Chloroplast Translation: Structural and Functional Organization, Operational Control, and Regulation

Abstract

Chloroplast translation is essential for cellular viability and plant development. Its positioning at the intersection of organellar RNA and protein metabolism makes it a unique point for the regulation of gene expression in response to internal and external cues. Recently obtained high-resolution structures of plastid ribosomes, the development of approaches allowing genome-wide analyses of chloroplast translation (i.e., ribosome profiling), and the discovery of RNA binding proteins involved in the control of translational activity have greatly increased our understanding of the chloroplast translation process and its regulation. In this review, we provide an overview of the current knowledge of the chloroplast translation machinery, its structure, organization, and function. In addition, we summarize the techniques that are currently available to study chloroplast translation and describe how translational activity is controlled and which cis-elements and trans-factors are involved. Finally, we discuss how translational control contributes to the regulation of chloroplast gene expression in response to developmental, environmental, and physiological cues. We also illustrate the commonalities and the differences between the chloroplast and bacterial translation machineries and the mechanisms of protein biosynthesis in these two prokaryotic systems.

INTRODUCTION

Chloroplasts are the characteristic organelle of plant cells. They host numerous essential metabolic pathways including photosynthesis, which makes chloroplasts the primary source of chemical energy on earth. All chloroplasts are likely derived from a single ancient photosynthetic cyanobacterium that was engulfed by a mitochondriate eukaryotic cell more than a billion years ago. During subsequent host-endosymbiont coevolution, the genome of the endosymbiont shrank significantly (Timmis et al., 2004). While some genes were lost, many others were transferred to the host genome (Martin et al., 2002; Bock and Timmis, 2008). The proteome of today’s chloroplasts consists of ∼3000 proteins, most of which are nucleus-encoded and posttranslationally imported into the organelle.

Present-day chloroplasts still harbor a genome, which comprises ∼120 genes in green plants. Most plastid genes are essential for plant viability because they encode crucial components of the photosynthesis machinery (the large subunit of Rubisco and approximately half of the subunits of the thylakoidal protein complexes involved in the light reactions: photosystem I and II [PSI and PSII], cytochrome b₆f complex [Cyt b₆f], and ATP synthase) and the gene expression system of the plastid (including a complete set of bacterial-type RNA polymerase core subunits, rRNAs and tRNAs, and approximately one-third of the ribosomal proteins; Allen et al., 2011; Green, 2011). The reasons for retention of this particular set of genes in the plastid genome are not fully understood. Several hypotheses that are not necessarily mutually exclusive have been put forward, including constraints on importability of proteins (and RNAs) into plastids and requirements for efficient and organelle-specific redox regulation of gene expression (Allen, 2015). An intriguing consequence of endosymbiont-host coevolution is multimeric chloroplast protein complexes, whose subunits are encoded in different compartments (i.e., the plastid and the nucleus). This necessitates the tight orchestration of nuclear and chloroplast gene expression (Jarvis and López-Juez, 2013; Kleine and Leister, 2016).

The bacterial origin of chloroplast gene expression is evident, for example, from the operon-like structure of plastid gene clusters and the highly similar composition of the translation machinery. However, several features clearly distinguish chloroplast gene expression from that of bacteria. For example, chloroplasts possess an astoundingly complex RNA metabolism that includes the usage of different RNA polymerases and extensive posttranscriptional RNA processing by splicing, editing, end processing, and intercistronic processing of polycistronic RNAs (Barkan, 2011; Lyska et al., 2013; Börner et al., 2015). These processes are nearly exclusively conducted by nucleus-encoded protein factors, most of which were likely established during host-endosymbiont coevolution (Barkan, 2011; Lyska et al., 2013; Pfalz and Pfannschmidt, 2013). Also different from bacteria, the regulatory influence of transcription is limited and posttranscriptional and translational events represent key points in controlling chloroplast gene expression (Barkan, 2011; Sun and Zerges, 2015).

In recent years, numerous specific features of chloroplast translation were uncovered, including its interconnection with cotranslational processes in RNA and protein metabolism, its regulation in response to internal and external triggers, and the presence of unusual components of the translation machinery. Genome-wide analyses have unraveled the suborganellar localization of translation and its participation in controlling the developmental program of chloroplast gene expression (Zoschke and Barkan, 2015; Chotewutmontri and Barkan, 2016). Studies of sequence-specific chloroplast RNA binding proteins that comprise helical repeat domains, especially the pentatricopeptide repeat (PPR) proteins, revealed their concerted function in RNA metabolism and promotion of translation (Barkan and Small, 2014; Hammani et al., 2014). Last but not least, high-resolution structural analyses provided a detailed three-dimensional picture of the plastid ribosome (Graf et al., 2016; Bieri et al., 2017). These discoveries were largely enabled by novel approaches toward the quantitative transcriptome-wide analysis of translational activity, the identification of specific factors that control protein synthesis, and the structural elucidation of the translational apparatus in plastids (e.g., by ribosome profiling techniques, RNA coimmunoprecipitation assays, and refined methods for 3D structural analysis). In the light of these and other findings, we now can reevaluate classical models of chloroplast translation and reassess controversially discussed hypotheses.

Chloroplast translation has been mainly studied in the unicellular green alga Chlamydomonas reinhardtii and in model seed plants such as Arabidopsis thaliana, maize (Zea mays), and tobacco (Nicotiana tabacum). In this review, we focus on embryophytes and, wherever appropriate, refer to breakthrough discoveries made in Chlamydomonas. For a broader overview of chloroplast translation in Chlamydomonas, the interested reader is referred to comprehensive review articles (Stern et al., 2009; Nickelsen et al., 2014; Sun and Zerges, 2015).

METHODS TO ANALYZE CHLOROPLAST TRANSLATION

Methods to determine translational activity either indirectly examine the ribosome coverage of the translation template (i.e., the mRNA) or directly measure the accumulation of newly synthesized proteins.

Classical Methods to Analyze Translation

Pulse labeling is the method of choice to directly measure translational activity in vivo. In this approach, isolated chloroplasts, cells, or intact plant tissues are fed with the ³⁵S-radiolabeled amino acids methionine and/or cysteine (Barkan, 1998). The isotopes are incorporated into newly synthesized proteins to an extent that mirrors their synthesis rate. Subsequently, the radiolabeled proteins can be separated by gel electrophoresis, visualized, and quantified (Figure 1A). An advantage of pulse labeling is that it has the potential to measure protein synthesis rates independent of the dynamics of ribosome movement along the mRNA. By pulse labeling, the synthesis rate of especially the large plastid-encoded core subunits of the photosynthesis machinery can be readily quantified. However, the method has several limitations: (1) Small subunits and subunits of similar molecular weight are difficult to resolve in protein gels and may require selective purification by immunoprecipitation (Barkan, 1998). (2) The synthesis rates of many plastid-encoded proteins cannot be determined by pulse labeling approaches due to their low expression levels. (3) The measured quantities of labeled proteins are determined by their synthesis and degradation rates. Consequently, for proteins with high turnover rates (e.g., PsbA, the D1 protein of PSII), results of pulse-labeling experiments are often difficult to interpret (even if followed by a chase with unlabeled amino acids to examine the stability of the labeled protein). Also, (4) in multicellular organisms, neither the pulsing nor the chasing occurs homogenously in all cells, thus making quantitative comparisons very challenging.

Figure 1.

Common Methods to Analyze Chloroplast Translation.

(A) Pulse labeling. Plant cells (chloroplast, large green oval; nucleus, white circle) are fed with radiolabeled cysteine and/or methionine (red dots), which is incorporated together with unlabeled amino acids (black dots) into nascent peptides by translation (for simplicity, only chloroplast ribosomes are shown). Proteins are then isolated, separated by gel electrophoresis, and visualized/quantified by radio-detection methods. (Gel picture kindly provided by Karin Meierhoff.)

(B) Polysome analysis. Plant cell lysates are loaded on sucrose gradients (white to black: low to high concentration) to separate RNPs according to their molecular weight by ultracentrifugation. RNA is isolated from gradient fractions and examined by RNA gel blot analysis to determine the ribosome loading of specific mRNAs.

(C) Ribosome profiling. Plant cell lysates are treated with nuclease to degrade ribosome-free mRNA sequences. This generates monosomes, whose protected mRNA fragments (ribosome footprints) are subsequently purified. The positions and abundances of the ribosome footprints are determined by next-generation sequencing or microarray hybridization and reflect protein synthesis rates.

Polysome analysis is a widely used method that indirectly measures translational activity using the association of mRNAs with ribosomes as a proxy. Polysomes are high molecular weight assemblies of actively translating ribosomes held together by the strands of mRNA being translated. They can be separated from free mRNAs and ribosomes (monosomes) by ultracentrifugation in sucrose density gradients (Figure 1B). Following RNA extraction from gradient fractions and RNA gel blot analysis, the distribution of specific mRNAs across the gradient is visualized and provides a qualitative measure of their translational activity (Barkan, 1998). For genome-wide analysis (translatomics), mRNAs recovered from different density fractions can be examined by microarray hybridization (Kahlau and Bock, 2008). Polysome analyses are often complicated by the operon-like organization of genes in the plastid genome. The processing of transcripts produced from polycistronic transcription units frequently gives rise to a multitude of mono-, oligo-, and polycistronic RNA species, all of which represent potential translation templates (Barkan, 1988). Due to the physical linkage of reading frames located on the same transcript, their individual translation rates cannot be resolved. In addition, only translational regulation at the level of initiation can be detected because the molecular weight of mRNAs loaded with actively elongating ribosomes is indistinguishable from those with paused or stalled ribosomes.

Other elegant though labor intensive methods have been used to examine the regulatory capacity of cis-elements in chloroplast translation: (1) Chloroplast in vitro translation systems have been established and used to analyze the regulatory influence of putative cis-elements residing in the 5′ untranslated region (5′UTR) on translation (Hirose and Sugiura, 1996; Yukawa et al., 2007). (2) Reporter genes (e.g., GFP or GUS) have been fused to different presumed cis-elements and inserted into the plastid genome by chloroplast transformation to examine the translational activity conferred by these sequences (Staub and Maliga, 1994; Eibl et al., 1999; Drechsel and Bock, 2011).

Ribosome Profiling: Genome-Wide Analysis of Translation at High Resolution

The above described classical methods have been informative, but they are labor intensive and limited in resolution, and none of them is suited to genome-wide and/or high-throughput analyses. These deficits were addressed by ribosome profiling, an approach that enables the quantitative genome-wide analysis of translation in unprecedented depth and resolution (Ingolia et al., 2009). Ribosome profiling takes advantage of the remarkable stability of translating ribosomes, which protect the mRNA sequence they physically cover from attack by nucleases, thereby producing protected fragments, so-called ribosome footprints (Wolin and Walter, 1988; Figure 1C). Next-generation sequencing analysis of these footprints determines the in vivo positions and abundances of translating ribosomes. Considering that each elongating ribosome produces one protein, ribosome footprint abundances reflect the protein synthesis rate for each reading frame (Ingolia et al., 2009). Footprint abundance is typically normalized to mRNA abundance (assayed by RNA sequencing), so that relative translation efficiencies can be inferred (Ingolia et al., 2009). Consequently, the approach measures the two determinants of gene expression that define the final protein output: transcript amount and translational activity. In recent years, ribosome profiling has been extensively used to study translation in prokaryotes and eukaryotes (Ingolia, 2016).

In chloroplasts, ribosome profiling was first applied in a modified approach, exchanging the next-generation sequencing analysis of footprints by microarray hybridization (Zoschke et al., 2013a; Figure 1C). More recently, deep sequencing was used to study chloroplast translational dynamics in maize and Arabidopsis (Chotewutmontri and Barkan, 2016; Lukoszek et al., 2016; Gawroński et al., 2018).

Despite the compelling attractions of ribosome profiling, it should be noted that the method cannot distinguish actively translating from paused ribosomes. This may be problematic if translation is regulated at the level of elongation, as described for some chloroplast genes (see below). Application of inhibitors of initiation or early elongation (e.g., lincomycin) and examination of the run-off kinetics of ribosomes over time should allow distinguishing pausing from elongating ribosomes.

THE PLASTID TRANSLATION MACHINERY: VARIATIONS ON A BACTERIAL THEME

A Bacterial-Like Translation Machinery Whose Components Are Encoded by Two Genomes

Chloroplast translation is performed by prokaryotic-type 70S ribosomes that are composed of a small 30S and a large 50S subunit and contain orthologs of most proteins and all rRNAs of the Escherichia coli reference ribosome (Kössel et al., 1985; Yamaguchi and Subramanian, 2000; Yamaguchi et al., 2000). All rRNAs, a complete set of ∼30 tRNAs, approximately half of the ribosomal proteins of the 30S subunit, and one-quarter of the proteins of the 50S subunit are encoded in the plastid genome (Sugiura, 1995). The remaining ribosomal proteins are nucleus-encoded (Tiller and Bock, 2014; http://www.i2bc.paris-saclay.fr/spip.php?article1261&lang=fr). Interestingly, the genomic distribution of genes for organellar ribosomal proteins is, to some degree, evolutionarily conserved, suggesting constraints in ribosome assembly that require on-site coexpression of rRNAs and organelle-encoded core ribosomal proteins (Maier et al., 2013). Most other components of the plastid translation machinery are nucleus-encoded (e.g., initiation/elongation/termination/ribosome recycling factors and aminoacyl-tRNA synthetases), except for initiation factor 1 (IF1), which is plastid-encoded in many plants (Millen et al., 2001).

Plastid Deviations from the Bacterial Ribosome and Their Structural and Functional Consequences

Despite the generally bacterial structure of the chloroplast ribosome, there are some features that clearly distinguish plastid ribosomes from the E. coli reference ribosome. Chloroplast ribosomes contain the full set of bacterial rRNAs (23S, 16S, and 5S rRNAs), which comprise the peptidyl transferase activity (23S rRNA) and the decoding center (16S rRNA) and serve as scaffold for ribosomal proteins during ribosome assembly (Shajani et al., 2011; Maier et al., 2013). However, the 23S rRNA gene was split into two genes in the plastid genome: a large 5′ portion encoded by the 23S rRNA gene and a small 3′ fragment encoded by the 4.5S rRNA gene (Whitfeld et al., 1978). In addition, the 23S rRNA is posttranscriptionally processed at two so-called “hidden breaks” into three fragments, whose abundances and precise sizes vary among species (e.g., ∼0.5, ∼1.2, and ∼1.1 kb from 5′ to 3′ in Arabidopsis plastids). These fragments are found in the mature 70S ribosome and held together by intermolecular base pairing (Kössel et al., 1985; Bieri et al., 2017). Despite their general homology, some structural elements of the E. coli 16S and 23S rRNAs are absent from chloroplasts and, conversely, the chloroplast 23S rRNA contains additional secondary structures (Kössel et al., 1985). Particularly well conserved are the catalytic domain V in the 23S rRNA, which carries the peptidyl transferase activity, and the anti-Shine-Dalgarno sequence in the 16S rRNA, which is crucial for translation initiation (Scharff et al., 2017).

In the course of evolution, significant changes also occurred in the proteinaceous part of the chloroplast ribosome. The Rpl25 and Rpl30 proteins were completely lost, and, in some species, the bacterial Rpl23 was replaced by its counterpart from the cytosolic 80S ribosome (Bubunenko et al., 1994; Yamaguchi and Subramanian, 2000). Furthermore, the first complete inventory of plastid ribosomal proteins identified six proteins that were assumed to lack bacterial orthologs and, consequently, were designated as plastid-specific ribosomal proteins (PSRP) 1-6 (Yamaguchi and Subramanian, 2000; Yamaguchi et al., 2000). However, later it was shown that PSRP1 is not a genuine ribosomal protein, but the ortholog of the bacterial cold-shock protein pY, which is associated with the small subunit of the ribosome but not a structural part of it (Sharma et al., 2007, 2010). PSRP4 also shows homology to a bacterial protein: THX, an intrinsic part of the 30S ribosomal subunit in Thermus thermophilus, which, so far, has not been found in other bacteria (Yamaguchi and Subramanian, 2003). At present, PSRP2 and PSRP3 in the 30S ribosomal subunit, and PSRP5 and PSRP6 in the 50S subunit are considered genuine plastid-specific ribosomal proteins; consequently, their renaming to RPS22/23 and RPL37/38 was suggested (Bieri et al., 2017). After their discovery, PSRPs were hypothesized to act in light regulation of translation (Yamaguchi and Subramanian, 2003; Manuell et al., 2007). Whereas plastid pY may indeed perform this function (see below), a number of studies have suggested that the major function of PSRPs lies in the structural compensation of evolutionarily modified rRNA domains (Sharma et al., 2007; Tiller et al., 2012; Ahmed et al., 2016; Graf et al., 2016; Bieri et al., 2017). However, this role does not necessarily exclude additional functions in translational regulation.

Some of the conserved plastid ribosomal proteins also exhibit N- or C-terminal extensions (or internal expansions) compared with their E. coli orthologs (Yamaguchi and Subramanian, 2000; Yamaguchi et al., 2000). Many of these extensions mediate new interactions with rRNAs or ribosomal proteins and may structurally compensate for missing or modified rRNA domains (Ahmed et al., 2016; Graf et al., 2016; Bieri et al., 2017). However, the extensions of some ribosomal proteins (e.g., S2, S18, and S21) represent potential new contact sites with the mRNA and therefore were hypothesized to be involved in translational regulation (Manuell et al., 2007; Sharma et al., 2007; Graf et al., 2016). Notably, several alterations (e.g., extensions of RPS5 and RPS1 and an insertion in RPS4) narrow the mRNA entry site of the chloroplast ribosome compared with the E. coli reference ribosome (Bieri et al., 2017). Furthermore, structural changes in the polypeptide exit tunnel and the tunnel exit site were hypothesized to support the cotranslational binding of the chloroplast signal recognition particle (SRP), which diverges substantially from that of bacteria (Ahmed et al., 2016; Graf et al., 2016; Bieri et al., 2017).

Altogether, the chloroplast ribosome has a substantially higher protein mass (by ∼170 kD) and a slightly lower RNA content (by ∼0.4 kD) than the E. coli ribosome, resulting in a considerably increased protein to RNA ratio (∼2:3 compared with 1:3 in E. coli). The partial replacement of rRNA domains by protein elements in chloroplast ribosomes follows the general evolutionary trend of reducing RNA components in enzymatically active chloroplast ribonucleoprotein particles (RNPs) and substituting them by protein constituents (Barbrook et al., 2006). Other examples include the chloroplast tRNA processing enzyme RNase P that lost its RNA component in the Viridiplantae lineage (Pinker et al., 2013) and the chloroplast SRP of seed plants that lacks the SRP RNA (Ziehe et al., 2017). The tendency to lose RNA functions in chloroplasts may be driven by the evolutionary genome reduction and the massive transfer of protein-coding genes to the nucleus. Whereas proteins can be reimported posttranslationally into plastids, this route seems to be blocked for RNA components, which may have enforced their evolutionary loss, replacement by proteins, or retention in the chloroplast genome (in the case of rRNAs and tRNAs; Barbrook et al., 2006). Also, it was predicted that ribosome composition (including its relative protein and rRNA contents) is optimized for the production of the translation machinery itself, a process that strongly limits cell division rates in prokaryotes (Reuveni et al., 2017). However, in plastids, the synthesis of ribosomal proteins and rRNAs is partially uncoupled from translation due to the transfer of genes for ribosomal proteins to the nucleus and the presence of nucleus-encoded RNA polymerases, both of which are produced by cytosolic rather than organellar ribosomes. Consequently, the evolutionary constraints on ribosome composition are somewhat relaxed in organelles, which may have facilitated the observed shifts in protein-to-rRNA ratios in plastid and mitochondrial ribosomes (Reuveni et al., 2017).

Some Ribosomal Proteins Are Dispensable under Standard Growth Conditions

Most constituents of the plastid translation machinery are essential for chloroplast biogenesis and, consequently, for plant viability (Tiller and Bock, 2014). In many species, plastid translation is essential even under heterotrophic growth conditions (Ahlert et al., 2003; Sosso et al., 2012), presumably due to the necessity to express a few essential plastid genes such as accD, clpP, ycf1, and ycf2 (Bock, 2007). However, some ribosomal proteins and tRNAs are nonessential, at least under standard greenhouse conditions.

Nonessential ribosomal proteins have been identified in both E. coli and plastids. Surprisingly, despite the shared ancestry of bacterial and plastid ribosomes, the essentiality of ribosomal proteins is not fully conserved between the two systems (Tiller and Bock, 2014). The nucleus-encoded chloroplast ribosomal proteins RPL11, RPL24, RPS17, RPS21, PSRP3, and PSRP6 and the plastid-encoded Rps15, Rpl33, and Rpl36 are nonessential (Tiller and Bock, 2014). The phenotypes of viable ribosomal protein mutants range from wild-type appearance to very strong phenotypes with altered leaf morphology, variegated leaves, cold-induced bleaching, and retarded growth. Nonessential plastid ribosomal proteins have been speculated to facilitate ribosome assembly and structural integrity, or act in regulation, optimization, or localization of translation (Pesaresi et al., 2001; Yamaguchi and Subramanian, 2003; Tiller et al., 2012; Tiller and Bock, 2014). However, their distinct molecular functions remain to be elucidated. In bacteria, ribosomes can differ in their protein composition, for instance, under different growth conditions. Moreover, there is growing evidence that this ribosome heterogeneity (additionally involving posttranscriptional and posttranslational modifications of rRNAs and ribosomal proteins) creates regulatory capacity by conferring translational selectivity of ribosomal subpools for specific mRNAs (Sauert et al., 2015; Shi et al., 2017). However, whether nonessential plastid ribosomal proteins can act as modulators of ribosome affinity to specific sets of mRNAs is currently unknown.

Translation with a Minimal Set of tRNAs

Similar to some bacterial species, plastids do not encode the full set of 32 tRNAs that are required to serve the 61 codons by standard and wobble base pairing between codon and anticodon. For example, only 30 tRNA genes are present in the plastid genome of most seed plants. Since there is no evidence for tRNA import from the cytosol, superwobble base pairing has been considered as a mechanism enabling translation with a reduced set of tRNAs. A systematic reverse genetic screen revealed that the chloroplast tRNAs trnG-GCC, trnL-CAA, trnS-GGA, trnT-GGU, and trnV-GAC are not essential for plant viability (Alkatib et al., 2012, and references therein). These nonessential tRNAs decode codons with a pyrimidine in the third position and, for all of them, essential plastid isoacceptor tRNAs with a uracil (U) in the first anticodon position exist in plastids. These isoacceptor tRNAs can serve the respective codons by superwobble base pairing (i.e., the U in the wobble position of the anticodon can pair with all four nucleotides in third codon position; Alkatib et al., 2012). The diverse phenotypes of these mutants suggest that superwobble base pairing causes distinct molecular constraints on translation (Alkatib et al., 2012), possibly explaining the evolutionary conservation of some nonessential tRNAs.

CONSERVATION AND MODIFICATION OF BACTERIAL TRANSLATION MECHANISMS IN CHLOROPLASTS

The overall structural conservation of essential functional elements of bacterial ribosomes in chloroplasts is generally assumed to reflect the functional preservation of the bacterial translation mechanisms in plastids (Peled-Zehavi and Danon, 2007).

Initiation

The initiation process starts with the contact of a preinitiation complex consisting of the 30S subunit and the initiator tRNA (for N-formylmethionine [fMet]) to the initiation site in the mRNA. Bacterial initiation depends on the initiation factors IF1, 2, and 3, which facilitate initiator tRNA binding and ribosome subunit assembly. Functional chloroplast orthologs were identified for all three IFs (Sijben-Müller et al., 1986; Campos et al., 2001; Miura et al., 2007; Zheng et al., 2016). Interestingly, many plants contain two or more paralogous genes for plastid IF3, whose differential expression was proposed to regulate chloroplast translation initiation (Nesbit et al., 2015). Also similar to the standard bacterial translation initiation, approximately two-thirds of the chloroplast reading frames are preceded by the purine-rich Shine-Dalgarno sequence (SD) (Shine and Dalgarno, 1974; Scharff et al., 2011). In bacteria, the SD interacts by base pairing with a pyrimidine-rich sequence in the 16S rRNA (the anti-Shine-Dalgarno sequence [aSD]) to ensure proper positioning of the initiation complex at the start codon. The aSD is fully conserved in plant plastids, and evidence for its functionality has been provided for many genes with SD (Kim and Mullet, 1994; Hirose and Sugiura, 2004). Nevertheless, the significance of SD-dependent initiation in chloroplasts has been questioned (Fargo et al., 1998). To provide ultimate clarification, mutations were introduced into the aSD in the tobacco chloroplast genome and shown to cause reduced translation for many reading frames with upstream SDs, thus confirming the functionality of SD-aSD interactions (Scharff et al., 2017). On the other hand, roughly a third of the chloroplast genes and also many bacterial genes do not contain SD sequences (or contain a putative SD but display SD-independent translation initiation). It was shown that low amounts of mRNA secondary structure around the start codon facilitate SD-independent translation initiation (Scharff et al., 2011, 2017; Nakagawa et al., 2017). Furthermore, in bacteria, the ribosomal protein S1 preferentially binds polypyrimidine tracts and acts as an RNA chaperone that unfolds structured regions in mRNAs, thereby enabling efficient SD-independent translation initiation (Qu et al., 2012). Chloroplast S1 binds RNA with a preference for adenine or uracil-rich sequences, but whether this supports SD-independent translation initiation is unknown (Franzetti et al., 1992; Shteiman-Kotler and Schuster, 2000).

Studies in both bacteria and chloroplasts have pointed to a 5′-to-3′ ribosome scanning mechanism and the preferential utilization of the 5′-most start codon (Drechsel and Bock, 2011; Yamamoto et al., 2016). Moreover, extended interactions of mRNA sequences upstream of the start codon and adjacent to the SD with bases downstream of the anticodon of the initiator tRNA-fMet and bases next to the aSD, respectively, have been suggested to facilitate chloroplast translation initiation (Ruf and Kössel, 1988; Esposito et al., 2003; Kuroda et al., 2007). However, the exact mechanism and the quantitative contribution of these interactions to the efficiency of translation initiation remain to be elucidated. Similar to bacteria, the triplets AUG, GUG, and UUG can be utilized as start codons in chloroplasts (Hirose et al., 1999; Kuroda et al., 2007; Rott et al., 2011; Moreno et al., 2017), with the recognition efficiency of non-AUG start codons presumably depending on the sequence context (Boeck and Kolakofsky, 1994).

Elongation

After binding of the 50S subunit to the preinitiation complex, the functional 70S ribosome is completed and starts moving along the coding sequence of the mRNA to translate it into a polypeptide chain. Bacterial translation elongation depends on the factors EF-Tu, EF-G, and EF-Ts for which conserved chloroplast orthologs were identified (Breitenberger et al., 1979; Fox et al., 1980; Sreedharan et al., 1985). Notably, the expression of chloroplast elongation factors is regulated by light and other stimuli (temperature, phytohormones, and developmental cues), suggesting involvement of elongation factors in the regulation of translation (Akkaya and Breitenberger, 1992; Bhadula et al., 2001; Singh et al., 2004; Albrecht et al., 2006; Liu et al., 2010; Schröter et al., 2010).

Analogous to bacterial gene expression, many chloroplast genes are cotranscribed from operon-like gene clusters. The resulting polycistronic transcripts contain reading frames that are separated by (often short) spacer sequences or even overlap by a few nucleotides. In bacteria, translation of adjacent and overlapping reading frames is frequently coupled in that translation of the second coding sequence depends on that of the first one (Jackson et al., 2007). In some cases, the strong RNA helicase activity of the ribosome translating the upstream reading frame is needed to unfold RNA secondary structures that mask initiation elements (SD and/or start codon) of the downstream reading frame (Jackson et al., 2007). Recently, direct coupling of termination on the upstream reading frame (without ribosome disassembly) with subsequent scanning and reinitiation on the downstream reading frame also has been demonstrated (Yamamoto et al., 2016). In chloroplasts, cases of coupled (ndhC/K and psbD/C) and uncoupled (atpB/E) translation were identified by in vitro and in vivo analyses, respectively, but the detailed mechanisms of translational coupling are unknown (Yukawa and Sugiura, 2008; Adachi et al., 2012; Zoschke et al., 2013a). A recently designed in vivo expression system exploits coupled translation in chloroplasts and suggests that the RNA helicase function of the ribosome may also mediate translational coupling in chloroplasts (Martin Avila et al., 2016).

Termination and Ribosome Recycling

When one of the three stop codons is reached, the orthologous release factors RF1/PrfA (serving UAA and UAG) and RF2/PrfB1 (serving UAA and UGA) set the synthesized protein free by hydrolysis of the ester bond (Buckingham et al., 1997; Meurer et al., 2002; Motohashi et al., 2007). Additionally, PrfB1 is involved in the stabilization of plastid mRNAs containing reading frames with UGA stop codons (Meurer et al., 2002). PrfB3, a nonfunctional chloroplast-targeted paralog of PrfB1, lacks domains that are essential for stop codon recognition and hydrolytic activity of release factors. Remarkably, PrfB3 gained a new function in transcript stabilization of the petB mRNA (Stoppel et al., 2011). Another release factor, RF3, facilitates dissociation of RF1 and RF2 from the ribosome in bacteria (Buckingham et al., 1997) and likely also in chloroplasts (Beligni et al., 2004). In the final step, ribosome recycling factor (Rolland et al., 1999), EF-G, and IF3 facilitate the release of mRNA and tRNA, and the disassembly of the small and large ribosomal subunits, thereby recycling them for the next round of translation initiation(Kiel et al., 2007

PLASTID TRANSLATION IS INTERCONNECTED WITH RNA AND PROTEIN METABOLISM

Relaxed Coupling of Translation and Transcription

In bacteria, translation initiates and elongates cotranscriptionally, thus ensuring efficient transcription (e.g., by preventing RNA polymerase backtracking), conferring RNA stability (by translating ribosomes protecting the mRNA from ribonucleolytic attack), enabling timely translation, and maintaining genome integrity (e.g., by avoiding extended hybridization of RNA and DNA that would cause collision of the transcription and replication machineries; McGary and Nudler, 2013). A similar coupling of transcription and translation was proposed for chloroplast gene expression based on early electron micrographs that were interpreted as evidence for ribosomes being associated with nascent transcripts (Rose and Lindbeck, 1982; Figure 2). Additional evidence for coupling of transcription and translation has come from the findings that (1) ribosomal proteins are associated with the transcription machinery (Pfalz et al., 2006), and (2) translation factors and other proteins involved in translation are enriched in plastid nucleoids in a ribonuclease-sensitive manner, suggesting tethering by nascent transcripts (Majeran et al., 2012). Moreover, orthologs of the Nus proteins that couple transcription and translation in bacteria have been identified in chloroplast nucleoids (Majeran et al., 2012). However, chloroplast transcripts have a longer half-life than bacterial transcripts and are stable when not covered by ribosomes, and many translated RNA species are generated by RNA processing (see below). Together, this implies that, simply due to the kinetics of mRNA processing and turnover, there may be a quantitative shift toward posttranscriptional translation in chloroplasts.

Figure 2.

Overview of Internal and External Triggers That Cause Regulatory Adjustments of Translation in Chloroplasts, the Mechanisms That Control Translation, the Coupling of RNA and Protein Metabolism to Chloroplast Translation, and the Localization of the Chloroplast Translation Machinery.

Chloroplast translation is regulated in response to internal and external triggers (listed in the upper part). Nucleus-encoded factors are translated in the cytosol (shown in the upper left part) and imported into the chloroplast, where they control and/or regulate chloroplast protein synthesis directly (by altering chloroplast translation activity) or indirectly (by controlling cotranslational chloroplast RNA or protein metabolisms). Chloroplast translation occurs cotranscriptionally (left); however, due to the slow mRNA turnover, the majority of ribosomes act posttranscriptionally (right). RNA binding proteins assist cotranslational RNA processing and/or facilitate translation initiation. Ribosomes initiate and elongate regularly on both processed and unprocessed transcripts, the extent of which seems to mainly depend on the kinetics of the processing events (see text and Figure 3). Many of the factors involved in protein processing, folding, targeting, and assembly act cotranslationally on the nascent polypeptide. See text for details.

The chloroplast genome is regularly transcribed by bacterial-type and phage-type RNA polymerases that have very different properties (Börner et al., 2015). For example, the two polymerases transcribe with different speeds and recognize different promoters, thus producing primary transcripts with divergent 5′ ends. Notably, in phage-infected E. coli cells, the speedy transcription by the RNA polymerase of bacteriophage T7 is not coupled with translation and thus produces initially “naked” (i.e., ribosome-free) transcripts with a higher decay rate (Makarova et al., 1995). How transcription by the different plastid RNA polymerases is coordinated with translation and whether or not the utilized RNA polymerase influences the kinetics of protein synthesis are currently unknown.

Unprocessed Transcripts Can Be Translated

Primary chloroplast transcripts undergo extensive RNA processing, including splicing of group I and II introns, RNA editing (changing cytosine to uracil residues to restore codons for conserved amino acids or start or stop codons), 5′ and 3′ end trimming, and intercistronic processing that generates diverse transcript isoforms from polycistronic primary transcripts (Barkan, 2011; Lyska et al., 2013). There is no obvious spatial separation that would compartmentalize RNA metabolism and translation. Interestingly, many of the factors known to be involved in transcript processing (e.g., RNA binding proteins), stabilization, and translation colocalize with processed and unprocessed transcripts in nucleoids and transcription complexes (Pfalz et al., 2006; Majeran et al., 2012; Lehniger et al., 2017). This raises the question whether unprocessed mRNAs, often referred to as “precursors” or “immature transcripts,” are utilized as templates for translation or whether any partitioning (e.g., temporal separation) exists between RNA processing and translation in plastids.

Up to 12 reading frames are interrupted by group II introns in chloroplast genomes of seed plants (Sugiura, 1995). In all of them, a substantial fraction of the coding region is located downstream of the intron; consequently, splicing is essential to produce functional proteins (Barkan, 2011). Surprisingly, recent ribosome profiling studies in maize chloroplasts have demonstrated that translation initiates on unspliced atpF, ndhA, ndhB, and ycf3 transcripts and also elongates (Zoschke et al., 2013a; Alice Barkan, personal communication; Figure 3A). Whether translation elongation pauses at the robust intron structure or, alternatively, the RNA helicase function of the ribosome allows translation to proceed into the intron until it terminates at the first in-frame stop codon, thus producing nonfunctional proteins, has not yet been possible to resolve. Nevertheless, it is clear that ribosomes initiate and elongate on unspliced transcripts, strongly arguing against a spatial or temporal separation of splicing and translation processes.

Figure 3.

Ribosomes Translate Unprocessed Chloroplast Transcripts (See Text for Details).

(A) Several chloroplast reading frames are interrupted by group II introns. Left: Translating ribosomes cover exon 1 of unspliced atpF, ndhA, ndhB, and ycf3 transcripts (Zoschke et al., 2013a; Alice Barkan, personal communication). Middle and right: Splicing releases the intron and ligates the exons. Consequently, both exons of the spliced transcript are occupied by ribosomes, producing full-length proteins (chain of black dots: nascent polypeptide).

(B) Chloroplast transcripts are edited at specific sites by modification of cytosine (C) to uracil (U) residues, often restoring codons for conserved amino acids (change from yellow to white dot in the nascent peptide). Actively translated mRNAs have the same editing status as the total transcriptome (Chotewutmontri and Barkan, 2016), indicating that, in a partially edited transcript pool, unedited transcripts also are translated.

(C) Polycistronic chloroplast transcripts often undergo posttranscriptional processing that generates smaller transcript isoforms (represented by the three monocistronic transcripts on the right; RF, reading frame). Often all transcript isoforms are used as translation templates. The extent to which transcript processing may enhance translation efficiency needs to be determined on a case-by-case basis.

High-resolution ribosome profiling analysis of maize chloroplast translation also demonstrated that the degree of editing in mRNA footprints of actively translating ribosomes is not substantially different from that in the general transcript pool (Chotewutmontri and Barkan, 2016). Specific editing sites that are only partially edited in the transcript pool showed a similar degree of partial editing in ribosome footprints. This is in line with the earlier finding that editing of tobacco rps14 is not a requirement for efficient in vitro translation (Hirose et al., 1998). Together, these data imply that ribosomes cannot distinguish between edited and unedited transcripts (Figure 3B). However, two exceptions were identified in maize: the transcripts of rpl2 and ndhA, which are only translated in their edited form (Chotewutmontri and Barkan, 2016). In rpl2, ACG-to-AUG editing restores the start codon (Hoch et al., 1991), thereby activating translation (Chotewutmontri and Barkan, 2016). This is consistent with the earlier finding that restoration of the tobacco ndhD start codon by editing is essential for efficient translation initiation in vitro (Hirose and Sugiura, 1997). In some instances, start codon restoration by editing is subject to tissue-specific or developmental variation, thus raising the intriguing possibility that editing may control translational activity (Ichinose and Sugita, 2016). In the case of ndhA, splicing of its group II intron is required to enable editing of the first editing site in exon II (Schmitz-Linneweber et al., 2001, and references therein). Consequently, ndhA transcripts that are unedited at this particular site are unspliced, and elongating ribosomes cannot get through to the editing site in the second exon (Chotewutmontri and Barkan, 2016). Leaving aside these special cases, it can be assumed that, normally, unedited and unspliced transcripts are translated (Figure 3) and potentially give rise to the synthesis of low amounts of nonfunctional and potentially deleterious proteins. In accordance with this assumption, failure to restore conserved amino acids in subunits of PSII, the Cyt b₆f complex, and the ATP synthase by mRNA editing strongly impairs the function of the respective complexes (Bock et al., 1994; Zito et al., 1997; Schmitz-Linneweber et al., 2005b). Hence, rapid proteolytic removal of the dysfunctional proteins synthesized from unedited and unspliced transcripts has to be assumed to prevent deleterious effects on chloroplast function.

End trimming and intercistronic processing of plastid transcripts have been proposed to enhance their translational activity (Drechsel and Bock, 2011). However, internal reading frames on polycistronic transcripts derived from the maize psbB-psbT-psbH-petB-petD transcription unit are actively translated, despite the fact that they exist also as 5′ reading frames on processed transcripts (Barkan, 1988). Likewise, unprocessed tobacco atpH and rbcL mRNAs were translated as efficiently as processed mRNAs in vitro (Yukawa et al., 2007). Furthermore, synthetic transcription units that were engineered into plastids gave rise to polycistronic transcripts that were efficiently translated in the absence of processing (Staub and Maliga, 1995). Finally, a genome-wide ribosome profiling study revealed that several polycistronic mRNAs are efficiently used as translation template, despite the known coexistence of monocistronic transcript isoforms (Zoschke and Barkan, 2015). In all of these examples, the translation of downstream reading frames in polycistronic mRNAs was not dependent on transcript processing into monocistronic units, indicating that internal start codons can be efficiently recognized. Also, a number of plastid reading frames are only present in di- or polycistronic transcripts, indicating that these must undergo translation. Altogether, these data provide compelling evidence that transcript end processing and intercistronic processing are not general requirements for efficient translation (Figure 3C). However, in a few cases, there is good evidence that transcript processing stimulates translation. For instance, in tobacco, a base-pairing interaction between the psaC coding region and the ndhD 5′UTR in the dicistronic transcript was shown to prevent efficient ndhD translation in vitro, whereas processed monocistronic transcripts were translationally active, suggesting that, in this operon, processing is required to activate translation (Hirose and Sugiura, 1997). Similarly, in vitro translation provided evidence that unprocessed atpB, psbB, and psbD transcripts from tobacco chloroplasts are less efficiently translated than their processed isoforms (Yukawa et al., 2007; Adachi et al., 2012). Moreover, in tobacco chloroplasts, heterologous expression of GFP from engineered polycistronic mRNAs was more efficient when GFP was placed at the 5′ end of the synthetic operon (Drechsel and Bock, 2011). Finally, in maize, the monocistronic forms of psaI and rps14 show the highest accumulation in those developmental stages where the highest translation rates of these reading frames occur, a finding that would be consistent with the monocistronic RNA species being better translatable (Chotewutmontri and Barkan, 2016).

In sum, although in some cases, translation is indeed stimulated by RNA processing, there is no general dependence of translation on processing. Unprocessed, unspliced, and unedited transcripts have been shown to be used as translation templates (Figure 3); therefore, these transcript isoforms are not necessarily “immature” or “precursors.”

Several RNA Binding Proteins Act Dually in Transcript Processing/Stabilization and Promotion of Translation

In recent years, plastid RNA binding proteins, many of them with helical repeat domains, were shown to be involved in specific RNA end trimming and intercistronic processing events (Barkan and Small, 2014; Hammani et al., 2014). Mutants of some of these factors (Table 1) displayed defects in RNA processing that were accompanied by translation deficiencies (Barkan et al., 1994; Felder et al., 2001; Hashimoto et al., 2003). Initially, these were interpreted as RNA processing-dependent translation defects (in that processing was required for efficient translation). However, later, it was observed that the knockout of PPR10, an RNA binding protein involved in the processing and stabilization of specific atpH transcript isoforms, caused much stronger translation and protein accumulation defects than expected from its RNA processing defect, suggesting a more direct role of PPR10 in translation of atpH (Pfalz et al., 2009). In vitro assays showed that PPR10 binds to the atpH 5′UTR and protects transcripts from 5′-to-3′ exonucleolytic degradation. Consequently, the PPR10 binding site defines the 5′ end of these transcripts (Prikryl et al., 2011). In addition, PPR10 binding remodels the atpH 5′UTR such that an RNA stem-loop structure that occludes the putative SD sequence of atpH is dissolved and the ribosome binding site becomes exposed (Prikryl et al., 2011). This suggests a dual function of PPR10 in RNA stabilization and stimulation of atpH translation. A similar mode of action was shown, or is discussed, for other chloroplast RNA binding proteins such as HCF107, PGR3, CRR2, and CRP1 (see Table 1 and references therein). The additional translation-promoting function should be independent of processing in that it should also occur in polycistronic transcripts where the target reading frame is located downstream of other reading frames. In fact, reanalysis of ppr10, crp1, pgr3, and hcf107 maize mutants by ribosome profiling revealed substantial translation defects in vivo (and less severe transcript accumulation defects) for the reading frames downstream of the RNA binding sites of the respective protein (in atpH, petD, petL, and psbH expression, respectively; Zoschke et al., 2013a; Alice Barkan, personal communication). In line with a dual function of some RNA binding proteins, PPR53, a member of the small PPR-SMR family, was recently described to be involved in both promotion of ndhA translation and processing/stabilization of transcript isoforms with ndhA as the 5′ reading frame (Zoschke et al., 2016).

Table 1. Factors Demonstrated or Suggested to Facilitate Translation of Specific Transcripts in Seed Plant Plastids.

Factor	Protein Domain(s)	Reading Frames with Translation Promoted	Species	References for Translational Function
ATP1	Unknown	atpB	Maize	McCormac and Barkan (1999); Zoschke et al. (2013a)
ATP4/SVR7	PPR, SMR	atpB	Maize, Arabidopsis	Zoschke et al. (2012)
CRP1	PPR	petA, petD*, psaC	Maize	Barkan et al. (1994); Zoschke et al. (2013a)
CRR2	PPR	ndhB*	Arabidopsis	Hashimoto et al. (2003)
HCF107	HAT	psbH*	Arabidopsis	Felder et al. (2001); Hammani et al. (2012)
HCF152	PPR	petB*?	Arabidopsis	Meierhoff et al. (2003)
HCF173	Atypical SDR	psbA*	Arabidopsis	Schult et al. (2007)
HCF244	Atypical SDR	psbA	Arabidopsis	Link et al. (2012)
PGR3	PPR	petL*, ndhA?	Arabidopsis	Yamazaki et al. (2004); Cai et al. (2011)
PPR10	PPR	atpH*	Maize	Pfalz et al. (2009); Prikryl et al. (2011); Zoschke et al. (2013a)
PPR53	PPR, SMR	ndhA*	Maize, Arabidopsis	Zoschke et al. (2016)

Asterisks indicate an additional function in stabilization of the transcript 5′ end upstream of the reading frame whose translation is stimulated. Question marks denote proposed but experimentally unconfirmed functions in stimulation of translation. SDR, short-chain dehydrogenase/reductase; HAT, half a tetratricopeptide repeat; SMR, small MutS-related.

These examples support the idea that RNA binding proteins acting in processing/stabilization of specific 5′ transcript ends can also directly promote translation of the reading frame downstream of their binding site (Figure 2). Such proteins provide a physical link between RNA metabolism and translation; therefore, co-occurrence of intercistronic processing and stimulation of translation does not necessarily imply a strict requirement of RNA processing to facilitate translation.

Notably, a widely employed sequence element in chloroplast biotechnology (IEE, for intercistronic expression element) that enhances the heterologous expression of reading frames located in polycistronic transcription units includes the HCF107 binding site (Zhou et al., 2007; Hammani et al., 2012). The insertion of the IEE between reading frames enhances the accumulation of monocistronic transcripts. This could be related to the RNA secondary structure formed by the IEE (Zhou et al., 2007), the action of RNase E (Walter et al., 2010), and/or the binding of HCF107 (Hammani et al., 2012; Legen et al., 2018).

The Turnover of Plastid Transcripts Is Not Determined by Their Translation Status

In E. coli, mRNAs are stabilized by translating ribosomes, presumably by ribosome coverage providing physical protection from ribonucleases (Laalami et al., 2014). By contrast, the study of maize and Arabidopsis mutants with transcript-specific translation defects has revealed that atpB, petA, psaC, and psbA transcripts are stable although their translation and, consequently, their ribosome occupancy was dramatically reduced (Barkan et al., 1994; McCormac and Barkan, 1999; Link et al., 2012; Zoschke et al., 2012, 2013b). Furthermore, exchange of the canonical AUG start codon by the nonstandard initiation codon GUG or UUG in the atpB, clpP, and psbD reading frames in tobacco chloroplasts diminished translation initiation and protein synthesis but did not destabilize the transcripts (Rott et al., 2011; Moreno et al., 2017; Mark A. Schöttler, personal communication). Also, mutants with general impairments in chloroplast translation do not exhibit general transcript accumulation defects (Barkan, 1993; Scharff et al., 2017). Similarly, the treatment of wild-type Arabidopsis plants with lincomycin, an antibiotic that disturbs 70S translation elongation only at the earliest steps, thus causing runoff of ribosomes (Kallia-Raftopoulos and Kalpaxis, 1999), did not cause a substantial decrease in the accumulation of any of the analyzed chloroplast transcripts (Meurer et al., 2002; Stoppel et al., 2011). Accumulation of chloroplast transcripts was also not increased after treatment with chloramphenicol, a 70S elongation inhibitor that arrests ribosomes and inhibits their release, thus resulting in densely ribosome-covered transcripts (Nierhaus and Wittmann, 1980).

Altogether, the available data demonstrate that chloroplast mRNAs are stable in the absence of translation and do not require physical protection by ribosomes. This may not be surprising given the evolutionary switch from a largely transcriptional regulation of gene expression, as found in bacteria, to predominantly posttranscriptional regulation in chloroplasts (which strongly depends on transcripts with long half-lives). Chloroplasts harbor many endo- and exoribonucleases that potentially could degrade “naked” transcripts (Germain et al., 2013). This implies that nontranslated chloroplast mRNAs must be somehow protected against nuclease attack. A small family of RNA recognition motif domain-containing proteins, the chloroplast RNPs (cpRNPs), were shown to be involved in different steps of mRNA metabolism, including transcript stabilization (Ruwe et al., 2011). Taking into account the high abundance of these proteins in the chloroplast, their broad RNA binding activity in vivo (in that they associate with virtually all mRNAs), and the fact that they are specifically bound to nonpolysomal mRNAs (Nakamura et al., 2001; Kupsch et al., 2012; Teubner et al., 2017), cpRNPs are strong candidates for providing stability to untranslated mRNAs. Notably, mutants of the cpRNPs CP29A and CP31A show a conditional cold-sensitive phenotype, presumably caused by a reduction in the stability of many mRNAs (Kupsch et al., 2012). A possible explanation is a runoff of translating ribosomes in the cold and subsequent transcript degradation in the absence of stabilizing cpRNPs. A recent structural analysis revealed a narrowed mRNA entry site of the chloroplast ribosome compared with that of E. coli (Bieri et al., 2017). It seems tempting to speculate that this is because chloroplast ribosomes, different from their bacterial counterparts, need to strip off abundant RNA binding proteins such as cpRNPs when translation reinitiates and elongates on previously “stored” (i.e., untranslated) mRNAs or even on normally translated mRNAs with low initiation rates (resulting in larger ribosome spacing).

In sum, untranslated chloroplast mRNAs are stable, and cpRNP binding may protect them against ribonucleolytic attack, thus causing the observed uncoupling of mRNA stability from translation. However, a direct functional connection between translational activity, cpRNP (un)binding, and mRNA stability remains to be demonstrated.

Cotranslational Folding, Maturation, Targeting, and Assembly of Proteins

In bacteria, several steps in protein metabolism, including proteolytic processing, chemical modification, cofactor binding, folding, targeting, and assembly, can occur cotranslationally (Gloge et al., 2014). It seems clear that this is also the case in chloroplasts, although the knowledge about the intersections between plastid translation and protein metabolism is still scarce (Giglione et al., 2015; Breiman et al., 2016; Figure 2). Removal of the N-terminal N-formylmethionine often represents the first step of nascent peptide chain processing in bacteria and chloroplasts. It occurs cotranslationally by the consecutive reactions of peptide deformylase and methionine aminopeptidase (Breiman et al., 2016). Likewise, the N-terminal signal peptide for thylakoid targeting of PetA (cytochrome f) is cotranslationally cleaved (see below). Another widespread N-terminal modification of the nascent peptide is the N-α-acetylation of the penultimate amino acid (Zybailov et al., 2008; Breiman et al., 2016). A complete list of identified N-terminal processing events in plastid-encoded proteins is provided at http://www.i2bc.paris-saclay.fr/spip.php?article1261andlang=fr (Breiman et al., 2016).

In bacteria and eukaryotes, folding of the nascent peptide chain has been shown to start already in the ribosome exit tunnel (Bhushan et al., 2010; Gloge et al., 2014). With dimensions of 10 nm in length and 1 to 2 nm in width, the 70S ribosome exit tunnel has a sufficient size to shelter 30 to 60 amino acids (depending on the folding status) and allows the formation of small protein domains consisting of α-helices (Holtkamp et al., 2015). Protein folding in the exit tunnel is assisted by the ribosome itself through interactions of the nascent peptide with the 23S rRNA and ribosomal proteins (e.g., L4, L22, and L23; Gloge et al., 2014). Given the high conservation of the peptide exit tunnel, cotranslational folding is expected to occur also during chloroplast translation. The ribosome-associated chaperone trigger factor binds the nascent peptide upon exit from the ribosome and stabilizes it, thus preventing protein aggregation and assisting cotranslational protein folding in bacteria and, most likely, also in chloroplasts (Breiman et al., 2016; Ries et al., 2017). Subsequently, other chaperones are recruited and take over (Trösch et al., 2015).

In parallel to folding, cofactors such as chlorophylls, hemes, carotenoids, quinones, and metal ions can associate cotranslationally with chloroplast apoproteins (Schöttler et al., 2011, 2015; Nickelsen and Rengstl, 2013). Several studies suggest that the plastid-encoded apoproteins of PSI and PSII must bind chlorophylls cotranslationally to ensure faithful complex biogenesis, most likely, because chlorophyll binding is required for correct protein folding and assembly (Nickelsen and Rengstl, 2013). This is supported by evidence that chlorophyll stabilizes nascent chlorophyll binding proteins (Mullet et al., 1990; Kim et al., 1994b; Eichacker et al., 1996). Specific pausing sites during psbA, psaA, psaB, and psaC translation elongation were suggested to facilitate the cotranslational binding of chlorophyll and other cofactors such as pheophytin, quinone, iron sulfur, and manganese clusters (Kim et al., 1991, 1994a; Gawroński et al., 2018). In cyanobacteria, unassembled PsbB and PsbC apoproteins contain chlorophyll a and β-carotene, suggesting their early cotranslational association (Boehm et al., 2011).

Chloroplasts comprise different suborganellar compartments: stroma, thylakoid membrane, thylakoid lumen, inner and outer envelope membranes, and the intermembrane space. The targeting of some chloroplast-encoded proteins to the thylakoid membrane has long been recognized to occur cotranslationally (reviewed in Celedon and Cline, 2013; Figure 2). Early on, it was shown that puromycin treatment (which causes premature translation termination and thereby release of the nascent peptide) also releases ribosomes from the thylakoid membrane in chloroplasts, suggesting cotranslational protein targeting mechanisms (Yamamoto et al., 1981). In later studies, chloroplast subfractionation coupled with polysome analysis and pulse labeling studies revealed that polytopic proteins of PSI (PsaA and PsaB) and PSII (PsbA/D1, PsbB/CP47, PsbC/CP43, and PsbD/D2), and the bitopic cytochrome f subunit (PetA) of the Cyt b₆f complex associate with the thylakoid membrane cotranslationally (Margulies et al., 1987; Friemann and Hachtel, 1988; Kim et al., 1994b; van Wijk et al., 1996). However, the interpretation of the results from these experiments was sometimes controversial (Ibhaya and Jagendorf, 1984) and complicated by the fact that in chloroplasts, polycistronic transcripts can be used as translation templates. Consequently, one cotranslationally inserted polypeptide produced from a polycistronic transcript is sufficient to tether all cotranscribed cistrons to the thylakoid membrane. This difficulty was overcome in a recent study using ribosome profiling, a method that employs nucleases to degrade mRNAs in polysomes down to the footprints protected by monosomes (Zoschke and Barkan, 2015). By coupling this approach with fractionation of chloroplasts into thylakoid membranes and stroma, cotranslational membrane insertion could be comprehensively examined at a genome-wide scale. The study revealed that 19 of the 37 plastid-encoded intrinsic transmembrane domain-containing thylakoid proteins in maize insert cotranslationally into the membrane and supplied evidence that exposure of the first transmembrane domain provides the signal and/or the anchor for stable membrane association (with the sole exception of PetA, as described below). The data suggest a model for ribosome-mediated mRNA targeting, in which the nascent polypeptide exposed by the first “pioneer” ribosome anchors the translation machinery together with the translated mRNA at the thylakoid membrane. Continued translation by the following ribosomes keeps the mRNA tethered to the thylakoid membrane. A similar model was suggested for the cytosolic ribosomes that are associated with mitochondria and the endoplasmic reticulum (Jan et al., 2014; Williams et al., 2014). In addition, a recent ribosome profiling study in Arabidopsis correlated plastid ribosome pausing events with the synthesis and correct integration of transmembrane domains (Gawroński et al., 2018). Electron microscopic evidence indicates that membrane-associated chloroplast polysomes are connected with all unstacked thylakoid membrane regions (i.e., stroma lamellae and grana margins), but not with internal membranes in grana stacks, which are inaccessible due to their tight packing (Yamamoto et al., 1981). Consequently, once grana stacks are assembled during chloroplast biogenesis, plastid-encoded grana proteins (mainly PSII subunits) need to be transported posttranslationally from unstacked membrane regions into grana stacks, for example, during photosystem repair (Puthiyaveetil et al., 2014; Pribil et al., 2014).

The mechanisms involved in suborganellar protein targeting have been best studied for nucleus-encoded chloroplast proteins that are posttranslationally distributed (Celedon and Cline, 2013). Five major pathways with partially overlapping functions (and some shared components) have been described (Schünemann, 2007; Celedon and Cline, 2013). The secretory (Sec) pathway and the twin-arginine translocase (Tat) transport proteins across the thylakoid membrane into the lumen (reviewed in Schünemann, 2007). The chloroplast SRP (cpSRP) interacts with the cpSRP receptor cpFtsY and the insertase ALB3 to insert nucleus-encoded light-harvesting complex proteins into the thylakoid membrane (reviewed in Ziehe et al., 2017). Some proteins apparently insert spontaneously into the thylakoid membrane; finally, a recently discovered parallel Sec pathway targets proteins to the inner envelope membrane (Li et al., 2017b).

Much less is known about the mechanisms involved in cotranslational suborganellar targeting of chloroplast-encoded proteins. Most plastid-encoded proteins are found in either the stroma or the thylakoid membrane. Plastid-encoded membrane proteins likely utilize one of the above-mentioned targeting pathways, either co- or posttranslationally. So far, the cotranslational targeting mechanisms were elucidated in some detail for only two plastid-encoded proteins: PetA and PsbA. PetA is the only plastid-encoded protein containing a cleavable signal peptide at its N terminus, which is recognized by cpSecA. In vitro and genetic data suggest that PetA targeting to the thylakoid membrane occurs cotranslationally (Voelker et al., 1997; Röhl and van Wijk, 2001, and references therein). A recent ribosome profiling study showed that nascent PetA engages the thylakoid membrane long before its single transmembrane domain becomes exposed outside the ribosome exit tunnel, thus confirming cotranslational action of cpSecA-dependent targeting in vivo (Zoschke and Barkan, 2015). Only recently, SecA-mediated targeting in bacteria was demonstrated to also occur cotranslationally (Huber et al., 2016).

Cotranslational interaction of PsbA and cpSRP54 has been suggested based on in vitro cross-linking experiments (Nilsson and van Wijk, 2002, and references therein). However, Arabidopsis mutants lacking cpSRP54 have very mild phenotypes, arguing against an essential role of cpSRP54 in membrane targeting of PsbA or any other core subunit of the photosynthesis machinery (Amin et al., 1999; Tzvetkova-Chevolleau et al., 2007). Nonetheless, cpFtsY, the essential chloroplast SRP receptor homolog that had previously been implicated in PSII repair (Tzvetkova-Chevolleau et al., 2007; Asakura et al., 2008), was shown to be associated with nascent PsbA in vitro suggesting a role in cotranslational targeting (Walter et al., 2015). Furthermore, the translocons cpSecY and ALB3 and the multifunctional protein Vipp1, all of which are essential for thylakoid biogenesis (Sundberg et al., 1997; Roy and Barkan, 1998; Kroll et al., 2001), appear to interact with nascent PsbA in vitro, suggesting their involvement in cotranslational membrane targeting of PsbA (Zhang et al., 2001; Walter et al., 2015). Recent in vitro data suggest that the cotranslational targeting of PetB (cytochrome b₆) also involves the insertase ALB3 (Króliczewski et al., 2016).

The assembly of proteins into functional complexes often initiates cotranslationally (Natan et al., 2017). Biogenesis of the multiprotein complexes in the thylakoid membrane requires the tightly coordinated action of multiple assembly factors that guide the association of plastid-encoded and nucleus-encoded subunits (Schöttler et al., 2011; Nickelsen and Rengstl, 2013). Many plastid-encoded subunits of these complexes are believed to be assembled cotranslationally (Figure 2) because unassembled free subunits are usually condemned to rapid degradation. However, only in few cases, direct evidence for cotranslational assembly has been obtained. The PSII assembly process is best understood (Nickelsen and Rengstl, 2013), and several plastid-encoded core subunits of PSII have been suggested to assemble cotranslationally into the complex. During both de novo assembly and repair of PSII, nascent PsbA subunits are integrated into an early assembly intermediate that contains PsbD (Müller and Eichacker, 1999; Zhang et al., 1999). Whereas the first and second transmembrane domains of PsbA only weakly interact with PsbD, a robust association is established after synthesis of the fourth transmembrane domain (Zhang et al., 1999; Zhang and Aro, 2002). It is tempting to speculate that the operon-like organization of chloroplast genes enhances the efficiency of cotranslational targeting and assembly (e.g., in the dicistronic psaA/B and psbD/C transcripts), as this was suggested for bacteria (Natan et al., 2017).

In summary, a multitude of factors act cotranslationally as “welcoming committee” for nascent polypeptides and assist with the amazing metamorphosis of linear amino acid chains into functional proteins and protein complexes in specific suborganellar locations (Figure 2). We are just beginning to understand the complex interconnections of the diverse processes involved in cotranslational protein maturation, targeting, and assembly, but it is becoming increasingly evident that the ribosome acts as a central hub in the coordination of these processes.

OPERATIONAL CONTROL AND REGULATION OF CHLOROPLAST TRANSLATION

Upon stress and under changing environmental conditions, the thylakoid membrane system is adjusted to achieve optimum photosynthetic performance and prevent photooxidative damage. These acclimation responses require integration of multiple external and internal signals (Figure 2) and involve extensive regulation of chloroplast translation (Nickelsen et al., 2014; Sun and Zerges, 2015).

Translational Control Versus Regulation

The terms “translational control” and “translational regulation” are sometimes used synonymously. However, factors controlling translation (e.g., the strength of a ribosome binding site in the 5′UTR) do not necessarily also regulate translation (i.e., dynamically change protein synthesis rates in response to environmental stimuli or developmental programs).

The interplay of trans-acting protein factors and cis-acting RNA elements determines the translation output of chloroplast genes. The functional involvement of these elements in translational control typically is demonstrated by their genetic manipulation causing altered translational activity. However, this does not necessarily imply a regulatory function in translation, resulting in a change in the protein synthesis rate during adaptation processes. In other words, a given factor would have a regulatory function if it were to become limiting for translation under specific conditions, thus altering protein synthesis output. According to this definition, a true regulatory function of chloroplast translation factors has been established only in very few cases in Chlamydomonas (see below). It was proposed that many of the nucleus-encoded factors involved in chloroplast RNA metabolism simply suppress mutations that accumulate in the (asexually reproducing) plastid genome over time (Maier et al., 2008; Lefebvre-Legendre et al., 2014). Similarly, nucleus-encoded translation factors could be needed constitutively to fix chloroplast mutations at the RNA level (e.g., by resolving secondary structures around ribosome binding sites to facilitate translation initiation). However, on an evolutionary time scale, factors that control translation may also be recruited as true regulators that modulate translation in response to internal and external triggers.

Translation Plays a Major Role in the Control and Regulation of Chloroplast Gene Expression

Translation is an extremely resource-consuming process due to the energy and nutrient demands involved in the assembly of ribosomes, the synthesis of amino acids, the expression, processing, and charging of tRNAs, and the GTP-dependent reactions during initiation and elongation. Dividing bacterial cells use ∼50% of their energy for protein synthesis (Russell and Cook, 1995). Millar and coworkers calculated the cellular energy budgets used for protein synthesis in Arabidopsis leaves (Li et al., 2017a). Their estimate is that, dependent on the developmental stage, 13 to 38% of the cellular ATP is used for protein synthesis, with plastid translation accounting for ∼70% of the costs of cellular protein synthesis. Synthesis of RbcL alone accounts for more than 15% of the cellular ATP equivalents used for protein synthesis (Li et al., 2017a). In view of the high costs of translation, the rate of protein synthesis is tightly coordinated to the cellular demands in all domains of life. In addition, translational regulation has several advantages. As translation is the final synthesis step in gene expression, its regulation (1) can mediate rapid responses to internal or external stimuli, (2) most directly affects the protein accumulation levels, and (3) can be readily exploited to enhance or attenuate changes in upstream steps of gene expression (i.e., transcription and transcript accumulation). Importantly, translation can also be dynamically localized to the site where the synthesized protein is needed, whereas transcription is largely bound to the position of the genomic DNA. Finally, although in bacterial systems transcriptional coregulation is easily achieved by combining genes in operons, this genomic organization is rather static and, in contrast to translationally coregulated mRNAs (qualifying as “regulons”), cannot readily mediate dynamic responses to different cues (Keene, 2007).

It is generally accepted that chloroplast gene expression is largely controlled and regulated at the posttranscriptional and, especially, the translational levels, contrasting gene expression in cyanobacteria, which is mainly transcriptionally regulated. Several lines of evidence led to this conclusion. First, bacterial transcripts tend to be unstable (with half-lives in the range of minutes), enabling transcriptional regulation (Pedersen and Reeh, 1978; Klug, 1993). By contrast, the half-lives of chloroplast transcripts are in the range of hours or even days, thus disabling fast transcriptional responses (Mullet and Klein, 1987; Klaff and Gruissem, 1991). Second, although chloroplast gene clusters are reminiscent of bacterial operons, they often are transcribed from different promoters (including operon-internal promoters) and, frequently, the resulting primary transcripts are further processed into smaller units (Barkan, 2011; Lyska et al., 2013; Börner et al., 2015). Also, chloroplast polycistronic transcription units often comprise functionally unrelated genes (Sugiura, 1995), unlike bacterial operons, where cotranscription serves to couple the expression of functionally related genes. Third, the translation of some chloroplast mRNAs coding for core components of the photosynthesis machinery is induced by light, whereas their mRNA levels remain virtually unaffected (Berry et al., 1988; Mühlbauer and Eichacker, 1998). Fourth, translation of many chloroplast mRNAs was shown to be the rate-limiting step in gene expression in Chlamydomonas (Eberhard et al., 2002), although the situation may be different in angiosperms (Udy et al., 2012). Moreover, in a regulatory mechanism termed control by epistasy of synthesis (CES), the translation rate of some subunits of photosynthetic protein complexes is regulated by the presence or absence of their assembly partners (Choquet and Wollman, 2009; see also below). Fifth, in recent years, a number of factors required for the translation of specific chloroplast reading frames were discovered, which may be suggestive of extensive translational regulation (Table 1).

RNA cis-Elements Controlling Translation

RNA sequence or structural elements that are located in cis, typically upstream of the reading frame, play crucial roles in translational control. A major cis-acting sequence element for translation initiation in chloroplasts is the SD sequence (Scharff et al., 2017; see above). Recent ribosome profiling studies have suggested that, in bacteria and chloroplasts, SD-like sequences that are located within reading frames cause programmed pausing of elongating ribosomes (Li et al., 2012; Zoschke et al., 2013a; Gawroński et al., 2018). It has been proposed that SD-dependent pausing facilitates cotranslational folding and targeting of nascent polypeptides (Li et al., 2012; Fluman et al., 2014). Additional chloroplast cis-acting sequence elements for translational control are the binding sites of trans-acting factors that stimulate translation of specific reading frames (Table 1).

The degeneracy of the genetic code offers another elegant means of influencing the translation rate in cis. Since an organism’s codon usage is usually adapted to the relative abundances of isoaccepting tRNAs, the choice of synonymous codons can potentially control translation efficiency (Supek, 2016). Whether or not this also applies to chloroplasts is currently controversially discussed (Sugiura, 2014; Suzuki and Morton, 2016; Gawroński et al., 2018). Although initial ribosome profiling studies could not confirm a robust correlation between codon adaptation and translational speed/efficiency, recent methodological and bioinformatic refinements uncovered the suspected relationship in several organisms (Dana and Tuller, 2014; Nakahigashi et al., 2014).

In addition to primary sequence elements, features of mRNA 2D or 3D structure (or lack of structure) can represent cis-elements that influence the translation process (Mauger et al., 2013). For example, the lack of secondary structures at the start codon facilitates efficient SD-independent translation initiation in bacteria and chloroplasts (Scharff et al., 2011, 2017). Furthermore, translation initiation in the chloroplast atpH and psbH mRNAs is stimulated by resolving secondary structures that mask ribosome binding site and start codon, respectively (Prikryl et al., 2011; Hammani et al., 2012). A recent ribosome profiling study also revealed transcriptome-wide correlations between ribosome pausing and secondary structures in chloroplast mRNAs (Gawroński et al., 2018).

Riboswitches are structured RNA elements that act as sensors for small molecules (metabolites). Metabolite binding triggers a conformational switch that regulates transcription (typically by inducing termination or antitermination) or translation (by exposing or sequestering the ribosome binding site; Serganov and Nudler, 2013). Although riboswitches have not yet been identified in plastids, they have been exploited as tools to control the translation of transgenes in chloroplast biotechnology (Verhounig et al., 2010). The use of improved algorithms for the computational prediction of riboswitches (Philips et al., 2013) should help to clarify whether endogenous riboswitches exist in chloroplast genomes.

Internal RNA structures within reading frames can adjust the translational speed by decelerating elongating ribosomes in bacteria and chloroplasts (Mauger et al., 2013; Gawroński et al., 2018), a mechanism that has been suggested to aid cotranslational protein-complex maturation steps. Specific translational pausing sites have been identified in the chloroplast psbA and atpA mRNAs (Kim et al., 1991; Kim and Hollingsworth, 1992). However, although these pausing events appear to be influenced by light or temperature (Kim et al., 1994a; Grennan and Ort, 2007), they could not be assigned to particular sequences or structural elements of either the mRNA or the nascent polypeptide chain nor could any obvious molecular function directly be ascribed to them. In bacteria, structured RNA elements are also involved in programmed frameshifting and translational coupling of neighboring reading frames on the same transcript (Jackson et al., 2007; Mauger et al., 2013).

cis-Acting Elements in the Nascent Peptide

Translational cis-elements can also be located in the nascent peptide. For example, translation of consecutive proline codons is complicated due to the physicochemical properties of the secondary amino acid proline, which make it a weak peptidyl donor and acceptor. Therefore, reading of consecutive proline triplets causes translational pausing (Artieri and Fraser, 2014, and references therein). Specific elongation factors such as EF-P in bacteria facilitate the translation of consecutive proline codons (Doerfel et al., 2013; Ude et al., 2013). EF-P is conserved in chloroplasts suggesting a similar function (Manuell et al., 2007). More than two consecutive proline codons are usually avoided in proteins of all organisms. For example, the entire tobacco chloroplast genome does not encode a single stretch of three prolines.

A recent study also provided evidence for other specific peptide motifs (e.g., domains comprised of small polar residues) inhibiting peptidyl transfer during elongation or peptide release during termination in bacteria (Woolstenhulme et al., 2013). Furthermore, specific translational pausing events that are induced by interactions of the prokaryotic ribosome with the nascent peptide have emerged as programmed switches that regulate translation in response to small metabolites or the availability of protein translocation factors (Ito et al., 2010). In chloroplasts, ribosome pausing events were correlated with positively charged amino acids (Gawroński et al., 2018).

Protein Factors That Control Translation in cis or trans

Proteins that bind the mRNA template, the ribosome, or the nascent peptide chain can influence translation. These proteins act either in an mRNA-specific manner or as more general factors that control translation of many transcripts.

All mRNA-specific chloroplast translation factors characterized to date act in trans by binding the 5′UTRs of one (or a few) mRNAs and promoting translation of the downstream reading frame (see above; Table 1). More generally acting trans-factors control the translation of several reading frames. These factors are either genuine or auxiliary constituents of the ribosome or bind the nascent peptide. Hence, these factors act in cis or trans. cis-regulatory functions have been suggested for several ribosomal proteins or ribosome-associated proteins such as PSRP2-6, pY, and RPS1 (Yamaguchi and Subramanian, 2003; Manuell et al., 2007; Sharma et al., 2010). In bacteria, Rps1 is involved in translation initiation of specific transcripts and a similar function has been suggested for chloroplast RPS1 (see above). Chloroplasts also contain an ortholog of the bacterial ribosome-associated protein pY that binds 70S ribosomes or 30S subunits under cold-shock conditions, thus inactivating translation and stabilizing monomeric 70S ribosomes (Vila-Sanjurjo et al., 2004, and references therein). Similarly, it has been proposed that plastid pY could be involved in the deactivation of translation and the stabilization of chloroplast ribosomes in the dark (Sharma et al., 2010, and references therein). Also, the expression of chloroplast translation elongation factors is regulated by light, suggesting their contribution to translational regulation. However, direct evidence for a regulatory function of chloroplast elongation factors RPS1 or pY is lacking.

In both bacteria and eukaryotes, the binding of specific protein factors can create subpools of specialized ribosomes that selectively translate a specific set of mRNAs (Sauert et al., 2015). More systematic and quantitative proteomic studies of chloroplast ribosomes will be required to determine whether specialized ribosomes exist also in plastids.

Translation Is Regulated at the Initiation and Elongation Levels

In theory, translation can be regulated at any of its three stages: initiation, elongation, or termination (Hershey et al., 2012). However, by far most common is the regulation of initiation. All chloroplast factors characterized to date that promote translation of specific reading frames act at the level of initiation (Table 1). However, in some cases, elongation also appears to be regulated. For example, elongation of psaA, psaB, psbA, and rbcL translation is regulated by light, and the synthesis of PsbA was suggested additionally to depend on the availability of cofactors and assembly partners (Berry et al., 1988; Klein et al., 1988b; Mullet et al., 1990; Kim et al., 1994a; Mühlbauer and Eichacker, 1998; Kim and Mullet, 2003).

Light- and Redox-Dependent Regulation of Translation

Light regulation of translation ensures coordination of the major energy-producing process (photosynthesis) with the major energy-consuming process (protein synthesis) in the chloroplast. In addition, components of the photosynthetic apparatus (especially PsbA) are damaged by light, necessitating specific repair synthesis.

The synthesis of several chloroplast proteins was shown to be activated at the translational level in response to increased light intensity. Upon illumination, translation of rbcL and psbA is activated at the level of elongation, a mechanism that was proposed to be regulated by the light-dependent generation of a proton gradient across the thylakoid membrane, the ATP status of the chloroplast and/or redox signal(s) generated by photosynthetic electron transfer (Taniguchi et al., 1993; Mühlbauer and Eichacker, 1998; Zhang et al., 1999). In line with regulation of elongation, the transfer of amaranth plants into the dark caused a decline in RbcL protein synthesis that was not accompanied by the loss of rbcL mRNA association with ribosomes (in polysomes), suggesting that elongating ribosomes paused (Berry et al., 1988). In addition, psaA and psaB were shown to be light-regulated at the elongation level during deetiolation (Klein et al., 1988b). A potential mechanism for the general light-dependent activation of chloroplast translation elongation could rely on pY (see above) or elongation factor Tu, which was identified as redox-regulated (Schröter et al., 2010).

During deetiolation, psbA translation is also stimulated at the level of initiation (Klein et al., 1988b; Kim and Mullet, 1994). Evidence currently available suggests that psbA translation for PSII biogenesis is regulated at the initiation level, whereas PsbA synthesis for PSII repair is regulated at the (presumably faster responding) elongation level, a model supported also by data from Chlamydomonas (reviewed in Nickelsen et al., 2014). In seed plants, psbA translation initiation is controlled by specific cis-elements in the 5′UTR (Staub and Maliga, 1994; Hirose and Sugiura, 1996; Eibl et al., 1999) and by the trans-factors HCF173 and HCF244, two putative oxidoreductases that promote psbA translation cooperatively (Schult et al., 2007; Link et al., 2012). All of these elements may also play a role in the light-dependent translation initiation of psbA.

Variations in light quantity and quality can cause imbalances in the activities of PSI and PSII, changing the redox state of the chloroplast and potentially generating harmful reactive oxygen species (Pfalz et al., 2012). The chloroplast redox state controls transcriptional and posttranscriptional adaptation responses in the chloroplast (Allen and Pfannschmidt, 2000; Rochaix, 2007). In Chlamydomonas, a redox-dependent regulatory mechanism has recently been suggested for the light-activated translation of psbD (Schwarz et al., 2012). In this model, the RNA binding proteins Nac2 and RBP40 form a disulfide bridge-connected complex in the light that stabilizes psbD mRNA and activates its translation (Schwarz et al., 2012, and references therein). In the dark, the complex is reduced and disassembles, and, as a result, PsbD synthesis is inactivated. The electrons are likely provided by the NADPH-dependent thioredoxin reductase C, an enzyme shown to reduce disulfide bonds in the dark by utilizing NADPH generated in the oxidative pentose phosphate pathway (Schwarz et al., 2012).

Plants perceive light independently of photosynthesis by photoreceptors that are located outside of chloroplasts. However, light-induced regulation of translation was also observed in isolated chloroplasts, arguing against a crucial role of photoreceptors (Sun and Zerges, 2015).

Altogether, our current knowledge about light-induced translational regulation in seed plant chloroplasts is restricted to very few photosynthesis-related genes, with psbA translation being best studied. However, even for psbA, the underlying mechanisms and the mode of action of the factors involved are not well understood.

Developmental Regulation of Translation

Developmental regulation of plastid gene expression is crucial for the differentiation and interconversion of plastid types. For example, conversion of proplastids to chloroplasts requires establishment of the thylakoid system and must be tightly coordinated with cell division and the emergence of photosynthetic tissues from meristems (Jarvis and López-Juez, 2013). The mechanisms involved in chloroplast differentiation were identified in classical experiments that analyzed the process of deetiolation. During deetiolation, a rapid rise in translational activity was observed by pulse labeling experiments for several plastid mRNAs encoding subunits of PSI (psaA/B), PSII (psbA/B/C/D), and Rubisco (rbcL; Klein and Mullet, 1986, 1987; Kim and Mullet, 2003; Kleffmann et al., 2007). The initial steep increase in translation of these mRNAs was followed by a slow decline as chloroplast differentiation continued in the light. Polysome analyses demonstrated that psaA/B, rbcL, and psbA transcripts are found in polysomes throughout the chloroplast differentiation process, suggesting that the regulation of their translation occurs, at least in part, at the elongation level (Klein et al., 1988b). This conclusion was further substantiated by the observation that psaA, psbA, and rbcL translation initiation remains unaltered during light-dependent chloroplast differentiation (Kim et al., 1994b; Kim and Mullet, 2003).

There is also evidence for developmentally activated translation initiation of plastid mRNAs. Initiation of psbA translation is induced during deetiolation (Eichacker et al., 1992; Kim and Mullet, 1994), and this induction is likely controlled by cis-elements in the psbA 5′UTR (Staub and Maliga, 1994, and references therein). It was further proposed that the light-induced synthesis of chlorophyll controls both the accumulation and the translation of chlorophyll binding apoproteins of PSI and PSII during deetiolation (Eichacker et al., 1992). However, this conclusion was mainly based on the observed lack of accumulation of chlorophyll binding apoproteins in the absence of chlorophyll, as determined by protein immunoblotting and pulse labeling experiments in vivo and in vitro (Klein et al., 1988a, 1988b; Eichacker et al., 1990, 1992). Although pulse labeling can reveal protein synthesis rates, it cannot unambiguously distinguish between the absence of synthesis and rapid degradation of newly synthesized proteins (especially not for proteins with high turnover rates, such as PsbA). A recent ribosome profiling analysis of plastid translation in a maize mutant with knocked-out chlorophyll synthesis showed that the synthesis of plastid-encoded chlorophyll binding apoproteins is virtually unaltered in the absence of chlorophyll (Zoschke et al., 2017). Even the cotranslational thylakoid membrane engagement of nascent apoproteins was shown to be independent of chlorophyll synthesis (Zoschke et al., 2017). However, apoproteins were shown to undergo rapid degradation in the absence of chlorophyll (Mullet et al., 1990; Kim et al., 1994b; Eichacker et al., 1996).

The above described deetiolation of seedlings represents a dramatic change in plant development and reflects the situation during germination, when light-dependent and developmental programs operate simultaneously. Another useful experimental system to study translational regulation during chloroplast differentiation is provided by the longitudinal developmental gradient of young leaves in grasses such as maize, rice (Oryza sativa), or barley (Hordeum vulgare). At the leaf base, meristematic, proplastid-containing tissue is found, followed by a developmental gradient toward the leaf tip that includes etioplasts, differentiating chloroplasts (etiochloroplasts), and fully differentiated chloroplasts (Leech et al., 1973). This natural developmental gradient has been exploited to systematically study the dynamic changes in gene expression at the level of transcript and protein accumulation (Barkan, 1989; Baumgartner et al., 1989; Cahoon et al., 2004; Li et al., 2010; Majeran et al., 2010).

Recently, ribosome profiling enabled the transcriptome-wide examination of plastid translation in the longitudinal developmental gradient of maize leaves (Chotewutmontri and Barkan, 2016). This pioneering study comprehensively determined the relative contributions of changing transcript levels and translational activity to the protein synthesis output (Chotewutmontri and Barkan, 2016). The data revealed that the synthesis rates of most plastid-encoded proteins increases early in development and drops later, once the photosynthetic machinery is set up. Strikingly, PsbA and PetD, two proteins that do not follow this general rule and instead display increasing synthesis levels that peak in the latest examined developmental stage, represent the subunits with the highest turnover rates in their complexes (Li et al., 2017a), likely explaining their continued production.

In general, the regulation of plastid gene expression during leaf development in maize is achieved by changes in transcript levels that are often tuned by changes in translation efficiency (Chotewutmontri and Barkan, 2016). Remarkably, the developmental dynamics in protein synthesis that was discovered revealed two major regulons that are defined by changes in RNA accumulation and translation efficiency: Early in development, proteins needed for chloroplast gene expression have the highest synthesis output (e.g., RNA polymerase subunits and ribosomal proteins), whereas in later developmental stages, proteins required for photosynthesis are extensively synthesized (i.e., subunits of PSII, Cyt b₆f, PSI, ATP synthase, and NDH complexes). To achieve this developmental pattern in plastid protein synthesis, translational regulation is especially important for polycistronic transcription units that encode proteins of both regulons. Indeed, in these transcripts (e.g., the psaA/psaB/rps14 transcription unit encoding two PSI subunits and a ribosomal protein), strong differential translational regulation was observed (Chotewutmontri and Barkan, 2016).

Unfortunately, none of the factors involved in the complex developmental regulation of chloroplast translation has been identified so far. It would be particularly exciting to reveal the mechanisms underlying the translational switch between these two major regulons during leaf development. A possible explanation could lie in the known developmental switch in the usage of the two distinct types of RNA polymerases during chloroplast differentiation (Börner et al., 2015). In theory, the two polymerase activities could produce transcripts coding for identical proteins but differing in their translational competence. Alternatively, RNA binding proteins that facilitate translation (see above and Table 1) could mediate the developmental regulation of translation.

Translational Autoregulation and Feedback Regulation

Control of protein synthesis by translational autoregulation or feedback regulation represents an elegant means of robustly fine-tuning protein synthesis levels to changing cellular requirements. In the chloroplast of Chlamydomonas, the translation of specific subunits of the four photosynthetic complexes (PSII, Cyt b₆f, PSI, and ATP synthase) was shown to be feedback regulated by the assembly state of the respective complex, a process termed control by epistasy of synthesis (CES; Choquet and Wollman, 2009). CES regulation ensures the stoichiometric production of chloroplast subunits that reside in oligomeric complexes according to the requirements of their sequential assembly (Choquet and Wollman, 2009). The best-studied case mechanistically is the CES involved in petA gene expression. PetA is a subunit of the Cyt b₆f complex, whose synthesis is strongly reduced in the absence of its assembly partners PetB (cytochrome b₆) and PetD (subunit IV; Kuras and Wollman, 1994). The protein factors MCA1 and TCA1 were shown to be cooperatively associated with the 5′UTR of the petA mRNA, where they stabilize the transcript (MCA1) and promote translation initiation (TCA1; Wostrikoff et al., 2001; Loiselay et al., 2008). Boulouis et al. (2011) discovered a regulation mechanism in which a C-terminal domain of PetA that is exposed only in the unassembled PetA protein binds to MCA1 and causes its proteolytic degradation. In turn, this leads to downregulation of petA gene expression at both the level of transcript accumulation and the translational level. Conversely, if the assembly partners PetB and PetD are available, PetA is assembled into the Cyt b₆f complex, the C-terminal domain is inaccessible, and, hence, MCA1 remains stable and facilitates petA expression in complex with TCA1. In this way, the CES mechanism ensures the adequate synthesis of PetA according to the availability of its assembly partners in the Cyt b₆f complex.

In chloroplasts of seed plants, evidence for a similar feedback regulation mechanism has been obtained in only a single case. The synthesis of the plastid-encoded large subunit of Rubisco (RbcL) in tobacco is adjusted to that of its nucleus-encoded assembly partner, the small subunit RBCS, by a mechanism that is similar to the CES-regulated rbcL expression in Chlamydomonas (Rodermel et al., 1988, 1996; Khrebtukova and Spreitzer, 1996). Furthermore, evidence was provided that unassembled RbcL represses its own translation, possibly through direct RNA binding (Yosef et al., 2004; Wostrikoff and Stern, 2007). In a systematic genetic approach, the essential plastid-encoded Cyt b₆f subunits PetA, PetB, and PetD were knocked out to disrupt complex assembly and test for potentially conserved CES regulatory mechanisms in tobacco (Monde et al., 2000). In the mutants, only a mild effect on translation of the polycistronic petA transcript was detected by polysome analyses, suggesting that the CES mechanism is not conserved in seed plants (Monde et al., 2000). This may not be entirely surprising, given that the regulatory trans-factors involved, MCA1 and TCA1, are not found in seed plants. Moreover, a ribosome profiling analysis of two maize mutants with defective AtpB synthesis did not show a substantial effect on atpA translation (Zoschke et al., 2013a). This suggests that also the AtpB-dependent translation of atpA, the only trans-activating CES mechanism observed in Chlamydomonas (Drapier et al., 2007), is not conserved in seed plants. However, in Arabidopsis, it was observed that mutation of the chloroplast RNA binding protein HCF107 that facilitates psbH translation also causes a reduction in PsbB (CP47) synthesis (Felder et al., 2001). Remarkably, when psbH expression was rescued by introducing a psbH gene copy into the nuclear genome of an hcf107 mutant, accumulation of the PsbB protein was rescued as well, despite the absence of functional HCF107 protein (Levey et al., 2014). A possible explanation for this observation is that PsbH accumulation is needed for PsbB synthesis, pointing to a potential CES mechanism in PSII biogenesis (Levey et al., 2014). Interestingly, recent transcriptome-wide analyses of translation in maize and Arabidopsis chloroplasts revealed that most proteins are synthesized in the amounts that correspond to (1) their steady state stoichiometry as subunits of protein complexes and (2) the abundance of the respective protein complex (Chotewutmontri and Barkan, 2016; Lukoszek et al., 2016). This suggests that precise regulation of gene expression is more important than posttranslational proteolytic adjustments of subunit stoichiometry. However, several alternative regulatory mechanisms can explain this behavior, and it remains to be examined whether or not CES is involved.

In many organisms, autoregulated proteins have been observed to comprise an intrinsic RNA binding activity. Prominent examples include several ribosomal proteins, the initiation factor IF3 and the β-subunit of the bacterial RNA polymerase (Choquet and Wollman, 2009). The RNA binding motifs allow these proteins to bind the 5′UTR of their own mRNA and inhibit translation initiation whenever their assembly partners are not available, thus generating a negative regulatory feedback loop. The chloroplast genome codes for numerous RNA binding proteins (e.g., orthologs of bacterial RNA polymerase subunits and ribosomal proteins). Hence, it is tempting to speculate that some of them may also autoregulate their own synthesis. A plastid-encoded protein that has been shown to associate with its own RNA in vivo is the putative splicing factor MatK (Zoschke et al., 2010). The maturase MatK is encoded in the intron of the trnK gene. MatK binds to the trnK transcript, most likely assisting in splicing (Zoschke et al., 2010, and references therein). Notably, a MatK-related bacterial maturase is translationally autoregulated (Singh et al., 2002), and an autoregulatory mechanism was also proposed for the expression of plastid MatK (Hertel et al., 2013).

In summary, whether autoregulation and/or feedback regulation of translation are common in seed plant chloroplasts or whether these regulatory mechanisms are specific to Chlamydomonas remains to be investigated. Pronounced differences between microalgae and seed plants may not be surprising if one considers the dissimilar evolutionary forces that act on regulatory circuits in a unicellular organism with one chloroplast per cell compared with multicellular organisms with numerous chloroplasts per cell.

Translational Regulation in Response to Other Internal and External Triggers

Temperature is a major abiotic signal influencing translational regulation. In bacteria, cold-stress conditions induce a general block of translation that is mediated by the pY protein. However, the synthesis of a small subset of proteins, such as RNA helicases, IF1-3, and trigger factor, is induced under these conditions (Barria et al., 2013). In chloroplasts, changes in temperature alter the speed of all enzymatic reactions (including the Calvin-Benson-Bassham cycle), but, unless extreme, they have virtually no impact on the light reactions of photosynthesis. Consequently, temperature changes can cause dramatic imbalances in the redox homoeostasis of photosynthesis (Crosatti et al., 2013). It has long been known that the chloroplast translation capacity is crucial to plant adaptation to chilling stress. This became evident by the discovery of numerous mutants with impaired chloroplast translation that exhibit cold-sensitive phenotypes (Barkan, 1993; Rogalski et al., 2008; Liu et al., 2010).

A recent ribosome profiling study comprehensively examined chloroplast translational regulation upon temperature stress (Lukoszek et al., 2016). Ignatova and coworkers observed in Arabidopsis that elevated temperature causes dramatic changes in the protein synthesis output of subunits of photosynthetic complexes (Lukoszek et al., 2016). This regulation was driven by both changes in transcript levels and translational regulation. Interestingly, in several cases, unbalanced changes in protein synthesis rates were observed for different subunits within a complex, in that the altered synthesis rates did not reflect the subunit stoichiometry. This finding may suggest an increased relevance of control at the level of protein turnover of at least some subunits during heat stress (Lukoszek et al., 2016). In sum, although it is well established that plastid translation is critical for acclimation to changing temperature conditions, the molecular mechanisms and the factors involved are currently unknown.

A special case of internal specialization of gene expression is the chloroplast dimorphism in C₄ plants. In C₄ species such as maize, light reactions and carbon reactions of photosynthesis are partitioned between mesophyll cells and bundle sheath cells. The division of labor between the two cell types is reflected by specialized chloroplast types, referred to as chloroplast dimorphism. At the molecular level, bundle sheath chloroplasts have a high content of Rubisco and NDH complex, whereas mesophyll chloroplasts are enriched in PSII (Majeran et al., 2010). This specialization is achieved by differential gene expression (Sharpe et al., 2011, and references therein). Recently, transcript accumulation and translation have been comprehensively analyzed in bundle sheath and mesophyll chloroplasts by ribosome profiling (Chotewutmontri and Barkan, 2016). This study confirmed that most of the observed differences in protein accumulation can be explained by differential protein synthesis, which in turn is predominantly achieved by differential expression at the RNA level. However, the synthesis of some proteins such as the PSII core subunits PsbA/B/C/D is additionally tuned by translational regulation. The underlying regulatory mechanisms are unknown. The study showed that the synthesis output of Rubisco, PSII, PSI, and NDH complex subunits reflects well their protein accumulation in the different chloroplast types. This is not the case for the subunits of the ATP synthase and the Cyt b₆f complex, suggesting more extensive posttranslational adjustments in these complexes.

Internal metabolic signals can also control translation (e.g., via riboswitches in bacteria; Serganov and Nudler, 2013). Several recent discoveries in Chlamydomonas have established interesting links between chloroplast metabolism and gene expression (Bohne and Nickelsen, 2017). DLA2, a subunit of the chloroplast pyruvate dehydrogenase complex, was shown to have RNA binding activity which enables its moonlighting function in translational regulation of the psbA mRNA (Bohne et al., 2013). Considering its functions in metabolism and translational regulation, DLA2 is anticipated to coordinate fatty acid and protein syntheses during thylakoid biogenesis (Bohne et al., 2013). The phylogenetic conservation of the DLA2 amino acid sequence and RNA binding properties suggests that it performs a similar dual function in seed plants (Bohne et al., 2013). A moonlighting function was also suggested for RbcL, which was shown to bind RNA (Yosef et al., 2004; Cohen et al., 2006) and recently has been proposed to be involved in the redox stress-induced localization of oxidized chloroplast RNAs in Chlamydomonas, independent of its function in the Rubisco complex (Zhan et al., 2015). Moreover, recent in vitro data suggest that PsbD protein synthesis in Chlamydomonas is metabolically controlled (Schwarz et al., 2012; see above). Whether similar regulatory connections between metabolism and chloroplast translation exist in seed plants remains to be investigated.

In bacteria, stress-induced regulation of gene expression is triggered by the so-called stringent response. Upon nutrient deprivation and other stresses, bacteria produce the effector molecule (p)ppGpp (guanosine tetraphosphate/pentaphosphate; also known as alarmone) that coordinates numerous cellular responses, including broad changes in transcription and translation (Gaca et al., 2015). Chloroplasts also harbor the enzymes needed for ppGpp synthesis (and degradation). Plastids produce alarmone not only in response to stress, but also to orchestrate chloroplast and nuclear gene expression (reviewed in Field, 2017). ppGpp has been shown to influence transcription, translation, and many other metabolic and physiological processes in plants (Field, 2017). In an in vitro chloroplast translation system from pea (Pisum sativum), ppGpp inhibited protein synthesis, suggesting conservation of the bacterial mode of ppGpp-mediated translational repression by interaction with IF2 and/or EF-G (Nomura et al., 2012). Furthermore, overaccumulation of ppGpp in vivo caused a dramatic reduction in the accumulation of many plastid-encoded proteins, mRNAs, tRNAs, and rRNAs, suggesting an overall reduced translation capacity (Sugliani et al., 2016). Whether the translational activity of specific chloroplast mRNAs is regulated by ppGpp and how this contributes to ppGpp-induced changes in chloroplast gene expression still needs to be determined.

In summary, considerable progress has been made in describing changes in translation in response to light and developmental signals, but comparably little is known about the mechanisms and factors that adjust plastid protein synthesis to other internal and external cues that affect protein homeostasis.

OUTLOOK

Over the last decades, extensive research on the structure and function of the translational apparatus of bacteria and chloroplasts has revealed many similarities but also substantial differences. Unfortunately, our knowledge about genuine regulatory processes, the underlying molecular mechanisms and the factors involved, is still limited. For example, the influence of internal and external triggers such as light, temperature, nutrient availability, osmotic status, redox state, phytohormones, alarmone, and diurnal and circadian rhythms on chloroplast gene expression has so far not been comprehensively examined at the translational level. Likewise, investigation of the interconnection of chloroplast translation with protein synthesis in the cytosol and the mitochondria is likely to provide fresh insights into intracellular signaling and crosstalk in response to changing environmental conditions and developmental programs. Last but not least, the detailed molecular function of RNA cis-elements and proteinaceous trans-factors in the regulation of chloroplast translation is mostly unknown. The exciting advent of chloroplast ribosome profiling and other in vivo ribonomic methods, such as the comprehensive examination of RNA secondary structures or the mapping of binding sites for RNA binding proteins, can be anticipated to address many of these open questions in the biology of plant organelles in the future.