-UCC, WHY1 is associated with DNA from the plastid genome and with a subset of plastid RNAs, and PPR2 functions in the synthesis or assembly of components of the plastid translation machinery. One possibility to reconcile the data sets is to propose that not only the mutation under investigation but also the genetic background influences the viability of the mutant embryos. This hypothesis is corroborated by the emb8522 mutation, which is embryo lethal in genotypes R-scm2 and A188 but not in genotype B73. Keeping in mind that genotype B73 was frequently used in the analysis of viable mutants, such as the ones listed above, it would be interesting to introgress these mutations into R-scm2 or A188 and to check the germination frequencies.

Since PPR8522 is required for embryo viability at least in certain genotypes, it is tempting to speculate that it targets the RNA of one or more plastid genes that are essential for survival. Such genes have been identified in dicots where the plastid genomes encode four open reading frames that cannot be knocked out, at least in tobacco: accD, ycf1, ycf2, and clpP (Drescher ; Kuroda and Maliga, 2003), and plastid translation in dicot embryos may simply be required to produce these proteins. In particular, accD involved in fatty acid biosynthesis is considered as the single, most important chloroplast gene required for embryo development in Arabidopsis (Bryant ). However, accD as well as ycf1 and ycf2 are missing in the plastid genomes of cereals and the essential status of clpP has been questioned by the Arabidopsis clb19 mutant in which the lack of clpP editing is not accompanied by an embryo-lethal phenotype (Chateigner-Boutin ). This leaves one more chloroplast gene that is possibly essential in both monocots and dicots, trnE (tRNA-Glu). Beyond its role in translation, tRNA-Glu is also a metabolite needed for 5-aminolevulinic acid (ALA) biosynthesis in plastids (Huang and Wang, 1986; Masuda and Fujita, 2008). ALA is the starting point for the biosynthesis of tetrapyrrole and haem, an essential prosthetic group of proteins involved in electron transport not only in plastids but also in mitochondria (Kanamaru and Tanaka, 2004) In addition, tRNA-Glu is a regulatory molecule inhibiting the transcriptional activity of NEP and mediating the switch in RNA polymerase usage from NEP to PEP during chloroplast development. (Hanaoka ). Consequently PPR8522 may be needed for the PEP-dependent transcription or stabilization of this tRNA.

The role of PPR8522 in embryo development

Independently of the genetic background, embryo development and in particular embryo size is clearly affected in the emb8522 mutant. Although the phenotypic analyses are based on the observation of a single allele due to the lack of additional alleles in sequence indexed reverse genetics collections, they lead to the conclusion that intact plastids are indeed required for normal embryo development in maize and that genetic backgrounds only influence the severity of the phenotype, including the fact whether the embryos are viable or not. The morphological and size differences of the embryos also suggest that the function of PPR8522 extends beyond photosynthesis and raises the question of the type of defect involved. The fact that the parallel development of the endosperm is not affected by the impaired plastids may be a hint for a signalling rather than metabolic defect, since the endosperm shares with the embryo not only the same genetic structure but also a rather similar heterotrophic metabolism.

Another consideration concerns the biogenesis of chloroplasts, more precisely the proplastid–chloroplast transition. In maize embryos, photosynthesis does not occur and plastid development is blocked at the proplastid stage, when the first thylakoid membranes are formed and the photosynthetic complexes are translated, but not yet assembled (Vothknecht and Westhoff, 2001). Light is the signal that triggers the assembly of the thylakoid network and chlorophyll production after germination (Waters and Langdale, 2009). In the emb8522 mutant, the ratio between the two types of proplastids is altered, since there are around 3-times more amyloplasts (sugars storage organs) than etioplasts (future photosynthetic organs), reversing the ratio seen in the wild type. This observation hints at a particular defect in early biogenesis of chloroplasts but not amyloplasts. Taking into account that, in the endosperm, proplastids turn essentially into amyloplasts, the altered transition into chloroplasts may have consequences only in the embryo but not the endosperm.

Genotype-dependent variability of emb8522 embryo morphology

The developmental profile of emb8522 embryos shows fundamental differences depending on the genetic background. In R-scm2 the mutant embryos are atypical tubular structures not seen during normal embryo development that undergo early necrosis (Clark and Sheridan, 1991; Heckel ), in A188 the mutant embryos have quite normal morphology, but are retarded in their development and severely reduced in size, and in B73 mutant embryos have a similar developmental delay as in A188, but are 3–4-times bigger and the majority of them are able to germinate as albino seedlings. The differences between the genetic backgrounds are independent of the direction of the pollination, ruling out cytoplasmic effects. Expression differences do not provide a plausible explanation either because qRT-PCR experiments revealed comparable PPR8522 expression levels in the three genotypes. Copy number variation, a frequently observed phenomenon in the ancient tetraploid maize (Schnable ; Springer ), cannot completely be ruled out as a possible explanation. Whereas B73, the least affected genotype, does not carry a second copy (near isogenic paralogue) of PPR8522 in its published sequence, the sequence is estimated to cover only 96% of the genome (Schnable ). Consequently the possibility remains that B73 carries a second copy that would be missing in the other two genotypes.

The most likely explanation of the phenotypic variability resides in the action of one or several modifier gene(s). Genotypes A188 and R-scm2 may either carry less efficient alleles of such a modifier gene or simply lack it. Maize is renowned for both its high level of single nucleotide polymorphisms within coding regions (Gore ; Chia ) and a substantial frequency of presence/absence variations (Messing and Dooner, 2006; Springer ). With respect to the type of modifier gene which might be involved, one obvious possibility is another member of the vast PPR family, which comprises more than 450 members in both Arabiodpsis and rice (Lurin ; O’Toole ). A potential candidate is the predicted gene GRMZM2G059449, the closest neighbour of PPR8522 in phylogenetic trees. On one hand, the presence of GRMZM2G059449 in B73, A188, and R-scm2 and the comparable expression levels in wild-type plantlets of the three genotypes seem to argue against a role as a modifier gene. On the other hand, the increased expression of GRMZM2G059449 in white emb8522 plantlets compared to green wild-type plantlets in background B73 hints at a compensatory role that may vary between genotypes. Obviously, other types of modifier genes acting either in the nucleus or in plastids cannot be excluded at this point in time.

Role of PPR8522 in chloroplast RNA metabolism

Several indications for the molecular function of PPR8522 can be inferred from sequence analysis, which identified PPR8522 as a member of the PPR protein superfamily. The PPR motif is generally recognized as a RNA binding domain and the function of most characterized family members is related to the stability, splicing, editing, or translation of organelle RNAs (Schmitz-Linneweber and Small, 2008). The tandem arrangement of the 10 PPR domains and the absence of distinctive E, E+, or DWY domains at the C-terminus suggest that PPR8522 belongs to the P subclass of PPR proteins, commonly thought to be involved in RNA stabilization, splicing, and translation rather than editing (O’Toole ).

Both in silico predictions and experimental evidence in N. benthamiana leaves clearly indicate that PPR8522 acts in plastids and not in other organelles such as mitochondria. Within plastids, the protein is located in a subdomain that may be the nucleoid. The nucleoid, which contains the plastid DNA and is the place of transcription, represents a possible destination for PPR proteins depending on their precise role in RNA metabolism. The presence of PPR8522 (under its previous ID AC185612.3_FGP001) in highly enriched nucleoid fractions of maize proplastids and mature chloroplasts (Majeran ) supports the hypothesis of a nucleoid localization. On the other hand, the co-localization experiments with the nucleoid marker PEND (Terasawa and Sato, 2005) revealed only partial overlap between PPR8522 and PEND signals. Beyond experimental pitfalls such as a differential penetration of the various fusion proteins into the compact nucleoid or contamination of the nucleoid fraction in the work by Majeran , the co-localization data could possibly be reconciled with the proteome data by the assumption that nucleoids are spatially complex, dynamic structures and that PEND and PPR8555 were either localized in different parts or different forms of the nucleoid. Alternatively, PPR8522 may not be part of the nucleoid.

Finally, sequence comparisons and phylogenetic analyses identified PPR8522 as the orthologue of DG1 from Arabidopsis (Chi ), which shows 44% sequence identity and 64% sequence similarity with PPR8522. The dg1 mutant (originally named emb1408 and complementing emb246) shares several features with emb8522, namely embryo lethality (Tzafrir ; Meinke ), lower expression of all tested PEP-dependent transcripts, and higher expression of all NEP-dependent transcripts. As for emb8522, dg1 mutant embryos can be rescued in vitro. However, whereas rescued emb8522 plantlets remain albino and die after a few weeks, rescued dg1 plantlets eventually turn green and develop normally (Chi ). In addition, DG1 was found to interact in yeast-two-hybrid, pull-down, and BiFC experiments with SIG6, a sigma factor necessary for the transcription of plastid genes by PEP. Genetic experiments based on the analysis of double mutants and over-expressing plants did not reach clear-cut conclusions and seem to indicate partially overlapping functions of DG1 and SIG6. As suggested for DG1 (Chi ), PPR8522 may associate with SIG6 to stabilize its binding to promoter elements, and the lack of this stabilization may readily explain the strong downregulation of PEP-transcribed genes in dg1 and emb8522 mutants. However, this appealing hypothesis addresses only the function of the C-terminal protein–protein interaction domain in DG1 (Chi ), but not that of the central PPR domain that likely acts as an RNA binding domain. In fact DG1 and consequently PPR8522 may have two functions, one related to its binding to SIG6 and a second one to its binding to one or several yet unknown RNA target(s). These functions may be interconnected, for example SIG6 may direct PPR8522 to nascent RNA molecules. Keeping in mind the opposite trends for PEP- and NEP-transcribed plastid genes in the emb8522 mutant, one may postulate that the target of PPR8522 is a transcript coding for one of the PEP subunits (Pfannschmidt ), a transcript coding for a ribosomal protein, since translation of PEP subunits is limited in ribosome-deficient mutants (Zubko and Day, 2002), or the above-mentioned trnE gene for tRNA-Glu. RIP-Chip or RIP-Seq experiments and parallel yeast-two-hybrid experiments will allow to clarify the type(s) of macromolecules bound by PPR8522.

Supplementary material

Supplementary data are available at JXB online.

Supplementary Fig. S1. Complementation of the emb8522 mutant

Supplementary Fig. S2 Alignment of PPR8522 orthologues in maize, rice, and Arabidopsis

Supplementary Fig. S3. Plastid gene expression in the emb8522 and ppr2-2 mutants

Supplementary Table S1. List of primers used