A photosynthesis operon in the chloroplast genome drives speciation in evening primroses

Abstract Genetic incompatibility between the cytoplasm and the nucleus is thought to be a major factor in species formation, but mechanistic understanding of this process is poor. In evening primroses (Oenothera spp.), a model plant for organelle genetics and population biology, hybrid offspring regularly display chloroplast–nuclear incompatibility. This usually manifests in bleached plants, more rarely in hybrid sterility or embryonic lethality. Hence, most of these incompatibilities affect photosynthetic capability, a trait that is under selection in changing environments. Here we show that light-dependent misregulation of the plastid psbB operon, which encodes core subunits of photosystem II and the cytochrome b6f complex, can lead to hybrid incompatibility, and this ultimately drives speciation. This misregulation causes an impaired light acclimation response in incompatible plants. Moreover, as a result of their different chloroplast genotypes, the parental lines differ in photosynthesis performance upon exposure to different light conditions. Significantly, the incompatible chloroplast genome is naturally found in xeric habitats with high light intensities, whereas the compatible one is limited to mesic habitats. Consequently, our data raise the possibility that the hybridization barrier evolved as a result of adaptation to specific climatic conditions.


Introduction
Incompatibility between the nuclear and organellar genomes represents a mechanism of reproductive isolation that is observed in a wide range of taxa (Levin, 2003;Chou and Leu, 2010;Greiner et al., 2011;Burton et al., 2013;Greiner and Bock, 2013;Sloan et al., 2017;Fishman and Sweigart, 2018;Bogdanova, 2020;Postel and Touzet, 2020). However, with the exception of cytoplasmic male sterility, a commercially important trait, little is known about the molecular and evolutionary mechanisms of cytoplasmic incompatibility (CI). Mechanistic studies are available from only a few cases (Schmitz-Linneweber et al., 2005;Harrison and Burton, 2006;Lee et al., 2008;Burton et al., 2013;Meiklejohn et al., 2013;Fishman and Sweigart, 2018), in contrast with the great biological importance of the phenomenon. Increasing evidence is accumulating that CI arises early in the separation of two genetic lineages (Levin, 2003;Fishman and Willis, 2006;Greiner et al., 2011;Burton et al., 2013;Barnard-Kubow et al., 2016;Postel and Touzet, 2020), and thus represents an initial barrier leading to reproductive isolation. This suggests that CI is a major determinant of species formation, highlighting the need for a mechanistic understanding of its molecular basis in the context of population genetics.
The evening primrose (Oenothera) is a model plant that is uniquely suited to address the mechanisms of reproductive isolation via hybrid incompatibility. Crosses between Oenothera species usually produce viable offspring that regularly display incompatibility between the chloroplast and the nuclear genomes (plastome-genome incompatibility [PGI]). These incompatibilities usually manifest in bleached plants, more rarely in hybrid sterility or embryonic lethality. They represent the only strong hybridization barrier between Oenothera species, which often co-occur in overlapping ecological niches within hybridization zones (Stubbe, 1964;Dietrich et al., 1997;Greiner et al., 2011;Figure 1). Hybrid incompatibility of nuclear loci is essentially absent (Stubbe and Raven, 1979b;Greiner and Köhl, 2014), underscoring the importance of CI as the cause of incipient isolation of the hybrid. In addition, Oenothera is a prime example for hybrid speciation (Cleland, 1957;Hollister et al., 2019), in that permanent translocation heterozygosity, a form of cross-inducible functional asexuality, can occur. Such genotypes display a meiotic ring and breed true upon self-fertilization. In crosses, this can lead to an immediate fixation of a hybrid (Cleland, 1972;Harte, 1994;Rauwolf et al., 2008;Golczyk et al., 2014) (see "Methods" and below).
In evening primroses (section Oenothera subsection Oenothera), based on phenotype, phylogeography, and behavior in crossings, three genetic lineages (A, B, and C) exist that are genetically separated by PGI (Stubbe, 1964;Cleland, 1972;Greiner et al., 2011). These genetic lineages occur in either homozygous (AA, BB, or CC) or in a stable heterozygous (AB, AC, or BC) state. They can be combined with five basic chloroplast genome types (I-V), which differ in their compatibility relations to the nuclear genotypes ( Figure 1B).
The presence of distinct nuclear and chloroplast genomes and the sexual separation of the species by PGI have led to the development of a genetic species concept describing 13 species for Oenothera (Stubbe, 1964;Raven et al., 1979;Stubbe and Raven, 1979b;Dietrich et al., 1997). This species concept (of the so-called American school) is in sharp contrast to the largely morphological species definition of the European school that recognizes about 80 species (Rosta nski, 1982;1985;Rosta nski et al., 2010;Wo zniak-Chodacka, 2018). According to the American school, however, specific combinations of nuclear and chloroplast genomes define the species (Dietrich et al., 1997). Other genome combinations can occur in weak or inviable hybrids, thus sexually separating the species ( Figure 1B). Strikingly, these chloroplast-mediated speciation barriers rely on photosynthesis, a trait that is under selection in changing environmental conditions (Arntz and Delph, 2001;Flood, 2019). This makes Oenothera an appealing model to understand the genetic basis of speciation. The incompatibility loci causing separation of the species are part of the species concept and, hence, relevant for speciation by definition. They may even be a result of adaptive evolution.
For example, hybridization between Oenothera elata (an AA-I species) and Oenothera grandiflora (a BB-III species) produces the incompatible combination AB-I that displays a bleached-leaf phenotype. This genetic incompatibility creates the major hybridization barrier between AA and BB species that appears to prevent colonization of western North America by B genome species ( Figure 1C; Greiner et al., 2011). It should be emphasized that hybridization is frequent in the genus in nature. As a rule, hybridization occurs if plastome-genome combinations permit it. Importantly, all viable combinations can be confirmed in hybrids in nature (Dietrich et al., 1997). Hence, PGI appears to act as a major mechanism preventing gene flow. In addition, ecological species separation is likely to occur. For example, existence of the green and compatible hybrid AA-II can be confirmed, but it does not establish stable populations in nature (Dietrich et al., 1997;Figure 1, A and B).
It is assumed that climate changes and periods of glaciation during the Pleistocene have shaped the genetic and ecological characteristics of the basic lineages A, B, and C in Oenothera (Cleland, 1972). This is well supported both by the estimated divergence time of the chloroplast genomes (Greiner et al., 2008b) and by nuclear genome variation (Levy and Levin, 1975;Hollister et al., 2015). Following this view, the three lineages originated from Middle America and reached the North American continent in several waves. The current lineages resemble the ancestral sexual and homozygous species AA-I, BB-III, and CC-V, and crosses between them usually result in PGI (Stubbe, 1964;Greiner et al., 2011). However, during the Pleistocene, hybridization between the basic lineages must have happened that produced viable offspring (Cleland, 1957(Cleland, , 1972Dietrich et al., 1997). Those were fixed in the structural heterozygous and functional asexual species (AB-II or BA-III, AC-IV, and BC-IV). Hence, plastomes II and IV can be seen as relict genotypes of earlier stages of plastome evolution (Stubbe, 1964;Dietrich et al., 1997). Interestingly, those relict genotypes are now quite likely maintained in adaptive niches, Figure 1 Distribution of Oenothera AA-I, BB-III and AB-II/BA-III species, and compatibility/incompatibility relations upon hybridization. A, Native distribution of Oenothera A and B genome species (pure colors) and their hybridization zones (two-color pattern). Yellow and magenta gradients indicate occurrence of distinct species or subspecies. B, Genetic species concept of evening primroses, based on plastome/nuclear genome compatibility and incompatibility, exemplified for the A and B genome species. Species are defined by their combinations of nuclear and chloroplast genomes (boxed in magenta), and are genetically separated by PGIs that occur upon hybridization and vary in the severity of the hybrid phenotype (BB-I white, AB-I and BB-II yellow-green, and AA-III bleaching leaf phenotype). C, Association of AA-I and B genome (BB and AB) species of Oenothera to the xeric and mesic habitats of North and Central America. Distribution maps redrawn from Dietrich et al. (1997). Climate data are from SolarGis and North American Environmental Atlas. since they are not outcompeted by the more recent, faster multiplying and more aggressive, chloroplast genomes (Stubbe, 1963(Stubbe, , 1964: Biparental chloroplast inheritance, which is the rule in evening primrose, leads to selfish evolution of aggressive or more competitive (e.g. faster multiplying) chloroplast genotypes upon hybridization (Grun, 1976;Greiner et al., 2015;Sobanski et al., 2019). In Oenothera, the most competitive chloroplasts are fixed as current evolutionary end points in the homozygote lineages AA-I, BB-III, and CC-V. These competitive chloroplast genotypes (I, III, and V) have the potential to outcompete the weaker ones (II and IV) in hybrid genotypes. For example, the common evening primrose (Oenothera biennis) harbors plastomes II and III in overlapping subpopulations (Cleland, 1972;Dietrich et al., 1997). Since plastids with plastome III are more aggressive than those with plastome II, they should outcompete plastome II in this species. Observed, however, is that separation of plastomes II and III in the two major subpopulations coincides with the expansion of the Wisconsin glacier (see Supplemental Figure S3 of Anstett et al., 2015; Figure 1A). This suggests that post-glaciation dispersal events and the presence of an ecological niche maintain the less aggressive chloroplast type in the northern parts of the O. biennis population.
Taken together, even recent plastome divergence in evening primrose appears to be a consequence of glaciation that is fixed by an interplay of adaptive and selfish evolution. The influence of genetic drift on hybrid incompatibly among the Oenothera populations is currently unknown.
Our aim was to understand the mechanism of the AB-I incompatibility that genetically separates the A and B lineages ( Figure 1). We were further asking to what extent the underlying incompatibility might have evolved as a result of adaptive divergence in the two lineages. Based on association mapping in the chloroplast genome, the dual promoter region in the intergenic spacer between the clpP operon and the psbB operon was proposed to be involved in the incompatibility AB-I (Greiner et al., 2008a). The clpP gene encodes the proteolytic subunit of the Clp protease, the psbB operon encodes core subunits of photosystem II (PSII) and the cytochrome b 6 f complex (Cytb 6 f). However, the molecular mechanism and the physiological effects underlying the incompatibility have remained enigmatic. In addition, it was unclear, if the ClpP protease is involved in the phenotype of the incompatibility.

Results
AB-I plants are unable to acclimate to higher light conditions AB-I plants display a yellow-green (lutescent) leaf chlorosis, caused by disturbed PSII activity (Greiner et al., 2008a;Figure 2, A and B). These photosynthetic defects occur specifically under high light (HL) intensities ( Figure 2B). At 300 mE m À2 s À1 (low light [LL]), the compatible wild-type AB-II and the incompatible hybrid AB-I are indistinguishable from each other, but higher light intensities cause severe photodamage in AB-I. Consistent with our previous study, already at 450 mE m À2 s À1 (HL), a substantial portion of PSII was photodamaged. Interestingly, AB-I plants are also unable to perform an efficient light acclimation response when shifted to HL conditions (Figure 2B; see Supplemental Text  in the Supplemental Data pdf file containing the  Supplemental Figures and Supplemental Tables): Whereas AB-II plants responded to the increased growth light intensity by strongly increasing their chlorophyll content (Supplemental Figure S1A) and the contents of all redox-active components of the electron transport chain ( Figure 2B), AB-I plants were incapable of performing this light acclimation response efficiently. This behavior, inhibited in AB-I plants, is a typical reaction when plants that were previously grown under light-limited conditions are transferred to higher light intensities (Schöttler and Tóth, 2014). Finally, this leads to a relative reduction of the components of the electron transport chain, namely PSI, PSII, and Cytb 6 f, but not the ATP synthase and plastocyanin, in AB-I plants compared to AB-II ( Figure 2, B and C; Supplemental Figure S1; Supplemental Text). In summary, AB-I plants display a light-dependent phenotype of photosynthetic acclimation that cannot be assigned to a single component of the electron transport chain. In addition, the disturbance in acclimation response is independent of ATP synthase and PC function.
The better adaptation of wild-type AA-I plants to HL is conferred by the chloroplast genotype To examine whether the genetic differences between plastomes I and II also have phenotypic effects under HL conditions in a compatible background (i.e. could be subject to selection in the parental species), we compared the lightacclimation response of green wild-type AA-I (O. elata) plants with green wild-type AB-II (O. biennis). In addition, to investigate whether potential differences are due to the chloroplast genome, we included the green chloroplast substitution line of the two genotypes (AA-II; Figure 1B). Since this experiment involved green material only, a harsher and more natural light shift from 300 to 600 mE m À2 s À1 (harsh HL [HHL]) could be analyzed. The latter condition already induces severe damage in the incompatible AB-I genotype (Supplemental Text). However, it should be emphasized that 600 mE m À2 s À1 is still a relatively moderate light intensity when compared to 1,000 to 1,500 mE m À2 s À1 that can be reached in open land at noon.
In 300 mE m À2 s À1 , all photosynthetic parameters investigated (chlorophyll a/b ratio, chlorophyll content, F V /F M , linear electron transport capacity, and chloroplast ATP synthase activity) were very similar among the three genotypes (AA-I, AA-II, and AB-II; Table 1). However, after the shift to HL, pronounced differences were observed. While most parameters were not, or only weakly, affected in AA-I plants, the AB-II genotype showed a drastic loss of electron transport capacity. This was accompanied by marked decreases in the chlorophyll a/b ratio, chlorophyll content per leaf area, and chloroplast ATP synthase activity. Strikingly, similar changes also occurred in AA-II plants, indicating that the genetic background of the plastome, and not the nuclear genome, causes these differences in light acclimation. However, in comparison to typical light-acclimation responses of angiosperms, which (due to degradation of the chlorophyll b binding antenna proteins) result in increased electron transport capacity, chlorophyll content, and chlorophyll a/b ratio (Schöttler and Tóth, 2014), the response of AA-I plants to increased light intensity is limited.
It, therefore, can be concluded that plastome I is better adapted to cope with HL conditions than plastome II, although at least under the conditions tested, the johansen Standard strain of O. elata, originally isolated in California (Cleland, 1935;Dietrich et al., 1997), does not behave like a typical HL plant.

RNA editing is not involved in the AB-I incompatibility
Chloroplast loci that cause the described phenotypes could be related to mRNA editing sites, which often display great Plants were cultivated at either 300 mE m À2 s À1 or 600 mE m À2 s À1 actinic light intensity. Values are means 6 SD (n ¼ 5-6 plants grown in parallel). Different lowercase letters indicate significant differences (P < 0.05) according to two-way ANOVA with interactions followed by Tukey's post-hoc testing (Supplemental File S4).
variability among even closely related species (Schmitz-Linneweber et al., 2005;Kahlau et al., 2006). RNA editing in chloroplasts of seed plants involves C-to-U conversions at highly specific sites (Takenaka et al., 2013). It is of particular interest, since the only previously described mechanism of PGI is based on an editing deficiency of the tobacco (Nicotiana tabacum) atpA transcript (encoding a core subunit of the plastid ATP synthase) when exposed to the nuclear genetic background of the deadly nightshade Atropa belladonna (Schmitz-Linneweber et al., 2005). However, as evidenced by our sequencing results of the chloroplast transcriptomes of O. elata (AA-I), O. biennis (AB-II), and O. grandiflora (BB-III), mRNA editing does not play a role in the AB-I incompatibility. As judged from our data, all compatible wild-type genome combinations of the three species share the same 45 mRNA editing sites (Supplemental Table S2 and updated GenBank records AJ271079.4, EU262889.2, and KX014625.1). This analysis also includes partially edited sites, whose biological relevance is doubtful (see Supplemental Table S2 and "Methods" for details). These results exclude the possibility that editing sites in the plastome and/or nucleus-encoded editing factors differ among the genotypes involved in the AB-I incompatibility, as this was reported for an experimentally produced cytoplasmic hybrid (cybrid) of Atropa and tobacco (Schmitz-Linneweber et al., 2005).

Association mapping to identify plastid loci causing the AB-I incompatibility
Having ruled out the involvement of mRNA editing, we performed an association mapping in the chloroplast genome of Oenothera to pinpoint the causative loci for the AB-I incompatibility. In contrast to the green alga Chlamydomonas, chloroplast genomes of higher plants are not amenable to linkage mapping (Chiu and Sears, 1985;Greiner et al., 2015). Hence, identification of functionally relevant loci is usually based on correlation of a polymorphism to a phenotype (e.g. Greiner et al., 2008a;Simon et al., 2016;Sobanski et al., 2019). In the case of the AB-I incompatibility, this can be achieved by manual inspection of an alignment of fully sequenced chloroplast genomes and a search for specific polymorphisms in plastomes I versus II, III, and IV. Those polymorphisms are considered candidates for causing the AB-I incompatibility because only plastome I confers the bleached lutescent phenotype in the AB nuclear genetic background, whereas plastomes II, III, and IV are all green when combined with the same nuclear genome (Stubbe, 1964;Greiner et al., 2008a). Our original analyses of the AB-I phenotype had included only 4 chloroplast genomes and yielded 16 candidate regions (Greiner et al., 2008a).
Using the power of next-generation sequencing (NGS) technologies, we were able to base the association mapping on 46 full chloroplast genomes whose genetic behavior had been determined by extensive crossing studies (Stubbe, 1959(Stubbe, , 1960(Stubbe, , 1963Wasmund, 1980) ("Methods"; Supplemental Table S1). The chosen strains represent the material used for generalization of the genetic species concept in Oenothera that is based on the basic A, B, and C nuclear and I-V chloroplast genotypes (Cleland, 1972;Dietrich et al., 1997) (see "Introduction"). The strains reflect the full natural distribution range of Oenothera and cover all species accepted at the time the crossing data were gathered. Altogether, the mapping panel included 18 chloroplast genomes representing plastome type I ("Methods"; Supplemental Table S1). Although the individual plastomes differ by various single-nucleotide polymorphisms and insertions/deletions (indels) (cf. Greiner et al., 2008aGreiner et al., , 2008b, only four polymorphisms were absolutely linked with the AB-I phenotype, that is, they were specific to plastome I and could potentially be involved in the AB-I incompatibility: (1) a 144-bp deletion in the clpP-psbB operon spacer region, (2) a combined 5-bp deletion/21-bp insertion in the psbM-petN spacer (genes encoding a PSII and a Cytb 6 f subunit, respectively), (3) a 194 bp deletion in the ndhG-ndhI spacer (two genes encoding subunits of NADH dehydrogenase complex), and (4) a 21-bp insertion in the trnL-UAA-trnT-UGU spacer (Supplemental Files S1-S3).
Due to the lack of measurable sexual recombination frequencies in chloroplast genomes of seed plants  (see above), genetic methods cannot be employed to further narrow down the causative loci for the AB-I incompatibility in plastome I. We, therefore, evaluated the remaining candidate polymorphisms with respect to their potential for causing the incompatible phenotype. Based on the deletions in the ndhG-ndhI and trnL-UAA-trnT-UGU spacers it was difficult to explain the observed light-dependent reduction of specific photosynthetic complexes in AB-I incompatible material ( Figure 2). The neighboring genes do not encode components of the electron transport chain, and, moreover, knockouts of NDH complex subunits lack a discernible phenotype (Burrows et al., 1998). From possible effects on the expression of trnL-UAA and/or trnT-UGU, two essential tRNAs, one might expect a more pleiotropic phenotype that does not depend on the light intensity. As judged from the functions of the genes affected, a contribution of the latter two polymorphisms to the AB-I phenotype is hence unlikely, although strictly speaking we cannot fully exclude this. In contrast, the polymorphisms affecting the psbB operon and the psbM/petN spacer are more likely candidates because they can potentially affect both PSII and Cytb 6 f, in agreement with the physiological data ( Figure 1B).

The psbN-petN spacer region makes a minor contribution to the AB-I incompatibility
To examine the contribution of the combined 5-bp/21-bp indel in the psbM-petN spacer, transcript and protein analyses were performed in incompatible AB-I plants and compatible controls under LL and HL conditions ( Figure 3). The indel is located in the 3'-UTR of both genes ( Figure 3A) and, therefore, could potentially affect the stability of their transcripts.
RNA blot analyses revealed that both genes are affected by the indel. For psbM, reduction of the 0.35-kb monocistronic transcript was observed for AA-II and AB-I under LL, and for AB-II and AB-I under HL conditions. Although there is no obvious explanation for this light-dependent effect, it is independent of the AB-I incompatibility because the levels of psbM mRNA cannot be linked to the AB-I phenotype ( Figure 3B and below). Moreover, as judged from knockout mutants in tobacco, even complete loss of the PsbM protein does not lead to a strong phenotype that would be comparable to the phenotype of AB-I plants (Umate et al., 2007). In contrast, petN encodes an essential subunit of the Cytb 6 f (Hager et al., 1999;Schwenkert et al., 2007), and reduced petN transcript stability, therefore, could affect Cytb 6 f accumulation. RNA gel blot analysis of petN mRNA accumulation detected a mature transcript of 0.3 kb ( Figure 3A). Under LL conditions, petN transcript accumulation is unaltered in the incompatible hybrid, whereas under HL, the petN mRNA is significantly reduced in AB-I material ( Figure  3C). Immunoblot analyses showed that this leads to a reduction of the protein level to $80% ( Figure 3D), an estimate that is well supported by our spectroscopic quantification of Cytb 6 f ( Figure 1B).
Taken together, these data do not exclude the possibility that the psbM/petN region influences the incompatibility phenotype, but suggest a rather minor contribution.  Figure 3 Molecular genetic analyses of the psbM-petN spacer region in compatible (AA-I, AA-II, and AB-II) and incompatible (AB-I) material under HL and LL conditions. A, Sequence context and indels (blue triangle) in the spacer that is specific to plastome I. Arrows indicate transcription start sites. B, C, RNA gel blot analyses of psbM (B) and petN (C) transcript accumulation under LL and HL conditions. These experiments were performed three times with similar results. D, Immunoblot analysis of PetN accumulation under HL. N.t. ¼ N. tabacum wild type, DpetN ¼ petN knockout in N. tabacum (Hager et al., 1999). 100% corresponds to 5 mg chlorophyll equivalent. Tobacco control lines were grown in tissue culture as described in "Methods." The experiment was performed independently three times with similar results.
However, involvement of psbM is unlikely, because downregulation of its mature transcript is observed also in compatible AA-II plants under LL conditions ( Figure 3B). Nonetheless, a role of petN is unlikely as well, since reduction of $20% of the Cytb 6 f content ( Figures 2B and 3D) does not affect accumulation of the PSs (Anderson et al., 1997;Hager et al., 1999;Schwenkert et al., 2007;Schöttler et al., 2007b). Hence, another chloroplast locus is responsible for the AB-I incompatibility.
The promotor region of the psbB operon is a major determinant of the AB-I incompatibility Next, we analyzed the transcript patterns of the clpP and psbB operons that flank the 144-bp deletion in the spacer region (Supplemental Figure S2A). RNA gel blot analyses revealed that accumulation of both the clpP precursor transcript and the mature clpP mRNA did not differ in control plants versus incompatible plants under HL conditions. All clpP transcripts accumulated to similar levels as in the compatible lines. Similarly, no difference in transcript accumulation of the remaining operon genes residing upstream of clpP (rpl20 and 5'-rps12) was observed (Supplemental Figure S2B). In addition, analyses of ClpP protein accumulation and the integrity of the plastid ribosomes revealed no difference between compatible and incompatible material (Zupok, 2015). Based on these findings, a contribution of the clpP operon to the incompatibility phenotype can be excluded.
In contrast, transcript accumulation of all psbB operon genes (psbB, psbT, psbH, petB, and petD) was found to be reduced in AB-I plants under HL, but not LL conditions (Figure 4, B and F; Supplemental Figure S2). Run-on transcription analyses revealed that this effect was due to impaired transcription rather than being an effect of altered transcript stability. Transcription of the psbB operon was specifically reduced under HL conditions in the incompatible hybrids ( Figure 4D). Consequently, in contrast to the green AA-I, AA-II, and AB-II plants, the deletion in plastome I in the AB background affected regulation of the psbB operon promoter in a light-dependent manner ( Figure 4D). Importantly, the same promoter was used in all genetic backgrounds, as evidenced by mapping of the transcription start sites ( Figure 4C), which are also highly conserved among species ( Figure 4A). The deletion does not affect the TATA box of the psbB operon promoter, but resides 7-bp upstream of the À35 box. This may suggest that polymerase binding is not affected, but instead that binding of auxiliary proteins such as sigma factors is impaired by the deletion in the incompatible hybrids (see "Discussion").
Interestingly, photosystem biogenesis factor 1 (pbf1, previously designated psbN), a gene involved in PSI and PSII assembly (Krech et al., 2013), is downregulated in AB-I incompatible plants under HL conditions ( Figure 4E). pbf1 is transcribed from the opposite strand by its own promoter, which lacks any polymorphism in all Oenothera plastomes sequenced so far (Supplemental Files S1-S3). Therefore, the reduction in pbf1 transcript accumulation might result from the sense-antisense interaction with the psbT mRNA, as previously described for Arabidopsis (Zghidi-Abouzid et al., 2011;Chevalier et al., 2015). Alternatively, it could be the result of a type of unknown feedback regulation. In any case, the interaction results in a strong reduction of Pbf1 accumulation ( Figure 4F) that is a result of a decreased translation activity of the reduced pbf1 transcript and its antisense counterpart psbT ( Figure 4H). Since pbf1 knockouts are extremely light-sensitive and show severe defects in PSII and, to a lesser extent, also in PSI accumulation (Krech et al., 2013;Torabi et al., 2014), it appears likely that the effect on the pbf1 mRNA also contributes to the incompatibility phenotype.

Discussion
PGI was reported from a large number of plant species and is an important phenomenon in plant speciation (Greiner et al., 2011). Our work reported here shows that light-dependent misregulation of a core photosynthesis operon leads to hybrid incompatibility, thus causing reproductive isolation and ultimately, speciation. Interestingly, the underlying genetic architecture was shaped during the last ice age by periods of glaciation, as noted in the Introduction. The mechanism we have discovered is different from that of the two other cases of PGI mechanistically studied so far. RNA editing of the atpA transcript was identified as the cause of chloroplast-nuclear incompatibly in an Atropa/tobacco synthetic cybrid (Schmitz-Linneweber et al., 2005) (see above). Variation in the coding regions of accD (the plastid-encoded subunit of the acetyl-CoA carboxylase, catalyzing the first step of fatty acid biosynthesis) was suggested as a genetic determinant of PGI in pea (Bogdanova et al., 2015). However, the Atropa/tobacco case represents an unnatural, artificial combination of the plastid and the nuclear genomes of two sexually incompatible species, and, unfortunately, the evidence for possible causative loci for the incompatibility in pea are currently not strong enough to judge their impact on natural populations (Nováková et al., 2019). Moreover, in both cases, the ecological relevance of the suggested PGI loci is unclear and cannot be deduced from the identified polymorphisms.
In contrast, our data indicate that the AB-I incompatibility might have evolved as a result of ecological selection, representing a genotype Â genotype Â environment interaction (GÂGÂE). Although cytoplasmic GÂE interactions were reported earlier in bread wheat (Triticum aestivum; Ekiz et al., 1998), especially for the chloroplast, cases in plants are hardly described. Recent work in animals, however, indicates that GÂGÂE interactions may be relatively common (Hoekstra et al., 2018;Hill et al., 2019;Rand and Mossman, 2020). That cytoplasmic incompatibly, in turn, can result from ecological selection is obvious from work in sunflower (using a cross between Helianthus annuus and H. petiolaris), in which common garden experiments in xeric and mesic habitats demonstrated maintenance of CI by positive selection (Sambatti et al., 2008). The underlying genes and physiology, however, have remained enigmatic.
Our study demonstrates that photosynthesis-related genes encoded in the chloroplast genome can establish hybridization barriers. The incompatible phenotype is only visible under HL condition-AB-I plants cannot perform the necessary acclimation response as described above. Strikingly, plastome I in the native AA nuclear genetic background of O. elata (a species adapted to the western United States and Mexico) copes better with HL conditions than does the AB-II genotype of O. biennis, which is native to the North American woodlands (Dietrich et al., 1997). This effect is plastome-dependent, since the green AA-II chloroplast substitution line behaves photosynthetically very similar to O. biennis, but fully resembles the O. elata morphotype.
However, to what extent the identified incompatibility locus is involved in a light acclimation response of AA-I species in their natural habitats remains to be addressed in further investigations.
Within their natural distribution ranges, species carrying plastome I colonized central and southwest North America (xeric habitats with HL irradiation), whereas species carrying plastomes II, III, IV, or V are limited to the mesic sites of eastern North America ( Figure 1C) (Cleland, 1972;Dietrich et al., 1997). Hence, plastome I seems to be required for colonization of habitats exposed to higher irradiation. This assumption is further supported by the fact that O. biennis (AB-II or BA-III), a species that spread west of the Great Plains after 1970, is only rarely found in the southern parts of the United States and is still absent from Mexico (Cleland, 1972;Dietrich et al., 1997).
Consequently, the loci underlying the AB-I incompatibility seem to prevent colonization of the south western parts of North America by the B genome by creating an asymmetric hybridization barrier between AA-I and AB-II, BA-III, and BB-III species. At the same time, the compatible nuclear-chloroplast genome combination AA-I may have facilitated physiological adaptation of the corresponding species by nuclearcytoplasmic coevolution. Therefore, it seems reasonable to assume that, as a result of higher light intensities (and/or light quality differences) in xeric habitats, the deletion upstream of the psbB operon promoter coevolved with nucleus-encoded proteins that interact with the (bacterialtype) plastid-encoded RNA polymerase (PEP).
Strong candidates for these interacting proteins are the PEP sigma factors, which were shown to regulate polymerase binding in response to both light quality and light quantity (Noordally et al., 2013;Chi et al., 2015). Moreover, regulation by sigma factors (e.g. through redox-induced phosphorylation) influences the stoichiometry of the protein complexes of the photosynthetic electron transport chain (Shimizu et al., 2010). Thus, the failure of AB-I plants to acclimate to HL intensities could be a direct consequence of disturbed transcriptional regulation by sigma factors. Coevolution and coordinated rates of molecular evolution of PEP core subunits and sigma factors could be a common principle in plant evolution, as suggested by recent findings in Geraniaceae, a family in which PGI is also widespread .
Finally, it should be emphasized that, although the psbB operon does not encode PSI-related genes, its transcriptional misregulation also explains the observed effect on PSI, due to either the antisense interaction with the pbf1 mRNA or an unknown mechanism of feedback regulation (Figure 4). Interestingly, while the psbB operon (including the pbf1 gene on the opposite strand) displays extremely high structural conservation from cyanobacteria to higher plants, light regulation of pbf1 transcript abundance was shown to be highly variable among species (Plöchinger et al., 2016).

Plant material
Throughout this work, the terms Oenothera and evening primrose refer to subsection Oenothera (genus Oenothera, section Oenothera, Onagraceae; 2n ¼ 2x ¼ 14) (Dietrich et al., 1997). Plant material used here is derived from the Oenothera germplasm resource harbored at the Max Planck Institute of Molecular Plant Physiology (Potsdam-Golm, Germany). This includes the living taxonomic reference collection of subsection Oenothera (Greiner and Köhl, 2014). Part of this reference collection is the Renner Assortment, a medium-sized collection of European lines that were thoroughly characterized by members of the genetic school of Otto Renner (Cleland, 1972;Harte, 1994). In addition, it includes the Cleland collection, a large set of North American strains of subsection Oenothera that was extensively studied by Ralph E. Celand and colleagues (Cleland 1972). Also present are North American accessions analyzed by Wilfried Stubbe et al., and these represent species of this subsection that were recognized later than the 1960s (Stubbe and Raven, 1979a;Stubbe, 1984, 1986;Wasmund and Stubbe, 1986;Wasmund, 1990;Schumacher et al., 1992;Schumacher and Steiner, 1993;Stubbe and Steiner, 1999). The availability of this material allowed us to employ the original source of lines on which the genetic species concept of subsection Oenothera was based (cf. Cleland, 1962) The lines employed for association mapping of the plastidic AB-I locus were extensively analyzed by classical genetics for the compatibility relations of their nuclear and chloroplast genomes (Supplemental Table S3 for details).
RNA editing analyses were performed with the wildtype strains of O. elata subsp. hookeri strain johansen Standard (AA-I), O. grandiflora strain Tuscaloosa (BB-III), and O. biennis strain suaveolens Grado (AB-II). Supplemental Table S3 contains a summary of all wild-type strains used in this work.
For most other genetic or physiological work presented here, the wild-type strains johansen Standard (AA-I) and suaveolens Grado (AB-II), or chloroplast substitution lines between them (AA-II and AB-I) were used. Here, AA-I refers to the wild-type situation, that is, strain johansen Standard with its native nuclear and chloroplast genomes. AA-II refers to the nuclear genome of johansen Standard combined with the chloroplast genome of suaveolens Grado. AB-II designates nuclear and chloroplast genomes of the wild-type strain suaveolens Grado, and AB-I the nuclear genome of suaveolens Grado equipped with the chloroplast genome of johansen Standard. Generation of AA-II and AB-I from the wild-types AB-II and AA-I is detailed below (see the "Summary" in Supplemental Table S4).
For ribosomal profiling and RNA-seq analyses incompatible (AB-I) and compatible (AB-II) plants were obtained from F1 seeds of crosses between the wild-type or mentioned chloroplast substitution lines of johansen Standard (AA-I or AA-II) as the seed parent and grandiflora Tuscaloosa (BB-III) as the pollen parent. Here, AA-I Â BB-III yielded AB-I plants, and AA-II Â BB-III produced AB-II plants. Selection against the paternally transmitted plastome III was done using appropriate markers (Rauwolf et al., 2008). The material produced largely recreates the AB-I genotype that was characterized by Greiner et al. (2008a). For the tobacco (N. tabacum) wild-type, the cultivar Petit Havana was used. The tobacco DpetN mutant was obtained from Hager et al. (1999).

Generation of chloroplast substitution lines
In Oenothera, the genetics of permanent translocation heterozygosity, combined with a biparental transmission of plastids, offer an elegant opportunity to substitute chloroplasts between species in only two generations while leaving the nuclear genome constitution unaltered (Stubbe, 1959(Stubbe, , 1960(Stubbe, , 1989. For the interested reader, general principles, including a detailed discussion of crossing examples, are presented in Rauwolf et al. (2008). The chloroplast substitution between the strains suaveolens Grado and grandiflora Tuscaloosa (described in Figure 6 of Rauwolf et al. 2008) resembles the chloroplast substitution between suaveolens Grado and johansen Standard used in this work.
In brief, due to reciprocal chromosomal translocations, many species of Oenothera form permanent multichromosomal meiotic rings. If all members of a given chromosome complement are involved in a single ring, they establish two regularly segregating sets of genetically linked chromosomes. This leads to formation of two superlinkage groups, each involving one complete parental haploid chromosome set (a and b). Suppression of homologous recombination avoids genetic reshuffling between the two haploid sets. Additional genetic properties, especially presence of gametophytic lethal factors that lead to sex-linked inheritance of a given haploid set, eliminate homozygous segregants (aÁa or bÁb). This results in permanent heterozygous progeny (aÁb) that is identical to the parental plant. The phenomenon of structural heterozygosity is a form of functional asexuality. However, truly sexual species also exist in Oenothera, that is, species that display bivalent-pairing and regular meiotic segregation. In contrast to the structurally heterozygous species, they lack lethal factors and are homozygous for their haploid sets (haploÁhaplo versus aÁb from above; Cleland, 1972;Harte, 1994).
As a consequence of this genetic behavior, entire haploid chromosome sets in evening primrose can behave as alleles of a single Mendelian locus. These so-called Renner complexes are designated using (Latin) names; for example, G albicansÁ G flavens (aÁb) for the structurally heterozygous strain suaveolens Grado or h johansen StandardÁ h johansen Standard (haploÁhaplo) for the homozygous line johansen Standard. A cross between them (suaveolens Grado Â johansen Stanard ¼ G ablicansÁ G flavens x h johansen StandardÁ h johansen Standard) yields in the F1 the offspring G albicansÁ h johansen Standard and G flavensÁ h johansen Standard.
To equip johansen Standard (AA-I) with the chloroplast of suaveolens Grado (AB-II), the F1 hybrid G albicansÁ h johansen Standard was used. (The other hybrid G flavensÁ h johansen Standard is not of interest and therefore discarded.) Due to biparental inheritance of chloroplasts in evening primroses, it carries the chloroplasts of both johansen Standard (I-johSt) and suaveolens Grado (II-suavG). (Note that nuclear genomes are indicated in italic and chloroplast genomes in nonitalic text.) Since G albicansÁ h johansen Standard I-johSt/II-suavG displays a full meiotic ring, this leads to suppression of homologous recombination as well as elimination of random chromosome assortment in meiosis (see above albicans is not realized due to a male gametophytic lethal factor in G albicans). When a G albicansÁ h johansen Standard plant homoplasmic for II-suavG is used for selfing, the johansen Standard plant in F2 now carries plastome II-suavG (AA-II). If AB-I plants are desired, the G albicansÁ h johansen Standard I-johSt/II-suavG hybrid in F1 is selected for I-johSt and backcrossed with suaveolens Grado ( G abicansÁ G flavens II-suavG). BC1 then reassembles G albicansÁ G flavens I-johSt/II-suavG which, due to the maternal dominance of biparental transmission in evening primrose (Schötz, 1954;Chiu et al., 1988;Sobanski et al., 2019), contains a major proportion of G albicansÁ G flavens I-johSt, that is, AB-I plants.

Plant cultivation, growth conditions, and tissue harvest
For crossing studies, plastome sequencing, and analysis of RNA editing, Oenothera plants were cultivated in a glasshouse as previously described (Greiner and Köhl, 2014). AA-I, AA-II, AB-I, and AB-II plants for molecular genetics and physiological analyses were cultivated in soil in growth chambers using a 16-h light/8-h darkness cycle and 24 C at LL intensities ($150 mE m À2 s À1 ). At the beginning of the The Plant Cell, 2021 Vol. 33, No. 8 early rosette stage, $21 days after germination (cf. Greiner and Köhl, 2014), plants were transferred to higher light intensities, that is, 300 mE m À2 s À1 (LL), 450 mE m À2 s À1 (HL), or 600 mE m À2 s À1 (HHL), and then kept under the same growth regime. Hence, in all subsequent experiments the material was analyzed during the early rosette stage, which is defined as the young rosette 21-30 days after germination (see Greiner and Köhl [2014] for details and Figure 1A). About 600 mE m À2 s À1 was used only for a single experiment, because it resulted in severe photodamage of the incompatible combination AB-I (see Supplemental Text). To avoid pleiotropic effects, the yellowish material of the bleached leaf tip, a typical characteristic of the lutescent AB-I incompatible phenotype ( Figure 1A), was excluded from all experiments. The tobacco DpetN mutant and its corresponding wild-type were cultivated as reported earlier (Hager et al., 1999).

Thylakoid membrane isolation from Oenothera leaves
For spectroscopic measurements and blue native-PAGE, an improved thylakoid membrane isolation protocol was developed for Oenothera leaf tissue that contains high amounts of mucilage and starch. All steps were performed at 4 C. Solutions were pre-chilled, leaves shortly placed in ice-cold water (H 2 O), and dried with a salad spinner. Approximately 10 g of mature leaf tissue that was dark adapted for 1 h was homogenized in a blender adding 200 mL of isolation buffer (330 mM sorbitol, 50 mM HEPES, 25 mM boric acid, 10 mM EGTA, 1 mM MgCl 2 , 10 mM NaF [optional, if protein degradation is apprehend]; pH 7.6 with KOH, and 5 mM freshly added sodium ascorbate). Then 100-mL aliquots of the homogenate were then filtered through a double layer of cheese cloth (Hartmann, Heidenheim, Germany), followed by filtering through a single layer of Miracloth (Merck, Kenilworth, NJ, USA). After that, the following procedure was applied twice: after adjustment of the solution to 200 mL with isolation buffer, it was centrifuged for 5 min at 5,000 g and the pellet was resuspended in 40 mL of isolation buffer using a 30-cm 3 Potter homogenizer (VWR, Radnor, PA, USA; mill chamber tolerance: 0.15-0.25 mm). Following the second homogenization step, the solution was adjusted to 200 mL with washing buffer (50 mM HEPES/KOH [pH 7.6], 5 mM sorbitol, and optionally 10 mM NaF) followed by a filtering step through one layer of Miracloth. Subsequently, the thylakoid homogenate was centrifuged for 5 min at 5,000 g, the pellet resuspended with a 30-cm 3 Potter homogenizer in 30 mL washing buffer and centrifuged for 5 min at 5,000 g. Then, after resuspending the thylakoids in 5 mL of washing buffer, the homogenate was placed on a 85% Percoll cushion (Percoll stock solution: 3% [w/v] polyethylene glycol 6,000, 1% [w/v] BSA, 1% [w/v] Ficoll 400, dissolved in Percoll; 85% Percoll: 85% PBF-Percoll stock solution, 330 mM sorbitol, 50 mM HEPES, 2 mM EDTA, 1 mM MgCl 2 ; pH 7.6 with KOH) in a 30 mL Corex tube and centrifuged for 5 min at 5,000 g. This step effectively removes starch from the isolation. Finally, thylakoids, which do not enter the Percoll cushion, are collected, washed in a total of 25 mL of washing buffer, centrifuged for 5 min at 5,000 g, and resuspended in the desired buffer and volume.

Spectroscopic methods
For quantification of isolated thylakoids, chlorophyll amounts were determined in 80% (v/v) acetone (Porra et al., 1989). The contents of PSII, PSI, Cytb 6 f, and PC were determined in thylakoids as described previously (Schöttler et al., 2004). PSI was quantified from P700 difference absorption signals at 830-870 nm in solubilized thylakoids using the Dual-PAM-100 instrument (Walz, Effeltrich, Germany) (Schöttler et al., 2007a(Schöttler et al., , 2007b. Contents of PSII and Cytb 6 f were determined from difference absorption measurements of cytochrome b 559 and Cytb 6 f, respectively. Measurement procedures and data deconvolution methods have been described previously in detail (Kirchhoff et al., 2002;Schöttler et al., 2007a). Maximum F v /F m values were measured in leaves adapted to darkness for 1 h. Chlorophyll fluorescence was recorded with a pulse amplitude-modulated fluorimeter (Dual-PAM-100) on intact plants at room temperature. An F-6500 fluorometer (Jasco, Tokyo, Japan) was used to measure 77 K chlorophyl-a fluorescence emission spectra using an amount of freshly isolated thylakoid membranes equivalent to 10 lg chlorophyll mL À1 . The sample was excited at 430 nm (bandwidth of 10 nm), and the emission spectrum was recorded between 655 and 800 nm in 0.5 nm intervals (bandwidth of 1 nm). Dark-interval relaxation kinetics of the electrochromic shift, a measure of the proton motive force across the thylakoid membrane, was used to determine the thylakoid conductivity for protons (gH þ ), which is a proxy for ATP synthase activity. Electrochromic shift signals were measured and deconvoluted using a KLAS-100 spectrophotometer (Walz) as previously described (Rott et al., 2011).

Protein analyses
Blue native-PAGE was performed as previously reported (Ossenbühl et al., 2004;Schwenkert et al., 2006). To avoid protein degradation, 10 mM of NaF was optionally added to all solutions. Thylakoid membranes were solubilized with dodecyl-b-D-maltoside at a final concentration of 1% and separated in 4%-12% polyacrylamide gradient gels. Protein equivalents of 30 lg chlorophyll were loaded in each lane.
For immunoblot analyses, thylakoids were mixed with sample buffer (50 mM Tris/HCl [pH 6.8], 30% [v/v] glycerol, 100 mM DTT, 4% [w/v] SDS, 10% [w/v] Coomassie Brilliant Blue G-250) and were denatured for 5 min at 95 C under continuous agitation. Then, samples were analyzed by Tricine-SDS-PAGE (16%T separation gel and 4%T stacking gel) followed by gel blotting onto a PVDF membrane (0.2 mm) using the semi-dry PEQLAB transfer system (PEQLAB Biotechnologie GmbH, Erlangen, Germany). After incubation with the secondary antibody, immunochemical detection was performed with the help of the ECL Prime Western Blotting Detection Reagent (GE Healthcare, Chicago, IL, USA) according to the supplier's recommendations. In the relevant figures, 100% loading represents a 3 mg chlorophyll equivalent.

Isolation and purification of nucleic acids
DNA and RNA isolations from evening primroses were performed employing protocols specially developed for their mucilage and phenolic compound rich tissue, as previously described in Massouh et al. (2016). For RNA-seq analyses, total RNA was purified from residual DNA contamination by digestion employing the Ambion V R Turbo DNA-free TM Kit (Thermo Fisher Scientific, Waltham, MA, USA).

Association mapping of the plastid AB-I locus
For association mapping in the chloroplast genome, we used 46 full plastome sequences of Oenothera for which precise genetic information is available (Supplemental Table S1 and "Plant material" section). To this end, we newly determined the sequences of 30 plastomes, now available from GenBank under the accession numbers KT881175.1, KX014625.1, MN807266.1, MN807267.1, and MN812468.1-MN812493.1. The new chloroplast genomes were annotated and submitted by GeSeq version 1.43 (Tillich et al., 2017) and GB2sequin version 1.3 (Lehwark and Greiner, 2019), respectively. The remaining 16 plastomes were previously published (see Supplemental Table S1 for details). Chloroplast genome sequencing from Oenothera total DNA was done as reported earlier (Massouh et al., 2016;Sobanski et al., 2019), but a higher version of the SeqMan NGen assembly software was used (version 14.1.0; DNASTAR, Madison, WI, USA). Also, in contrast to earlier work, 250-bp Illumina paired-end reads (instead of 100-bp or 150-bp) were generated, with the exception of KX014625.1 (100-bp paired-end) and MN807266.1 and MN807267.1 (both 150-bp paired-end). Subsequently, for association mapping, the redundant inverted repeat A (IR A ) was removed, sequences were aligned with ClustalW (Thompson et al., 1994) and the alignments manually curated in Mesquite version 3.61 (Maddison and Maddison, 2018). Polymorphisms specific to plastome I (i.e. polymorphisms that were present in all 18 plastome I genotypes but absent from all 28 plastomes II, III, and IV genotypes) were identified by visual inspection in SeqMan Pro version 15.2.0 (DNASTAR) (cf. Greiner et al., 2008a). For the original data file, see Supplemental File S2. For readers without access to the commercial SeqMan Pro software, Supplemental File S1 and S3 are provided. Those contain the multiple sequence alignment of the 46 plastomes in standard FASTA, as well as the annotation of its consensus in the widely used GenBank format, respectively. Both files together reassemble the information included in Supplemental File S2, but can be read by any (free) sequence analyses software.

RNA editing analyses
To determine the RNA editotype of the Oenothera chloroplast, RNA-seq samples of the 1kp project (Hollister et al., 2015;Leebens-Mack et al., 2019) of johansen Standard (AA-I), suaveolens Grado (AB-II), and grandiflora Tuscaloosa (BB-III; NCBI SRA Accession Numbers ERS631151, ERS631122, and ERS631139; also see "Plant material" section) were mapped against their respective chloroplast genomes (AJ271079.4, KX014625.1, and EU262889.2) from which the IR A had been removed. For this we employed the referenceguided assembly-special workflows pipeline of SeqMan NGen version 15.2.0. single-nucleotide polymorphisms were called in SeqMan Pro version 15.2.0. To deal with the heterogeneity of the mRNA population, partial editing, and sequencing errors, sites showing C-to-T (U) conversion of at least 30% were originally considered as mRNA editing sites. If editing could not be detected above this threshold at a given site in all three species, the sites were subjected to manual inspection of the original mapping data. In most cases, this procedure revealed mapping errors, however, in a few cases also partial editing <30% in at least one of the strains was uncovered.

Gel blot detection of RNA
RNA blot analyses were performed as previously described (Massouh et al., 2016). Gene-specific PCR products used as probes were obtained by using the primers listed in Supplemental Table S5. Total Oenothera DNA was used as template in standard PCR reactions.

Chloroplast run-on analyses
For slot-blot preparation of DNA probes, PCR-amplified DNA probes (Supplemental Table S5) were immobilized to a Hybond-Nþ nylon membrane (Amersham, Little Chalfont, UK) through a slot-blot manifold. For this, 1.5 lg of DNA was denatured in 0.5 M NaOH and heated for 10 min at 95 C. Then, the volume of the denatured DNA probes was adjusted with H 2 O to 100 lL per spot. After heating, the probes were cooled on ice for 2 min to prevent DNA renaturation, and briefly centrifuged to collect the condensate. To each sample, 20 lL of cold 0.5 M NaOH and 0.5 lL of cold 10Â DNA loading-dye (50% [v/v] Plant Cell, 2021 Vol. 33, No. 8 cyanol) were added. Subsequently, the samples were spotted to nylon membranes pre-hydrated with double-distilled H 2 O, and then 100 lL 0.5 M NaOH was applied to each spot. After drying the membrane at room temperature for 5 min, the DNA was cross-linked to the membrane with 0.12 J/cm 2 using the UV crosslinker BLX-254 (BIO-LINK, Liverpool, NY, USA).
To analyze strand-specific gene expression of pbf1, singlestranded pbf1 RNA probes were generated using the Ambion V R Maxiscript V R T7 Kit (Invitrogen, Carlsbad, CA, USA) according to the manufacturer's instructions and immobilized through a slot-blot manifold to a Hybond-Nþ nylon membrane (Amersham). The pbf1 gene of johansen Standard was amplified with the primer pair psbNRO_F 5 0 -AGCATTGGGAGGCTCATTAC-3 0 and psbNRO_R 5 0 -GGAAACAGCAACCCTAGTCG-3 0 and cloned into to pCRTM2.1-TOPO V R (Invitrogen). The vector was linearized with HindIII and in vitro transcription was performed according to the suppliers' protocol. About 1.5 lg of RNA was adjusted with nuclease-free H 2 O to a volume of 50 lL prior to incubation with 30 lL of 20Â SSC (0.3 M sodium citrate and 3.0 M sodium chloride) and 20 lL 37% formaldehyde at 60 C for 30 min. Samples were maintained on ice and spotted to nylon membranes pre-hydrated with doubledistilled H 2 O and 10Â SSC. Next, 100 lL 10Â SSC was applied per slot. After drying the membrane at room temperature for 5 min the RNA was cross-linked with an UV crosslinker as described above.
For in vitro transcription and hybridization to slot-blot membranes, chloroplasts from Oenothera leaves harvested 8-10 weeks after germination were isolated and counted according to a previously published protocol, applying the same minor modifications as described in Sobanski et al. (2019). Then, a chloroplast suspension containing 4.9 Â 10 7 chloroplasts was transferred to a fresh tube, centrifuged at 5,000 g for 1 min and the supernatant was removed. To start the in vitro transcription, 20 units of RNase inhibitor (Promega GmbH, Madison, WI, USA), 50 lCi of (a-32P) UTP, and 94 lL transcription buffer (50 mM Tris/HCl [pH 8.0], 10 mM MgCl 2 , 0.2 mM CTP, GTP, and ATP, 0.01 mM UTP, 10 mM 2-mercaptoethanol) were added, mixed, and incubated for 10 min at 25 C. Next, the reaction was stopped by adding 10 lL of stop buffer (5% [w/v] Na-lauroylsarcosine, 50 mM Tris/HCl [pH 8.0], 25 mM EDTA) followed by a RNA isolation protocol, where 100 lL of phenol/chloroform/ isoamyl alcohol (25:24:1) was added to the reaction, vortexed, incubated for 10 min at room temperature, and centrifuged at 18,000 g for 10 min at 4 C. Afterward, the upper phase was collected and nucleic acids were precipitated overnight at À20 C using 3 volumes of 100% (v/v) ethanol, 0.3 M sodium acetate, and 1 lL GlycoBlue TM (Invitrogen). On the next day, the sample was centrifuged at 20,000 g for 1 h at 4 C. After centrifugation, the pellet was washed in 75% (v/v) ethanol and dissolved in 50 lL of RNase-free H 2 O. Next, the RNA was denatured at 75 C for 15 min and cooled for 2 min on ice. Before hybridizing the slot blots with the isolated RNA, the membrane was pre-hybridized with 20 mL of Church buffer (1 mM EDTA, 7% [w/v] SDS, 0.5 M NaHPO 4 [pH 7.2]) in hybridization tubes at 65 C for 1 h. Hybridization was performed at 65 C overnight. Subsequently, the membrane was washed once with 1 Â SSC and 0.2% (w/v) SDS for 10 min, and once with 0.5 Â SSC and 0.2% (w/v) SDS for 10 min. After washing, the membrane was wrapped in a transparent foil and exposed to a storage phosphor screen for 5 days. The signals were detected using an Amersham Typhoon IP scanner.

5'-RACE experiments to map transcription start sites
TAP transcript 5'-end mapping in Oenothera chloroplasts was performed as previously described (Kühn et al., 2005). In brief, primary transcripts of bacteria and cell organelles have triphosphates at their 5 0 -ends, while processed transcripts possess monophosphates at the 5 0 -end. The tobacco acid pyrophosphatase (TAP) enzyme removes the additional phosphates from the 5'-end of primary transcripts. After this treatment, both primary and processed transcripts can serve as substrate for RNA ligase. This allows us to distinguish between primary and processed transcripts when þTAP and ÀTAP treated samples are compared. In ÀTAP samples, ligation products originating form primary transcripts are absent. Hence, to map the transcription start sites of the psbB operon and to distinguish them from processing sites in close proximity, a 5'-RACE from þTAP (Epicenter, Madison, WI, USA) and ÀTAP RNA samples was performed. For this, RNA samples of both treatments were ligated to an RNA linker (5'-GUGAUCCAACCGACGCGACAAGCUAAUGCAAG ANNN-3'). After cDNA synthesis with a psbB gene-specific primer (psbB_cDNA_jn 5 0 -GCTGGCTGTCCATATAATG CATACAGC-3 0 ), two PCRs were performed: the first PCR used the linker-specific primer RUMSH1 (5 0 -TGATCCAACC GACGCGAC-3 0 ) and the psbB-specific primer psbB_cDNA_jn. The second PCR used the linker-specific nested primer RUMSH2 (5 0 -ACCGACGCGACAAGCTAA TGC-3 0 ) and the primer psbB_5prime_jn (5 0 -GGAAAGGG ATTTTAGGCATACCAATCG-3 0 ). PCR products (30 mL of PCR solution) were run on 1% agarose gels (w/v) and, prior to sequencing, cloned into pCR2.1-TOPO (Invitrogen).

Ribosome footprint isolation
After nuclease digestion, the monosome pellet was suspended in 500 mL ribosome footprint isolation buffer (10 mM Tris [pH 8.0], 1 mM EDTA, [pH 8.0], 100 mM NaCl, 1% [w/v] SDS, 0.1 M EGTA [pH 8.0]) and transferred to a 2 mL centrifuge tube. To the monosome suspension 750 lL TRIzol TM was added, and the sample was vortexed and incubated for 10 min at room temperature. Subsequently, 150 lL of chloroform/isoamylalcohol was added; the sample was vortexed and centrifuged for 20 min at 20,000 g at room temperature. The supernatant was transferred into a 2 mL centrifuge tube containing 1 mL of isopropanol, and the RNA precipitated for 30 min at À20 C. Afterwards, the RNA was centrifuged for 40 min at 20,000 g at 4 C and the pellet was washed with 75% (v/v) ethanol. The pellet with the ribosome footprints was suspended in 20 lL of H 2 O. The RNA concentrations were measured using a NanoDrop 1000 spectrophotometer, and the RNA was stored at À80 C. After the incubation for 5 min at 45 C, the sample was denatured for 10 min at 75 C. For each lane, 30-40 mg of RNA dissolved in 40 mL loading buffer was loaded and 4 lL of Dynamarker Prestain Marker for small RNA Plus served as marker for the gel. The gel was electrophoresed in 1Â TBE buffer at a constant power of 30 W per gel until the dye front reached the end of the gel. During the run, the gel apparatus was connected to a cold machine and cooled to 8 C (Pharmacia & Upjohn, Kalamazoo, MI, USA), which avoids gel melting and, especially for small nucleic acids, leads to better electrophoresis results by decreasing mobility rates. Then a region representing RNAs between 20 and 50 nt was excised and the RNA was eluted from the gel in 4 mL TESS solution per lane (10 mM Tris [pH 8.0], 1 mM EDTA [pH 8.0], 0.1 M NaCl, 0.2% [w/v] SDS) overnight in a rotator at 4 C. The solution was transferred to a new tube containing 4 mL phenol/chloroform/isoamyl alcohol (25:24:1) and 2.5 lL GlycoBlue TM , vortexed and centrifuged for 5 min at 1,700 g at room temperature. Afterwards the supernatant was transferred into a new tube and RNA was precipitated at À20 C overnight by the addition of 2.5 volumes of 100% (v/v) ethanol. The next day, the sample was centrifuged at 15,000 g at 4 C for 1 h and the pellet was eluted in 500 lL H 2 O. Subsequently, the eluted ribosome footprints were transferred to a new tube containing 100 mM NaCl and 500 lL phenol/chloroform/isoamyl alcohol, vortexed and centrifuged at 15,300 g at room temperature for 20 min. Then the supernatant was transferred to a new tube containing chloroform/isoamyl alcohol, vortexed, and centrifuged at 15,300 g at room temperature for 20 min. To precipitate the RNA, the supernatant was transferred into new tubes with 2.5 volumes of 100% (v/v) ethanol and incubated overnight at -20 C. The next day, the sample was centrifuged at 20,000 g at 4 C for 1 h. Afterward the pellet was washed with 75% (v/v) ethanol, centrifuged at 20,000 g at 4 C for 10 min, and eluted in 20 lL of H 2 O. The size distributions of the ribosome footprints were measured using the Bioanalyzer (Agilent Technologies, Santa Clara, CA, USA) using the Agilent small RNA Kit, and the concentrations were determined with the Qubit 4 Fluorometer (Thermo Fisher Scientific) using the Qubit TM RNA Assay Kit. The ribosome footprints were stored at -80 C.

Validation
A detailed validation of the ribosomal profiling protocol established for evening primrose is reported in Kozul (2019) and will be published in a subsequent communication.

NGS library preparation and sequencing
To prepare the ribosome footprints for the next-NGS library preparation, 5'-termini phosphorylation of the ribosome footprints was performed using the T4 PNK kit (polynucleotide kinase, Thermo Fisher Scientific), following the manufacturer's instructions. Total RNA for RNA-seq libraries was isolated as described above. Subsequently, Ribo-seq and RNA-seq library preparations were performed according to the manufacturer's protocols both for the LL and HL treatments in three replicates using the NEXTflex Small RNA-Seq Kit version 3 (PerkinElmer, Waltham, MA, USA) and Zymo Research (Freiburg, Germany) Zymo-Seq RiboFree Total RNA Library Kit, respectively. To provide details for the Ribo-seq library preparation, to 10.5 mL phosphorylated ribosome footprints 100 ng 3'-4N adenylated adapter and 5'-4N adenylated adapter were ligated, followed by the reverse transcription of the 3'-and 5'-adapter ligated RNA into a first strand synthesis product. The PCR products were subsequently amplified (18 cycles; 10 s at 98 C, 30 s at 60 C, and 15 s at 72 C) by using a universal primer and different barcoded primer for each sample. Afterward, a gel-free size selection and clean-up were performed. The size distribution of all final libraries was determined with the Bioanalyzer using the Agilent High Sensitivity DNA Kit, and the concentration was measured with the Qubit 4 Fluorometer using the Qubit TM dsDNA HS Assay (Thermo Fisher Scientific). Multiplexing and sequencing of the NGS libraries were performed at the Max Planck Institute for Molecular Genetics (Berlin, Germany) using an Illumina V R Nova Seq 6000 machine (SP flow cell for Ribo-seq and S1 flow cell for RNA-seq libraries, 100 bp single-end reads).

NGS data analysis
Quality control of Ribo-seq (accessions GSM5288039 to GSM5288050) and RNA-seq (accessions GSM5288051 to GSM5288068) sequencing data, both included into the GEO Series GSE174154, was performed using FastQC version 0.11.8 (https://www.bioinformatics.babraham.ac.uk/projects/ fastqc/). The 3 0 -adapter trimming was applied to both data sets using Flexbar version 2.5 (Dodt et al., 2012) with default parameters and the corresponding adapter-barcode-combined sequences. To remove a potential PCR bias from the Ribo-seq data set, the four-base 3'-and 5'-unique molecular identifiers (UMIs) were removed from the read sequences and attached to the read identifier in order to make them available for further usage, while sustaining correct alignment of the reads. All Ribo-seq data, as well as RNA-seq data for computing translation efficiency, were then aligned against plastome I (I-johSt ¼ AJ271079.4) using STAR version 2.7.0a with slightly adapted indexing parameters (-genomeSAindexNbases 8 and -genomeChrBinNbits 17) (Dobin et al., 2013). Subsequently, samtools version 1.10 was used to remove secondary alignments, to sort and to index them, and to generate mapping, and gene-wise count statistics (flagstat and idxstat, respectively). Conversion of transcriptomic bam files (harboring only reads mapped to exons) to browser extensible data (bed) files was done using bamToBed from the bedtools version 2.29.2 (Quinlan and Hall, 2010), adding read length as additional column. Both file formats, bam and bed, contain coordinate-based information for reads mapped against a set of sequences. BAM is a compressed binary version of the Sequence Alignment/ Map format and bed files contain less but the most important information for read alignments compared to the BAM format. See https://genome-euro.ucsc.edu/FAQ/FAQformat. html#format1 and https://github.com/samtools/hts-specs/ blob/master/SAMv1.pdf, respectively, for format specifications and details. Usable reads for all follow-up analysis were defined by the following criteria: (1) removing tRNA and rRNA-associated reads as well as reads shorter than 20 bases and longer than 50 bases; (2) read de-duplication based on chromosomal location and same UMIs (by sorting bed files on the columns scaffold_id, start/end position and both columns having the UMIs using groupBy to take only the first entry of each group identical entries (-o first); (3) remain only reads with mapping quality equal to 255 (means unique mapped in the context of the STAR mapper).
Resulting bed files were loaded into R to generate different types of plots.
Counts per million reads normalization was done on the total number of reads that mapped to protein-coding genes of the chloroplast. Gene length normalization to obtain reads per kb of exon per million reads values was applied by summing up the exon lengths for each gene in kilobase. Translational output and log 2 ratio bar plots are based on gene-wise read counting in which only reads greater or equal to 25 and 35 were considered. Translational efficiency was computed by dividing the RF ratio by the RNA ratio where the ratios were calculated as follows: RF ratio (Ribo-seq sample1/sample2) and RNA ratio (RNA_seq sam-ple1/sample2).

Statistical analyses
All numerical results are reported as mean 6 SD. Statistical significance of the difference between experimental groups was analyzed by unpaired t tests using GraphPad Prism software ( Figure 2). Differences were considered statistically significant for P < 0.05 or P < 0.01. Two-way ANOVA with Tukey post-hoc testing was performed using SigmaPlot version 14.0 (Systat Software, San Jose, CA, USA; Table 1). Reports are provided in Supplemental File S4. RNA gel blots, immunoblots, and run-on analyses were repeated at least twice. Representative data are shown.
Supplemental Figure S1. Photosynthetic parameters of compatible AB-II and incompatible AB-I plants grown under LL or HL conditions Supplemental Figure S2. RNA gel blot and RNA-seq analyses of the clpP and psbB operons in compatible (AA-I, AA-II, and AB-II) and incompatible material (AB-I).
Supplemental Table S1. Accession numbers, genome sizes, genetic information, and corresponding nuclear genotypes of Oenothera plastomes used for association mapping Supplemental Table S2. Chloroplast mRNA editotype and cDNA mapping results of three Oenothera species from subsection Oenothera Supplemental Table S3. Origin and collector information for the Oenothera strains used in this work Supplemental Table S4. Chloroplast substitution lines, F1 hybrids, and corresponding wild-types used in this work Supplemental Table S5. Oligonucleotides used for the generation of probes for RNA gel blot and run-on transcription analyses Supplemental File S1. Multiple sequence alignment of the 46 Oenothera plastomes used for association mapping provided in the FASTA format (.txt file).
Supplemental File S2. Multiple sequence alignment and annotation of the 46 Oenothera plastomes used for association mapping provided as SeqMan Pro project file (DNASTAR) (.sqd file).
Supplemental File S3. Annotation of the alignment consensus of the 46 Oenothera plastomes used for association mapping provided in the GenBank format (.txt file).
Supplemental File S4. Two-way ANOVA reports generated using SigmaPlot14 software containing statistical analyses for Table 1, Figure 2B, and Supplemental Figure S2, A and C (.xls file).