2.1. Plants Observe the Scanning Model of Translation

The plant cytosolic translation apparatus consists of ribosomes, associated translation factors, and tRNAs. Most of the plant translation machinery resembles that of other higher eukaryotes such as human and yeast. Translation follows Kozak's scanning model (Figure 2) (Kozak, 1992; Hinnebusch and Lorsch, 2012). For the remainder of this section, recent references are cited simply to serve as a primer into the plant literature, since the factual statements qualify as pan-eukaryotic textbook knowledge. The 5' cap of the mRNA is recognized by the elF4F complex, which consists of the cap binding protein, elF4E, and a large scaffold protein, elF4G (Patrick and Browning, 2012). In preparation for initiation, the small (40S) subunit of the ribosome is loaded with a charged methionyl-tRNA, which is delivered by the trimeric GTP-binding protein, elF2. Several additional initiation factors, elF1 and elF1A, elF3, and presumably elF5 are also preloaded onto the 40S in the form of a multifactor complex (Dennis et al., 2009). The 40S subunit of the ribosome is brought to the cap by the largest initiation factor, the 13-subunit elF3 complex (Burks et al., 2001). Next, the 40S ribosome scans down the mRNA in a 5' to 3' direction, a process that is facilitated by the ATP-dependent helicase elF4A and its cofactor, elF4B (Bush et al., 2009; Khan et al., 2009; Mayberry et al., 2009; Vain et al., 2011). The ribosome scans until its initiator tRNA basepairs with an AUG start codon in a favorable context (Kozak, 1984, 1989; Sugio et al., 2010), a process controlled by the initiation factors elF1 and elF1A. Upon start codon recognition, the elF5 protein stimulates GTP hydrolysis by elF2. Most, if not all, elFs now dissociate from the mRNA. Next, the large 60S ribosomal subunit can join. Subunit joining is catalyzed by another GTPase, elF5B, thus completing translation initiation. At this stage, the methionyl tRNA resides in the P(eptidyl)-site of the assembled 80S ribosome. The A(minoacyl)-site of the ribosome is free to bind the second charged tRNA.

During the subsequent elongation phase, the enzymatic peptidyltransferase activity of the ribosome catalyzes peptide bond formation between the two amino acids. The ribosome translocates by one codon, thereby shifting the spent initiator tRNA into its E(xit)-site, and opening up a new, empty, A-site. The elongation factors eEF1 and eEF2 mediate tRNA delivery and ribosome translocation, respectively. The elongation cycle repeats until the ribosome reaches a stop codon.

At termination, peptide release requires ribosome release factors. Translation ends with the separation of the two subunits and their dissociation from the mRNA, a process known as recycling. The poly(A)binding protein (PABP) forms a bridge with the initiation factor elF4G, thus bending the mRNA into a closed loop, which is thought to stimulate the recycling of the post-termination ribosomal subunits to the 5' end of the mRNA for the next round of translation (Cheng and Gallie, 2007, 2010). Very recently, a potential translation termination and ribosome recycling factor homologous to metazoan ABCE1 was cloned as SIMPLE LEAF3/RLI2, because its mutation interferes with leaf lobing in the Arabidopsis relative, Cardamine hirsuta (Kougioumoutzi et al., 2013). Under exceptional circumstances the ribosome is able to resume scanning after termination and reinitiate translation farther downstream on the same mRNA, a process with regulatory significance (Wang and Wessler, 1998; Ryabova and Hohn, 2000; Roy et al., 2010; Schepetilnikov et al., 2011). The molecular events during reinitiation and recycling are just coming into focus in other organisms.

Figure 2.

The translation machinery and selected kinase pathways. The small subunit of the ribosome (40S) is charged with a methionyltRNAMet with the help of elF2 (ternary complex) and loaded with additional initiation factors in the form of a multifactor complex (MFC) to form the 43S preinitiation complex (PIC). The PIC is recruited to the 5' end of the mRNA by the cap-binding complex (elF4F), which consists of a member of the elF4E and elF4G protein families. elF4A is a loosely associated helicase and elF4B an accessory protein. The 43S complex scans the mRNA in search of an AUG start codon. Upon start codon recognition, which is thought to be facilitated by elF1, elF5 stimulates GTP hydrolysis by elF2. elF2 and the other initiation factors dissociate from the 40S. Joining of the 60S large ribosomal subunit is mediated by the GTPase elF5B. Translation elongation involves cycles of tRNA delivery (by eEF1), peptide bond formation, and ribosome translocation by one codon (triggered by eEF2). Translation termination requires release factors and leads to ribosome subunit dissociation from the mRNA. Kinases known to phosphorylate components of the machinery include casein kinase 2 (CK2), GCN2, and S6 kinase. Phosphorylation of elF2 by GCN2 inhibits the activity of elF2's guanine nucleotide exchange factor, elF2B.

2.2. Aspects of Translation Unique to Plants

The basic events of translation are conserved between fungi, metazoans, and plants, with a few well-documented plant-specific modifications. Most importantly, plants possess two quite distinct protein complexes for 5' cap recognition, a classical elF4F, and a smaller yet more abundant elFiso4F. Depletion of both elF4E and elFiso4E in tobacco with RNA interference results in semi-dwarf phenotypes and reduces polyribosome loading overall (Combe et al., 2005). The reduction in plant growth and development may be attributable to an overall reduction in protein synthesis. Likewise, mutation of Arabidopsis elFiso4G reduces the protein synthesis rate and causes slower growth and dwarfism (Mayberry et al., 2009; Lellis et al., 2010). It is noteworthy that the developmental phenotypes of a mutation in a core initiation factor, elFiso4G, do not look identical to those of typical ribosomal protein mutations (Horiguchi et al., 2012). Ribosomal mutations typically cause pointed leaves and a variety of developmental phenotypes that are not apparent in the eifiso4g mutant. In other words, defects in translation initiation can be genetically uncoupled from more global defects in ribosome biogenesis.

The plant elF4E and elFiso4E proteins each harbor a pair of cysteines, which, when oxidized, form a disulfide bond that may impede guanine nucleotide cap binding, suggesting that plant elF4E may function as a redox sensor of translation initiation (Monzingo et al., 2007).

The elF4F and elFiso4F complexes differ quantitatively in their activities across a panel of client mRNAs, which can be attributed largely to the respective elF4G isoform (Mayberry et al., 2009; Martinez-Silva et al., 2012). Different client specificities are also apparent for isoforms of elF4B (Park et al., 2004; Mayberry et al., 2009). Therefore any changes in the relative cellular abundance of these elF isoforms may alter the translation states of different mRNAs in the cell.

The elF4F complex in particular is targeted by viruses, which rely on the translation apparatus of their host. In several virushost pairs, the virus preferentially relies on a particular isoform of elF4E for optimal virulence; vice versa, a mutation of the preferred elF4E isoform confers resistance to the virus. This welldocumented phenomenon is not covered comprehensively here because it has been reviewed periodically elsewhere (Robaglia and Caranta, 2006; Wang and Krishnaswamy, 2012). When the plant offers two distinct elF4F isoforms, the virus, being under tremendous selection pressure, may evolve quickly to prefer one isoform to the other. In this situation, the isoform that is not coopted by the virus would remain available for translation of host proteins, including those needed to mount an antiviral defense.

2.3. The Ribosome

The Arabidopsis ribosome possesses a typical complement of four ribosomal RNAs and 80 ribosomal proteins, which mostly decorate the surface of the ribosomal RNAs. The proteins of the Arabidopsis small (RPS) and large (RPL) subunits are encoded by between two and seven paralogous genes each. While nearly all rRNAs and proteins are highly conserved with other eukaryotes, there is one plant-specific protein, P3, in the large subunit (Bailey-Serres et al., 1997). Naming of proteins is not entirely uniform in the literature, but a consensus was emerging (Barakat et al., 2001; Carroll et al., 2008; Klinge et al., 2012). The recent X-ray crystal and cryo-electron microscopy structures of eukaryotic ribosomes, including that of wheat (Armache et al., 2010a, b) have transformed our understanding of ribosome structure and will continue to shape future research on this most fundamental cellular machine (Klinge et al., 2012; Wilson and Doudna Cate, 2012). As a result, the locations of 74 ribosomal proteins are now known with precision. Proteomic and transcriptomic studies have yielded evidence that the isoform profile of ribosomal proteins can be modified by cell type, cellular localization, and environmental conditions such as pathogens, nitrogen, UV-B light, and sucrose (Szick-Miranda and Bailey-Serres, 2001; Whittle and Krochko, 2009; Falcone Ferreyra et al., 2010; Sormani et al., 2011a; Hummel et al., 2012). In a number of case studies, scientists have had reason to conclude that two different protein isoforms confer different functions to the ribosome; however, it is not easy to distinguish between the effect of different RNA expression characteristics and the slight differences in amino acid sequences (Barakat et al., 2001; Degenhardt and Bonham-Smith, 2008; Falcone Ferreyra et al., 2010; Horiguchi et al., 2012). Regulation of ribosome functions may also occur through posttranslational modifications. While several proteins are phosphorylated, differential phosphorylation is well established only for a small subset, primarily RPS6 (Bailey-Serres and Dawe, 1996; Williams et al., 2003; Turkina et al., 2011), and the biochemical consequence of such modifications on translation remains unknown.

3.1. Early Evidence for Translational Control in Plants

At a time when Arabidopsis was still in its infancy as a plant model organism, evidence for translational regulation of gene expression surfaced in a variety of other plant species. These early observations established a conceptual framework for translational regulation that would later be explored using the tools of Arabidopsis, as explained in a subsequent section. Regulation of translation by light was being studied in plants and algae (reviewed in (Wobbe et al., 2008)) and was first discovered for Rubisco small subunit genes (Berry et al., 1986; Elliott et al., 1989; Berry et al., 1990). Of note, the Rubisco model system also proved that translational control may be cell-type specific, given that translational control contributes to the preferential expression of Rubisco in bundle sheath versus mesophyll cells in the C4 plant, amaranth (Patel et al., 2006). The light harvesting chlorophyll binding proteins of photosystems I and II, as well as ferredoxin, are regulated translationally. When plants are shifted from light to darkness, these and other mRNAs are released from polysomes, some within less than 30 minutes (Petracek et al., 1997; Hansen et al., 2001). Upon reillumination, their reassembly into actively translating polysomes requires photosynthetic electron transport (Petracek et al., 1997; Petracek et al., 1998). RNA sequence elements that control posttranscriptional regulation of these mRNAs include a light regulatory element that mediates reversible translation repression and a CATT repeat in the 5' untranslated region (UTR) of pea ferredoxin, which affects mRNA stability (Hansen et al., 2001 ; Bhat et al., 2004). A heat shock protein may stimulate the translation of the ferredoxin mRNA under light (Ling et al., 2000).

Besides light and abiotic stresses to be covered below, several other environmental conditions were known to regulate translation, i.e. hypoxia, reactive oxygen, and heat. For example, under hypoxia, maize polysome loading in general is inhibited yet the alcohol dehydrogenase mRNA specifically is induced (Bailey-Serres and Dawe, 1996). In response to reactive oxygen, catalase mRNA is regulated by an unusual mechanism, methylation of the 5' cap (Schmidt et al., 2006). Heat shock triggers a rapid and broad inhibition of translation. The heat shock response has an interesting cell biological correlate. Upon heat shock, many mRNAs are sequestered transiently into stress granules. These are ribonucleoprotein complexes that contain heat shock proteins in which translation is stalled and mRNA is stabilized (Nover et al., 1989; Gallie et al., 1995; Gallie and Pitto, 1996; Stuger et al., 1999). The translational block upon heat stress is not universal. In heat-shocked barley aleurone cells, for example, mRNAs for secreted proteins appear to be affected preferentially. Translation of these mRNAs stalls after synthesis of the signal peptide and targeting to the ER membrane (Chu et al., 1997). In species where the translational block by heat shock is more global, it is the heat shock protein mRNAs that remain polysomal, due to sequence elements that may reside in the 5' UTR (Dinkova et al., 2005).

In the early years, only a small number of RNA sequence elements could be defined in detail, and it was not always obvious whether the elements affected only translation, or also RNA stability, or other events (Kawaguchi and Bailey-Serres, 2005). One class of such elements that undoubtedly affect translation are upstream open reading frames (uORFs), e.g. in maize R/Lc (Wang and Wessler, 1998, 2001). uORFs are most common in mRNAs for transcription factors and other regulatory genes and typically inhibit translation. Several early case studies demonstrated that uORFs can render translation sensitive to small molecules such as sucrose or polyamines (Rook et al., 1998b; Hanfrey et al., 2005) (discussed in more detail below).

When it comes to identifying sequence elements for translational control, viruses are a rich source. Because comprehensive coverage of viral translation mechanisms such as capindependent translational enhancement, alternative initiation, shunting, ribosomal frameshifting, and stop codon readthrough would go well beyond the scope of this chapter, the reader is referred to earlier reviews (e.g. (Dreher and Miller, 2006; Ryabova et al., 2006; Nicholson and White, 2011; Simon and Miller, 2013). Only a few concepts from Arabidopsis as a host species will be mentioned. Viruses vary in their use of a 5' cap and a poly(A) tail, utilizing, for example, a genome linked Vpg protein or a tRNA shaped structure instead. Viruses have also evolved unconventional mechanisms to stimulate translation in the absence of a 5' cap structure. Many utilize cap-independent translational enhancer elements in the 3' UTR or in the 5' UTR. While cellular mRNAs are thought to form a closed loop via the PABP-elF4G bridge, turnip crinkle virus attracts the 60S ribosomal subunit via a tRNA like structure in the 3' UTR, which then basepairs with a translational enhancer in the 5' UTR to stimulate translation (Stupina et al., 2008; Zuo et al., 2010; Stupina et al., 2011). It will be interesting to discover whether similar mechanisms are utilized by plant encoded mRNAs. Two separate, and fairly unconventional, translational control systems were discovered early on in cauliflower mosaic virus, (CaMV, early work reviewed in (Hohn and Futterer, 1992)). The CaMV 5' leader contains several short upstream open reading frames (uORFs). These do not simply disrupt scanning by 40S ribosomes. Instead, the uORFs set the stage for the 40S ribosome to bypass, i.e. ‘shunt’ past, a large section of the 5' UTR. The bypassed section of the mRNA forms an elaborate stem-loop structure, and the post-termination ribosome appears prone to scan past it without unfolding it (Fütterer et al., 1993; Dominguez et al., 1998; Pooggin et al., 2000; Ryabova and Hohn, 2000).

CaMV's second translational anomaly is translation reinitiation (Dixon and Hohn, 1984). The CaMV genome gives rise to a single long mRNA with multiple protein coding ORFs, an exceptional case for any eukaryote. CaMV accomplishes this by stimulating the ribosome to reinitiate after translation termination. A viral gene product, named transactivator (TAV), which is expressed from a monocistronic (single-ORF) mRNA, is required (Bonneville et al., 1989). TAV binds to polysomes in virus infected cells and directly interacts with the g subunit of elF3 and with the RPL24 protein in a manner that is sensitive to elF4B (Park et al., 2001; Park et al., 2004). Remarkably, TAV can stimulate the association of the elF3 complex with the 80S ribosome, an event predicted by many models of translation reinitiation (Park et al., 2001; Park et al., 2004; Ryabova et al., 2006). TAV does not act alone to stimulate translation reinitiation, but in association with a host factor, re-initiation supporting protein (RISP), whose biochemical interactions suggest that it can bridge between RPL24 on the 60S large subunit and elF3, which is typically bound to the 40S small subunit (Thiebeauld et al., 2009).

These early case studies established that translational control is accentuated in specific biochemical domains of cellular metabolism; that translation is regulated by external conditions, especially abiotic stresses; that gene-specific mRNA sequence elements must exist; that viruses are particularly poised to utilize translational control mechanisms; and finally, that translational control might be harnessed in a cell-specific context to regulate development.

3.2. Translational Control in Response to Abiotic Stress

Translation is arguably the most energy-intensive biochemical process supporting cell growth. It requires a large fraction of the energy budget as well as a reliable supply of all twenty amino acids. Moreover, the nascent proteins must be folded, assembled, covalently modified, and sorted to the correct cellular compartment. In many eukaryotic organisms, bulk translation is impeded when an unfavorable physical environment compromises translation or processing. Likewise, in plants, many abiotic stresses, such as osmotic stress (Kawaguchi et al., 2003; Matsuura et al., 2010b), cause a dramatic change in polysome loading that affects the bulk of cellular mRNAs. The effects on translation of several environmental conditions have been examined at the global level in Arabidopsis; heat and salt (Matsuura et al., 2010a), drought (Kawaguchi et al., 2004), cadmium heavy metal (Sormani et al., 2011b), hypoxia (Branco-Price et al., 2005; Branco-Price et al., 2008; Mustroph et al., 2009), light and darkness (Juntawong and Bailey-Serres, 2012; Liu et al., 2012), as well as sucrose deprivation (Nicolai et al., 2006). In general, the translational response is often widespread, yet not universal. In every case, a certain fraction of mRNAs appears to escape from the translational block under the stress. To give one example, upon heat stress, heat shock protein mRNAs continue to be actively translated, whereas other mRNAs are sequestered into stress granules and translationally suppressed (Nover et al., 1989; Matsuura et al., 2010a; Sormani et al., 2011a). Some sequence elements responsible for continued translation under heat shock reside in the 5' UTR (Matsuura et al., 2013). For example, in Arabidopsis HSP81-3, a 47 nucleotide (nt) pyrimidine rich element stimulates cap-independent translation after heat shock (Matsuura et al., 2008). There is evidence that the translation initiation machinery is remodeled during heat shock with the help of brassinosteroids (Dhaubhadel et al., 2002).

In response to the stress, certain functional classes of mRNAs change their translation state in a coordinated fashion. For example, mRNAs for ribosomal proteins are inhibited in response to drought and hypoxia (Kawaguchi et al., 2004; Branco-Price et al., 2008). Ribosomal protein mRNAs have also shown coordinated behavior after mutation of a translation factor, albeit in the opposite direction (Kim et al., 2007). Therefore, the ribosomal protein mRNAs are strong candidates for a translational “regulon”, defined as a cohort of mRNAs that are regulated translationally in a coordinated fashion.

Another possible translational regulon comprises mRNAs for photosynthetic proteins destined for the chloroplast. A sudden shift of plants from light to darkness in the middle of the day shifts many of these mRNAs into the untranslated mRNA pool. First detected in tobacco and pea, the ferredoxin-1 (Fed-1) mRNA dissociates from the polysomal ribosome fraction within 20 min of shifting plants from light to darkness, followed by mRNA destabilization (Petracek et al., 1997; Dickey et al., 1998; Petracek et al., 1998; Petracek et al., 2000; Hansen et al., 2001; Bhat et al., 2004). Transcriptome-wide analysis in Arabidopsis showed that the response reaches far beyond just photosynthetic proteins (Juntawong and Bailey-Serres, 2012; Liu et al., 2012).

We can now begin to ask whether the response profiles to the different stresses are essentially the same or different. Such analyses are not trivial, in part due to the inherent noise and uncontrolled variables in translational array data from different laboratories. However, the response of Arabidopsis to darkness overlaps substantially with the response to hypoxia (Juntawong and Bailey-Serres, 2012), when GC-rich 5' UTRs were specifically affected. Such a result might indicate shared signaling mechanisms, yet this remains a target for future studies.

Somewhat less dramatic shifts in translation state have been detected in response to virus infection (Moeller et al., 2012) and also under normal growth conditions, for example over the course of the diurnal cycle (Piques et al., 2009). The hallmarks of sugar deprivation are a general decrease in sugar metabolism, nitrogen assimilation, protein synthesis, and cell division (Yu, 1999). Accordingly, sugar deprivation was shown to cause translational repression, primarily affecting genes involved in cell proliferation (Nicolai et al., 2006).

Taken together, environmental conditions as well as nutrient status cause extensive differential translation of specific mRNAs. As such, translation in plants is sensitive to some of the same stresses as in metazoans and fungi. Strikingly, however, certain translational responses that are well established in other eukaryotes seem to be absent in plants. Following is one example. Under stress conditions that cause misfolding of proteins in the endoplasmic reticulum (ER), yeast and mammals phosphorylate elF2α, using a dedicated kinase (PERK), in order to globally reduce translation as part of the unfolded protein response. In contrast, the Arabidopsis unfolded protein response is not accompanied by elF2α phosphorylation (Kamauchi et al., 2005). In fact, no translational block has been detected. Moreover, the Arabidopsis genome lacks a recognizable PERK homolog, and the only known elF2α kinase in Arabidopsis, GCN2 (general control non-derepressible 2), does not seem to become activated (Urade, 2007).

What remains to be elucidated are the mechanisms that lead to the selective translation of mRNAs in plant cells. For the response to darkness, photosynthetic electron transport is involved (Petracek et al., 1997). In yeast and metazoans, a major classical signaling pathway for stress responsive translational repression operates via the conserved protein kinase GCN2. Upon phosphorylation, the GCN2 target protein, elF2α is inhibited because it remains bound to its guanine nucleotide exchange factor, elF2B. Because elF2α is the essential cofactor for loading a tRNA onto the 40S ribosome, translation initiation is broadly inhibited. There is evidence for this pathway in plants. First, upon amino acid starvation, which occurs when the synthesis of branched chain or other amino acids in the plastid is disrupted with a herbicide, the Arabidopsis GCN2 kinase is activated and phosphorylates elF2α (Zhang et al., 2003; Lageix et al., 2008; Zhang et al., 2008; Li et al., 2013a). This is accompanied by a widespread drop in polysome loading of mRNAs. elF2α phosphorylation and translational regulation also occurs in response to challenge with a bacterial pathogen (Pajerowska-Mukhtar et al., 2012). In summary, it appears possible that GCN2 plays a broader role in Arabidopsis than in metazoans in response to stresses apart from amino acid starvation, although the mechanism for how these stresses activate GCN2 remains unclear. Quite possibly, different stresses and different protein kinases communicate information from the environment onto the translation machinery through a network of interactions rather than separate pathways (Hey et al., 2010).

3.3. Translational Control in Development

Compared to animals, where translational control is known to operate in various cases of development and cell differentiation, for example, egg cells, embryos, and neurons, translational control of plant development is largely unexplored territory. During pollen development, for example, mRNAs that were synthesized prior to the dormancy of the pollen grain may be stored in an inactive, untranslated, form, only to be activated in response to pollen germination (Hulzink et al., 2002; Honys et al., 2009; Wang and Okamoto, 2009). Moreover, several translation factors and ribosomal proteins have pollen-specific isoforms, which may potentially contribute to translational regulation in the pollen. Likewise, in the male gametophyte of the water fern, Marsilea, mRNAs are translationally repressed due to intron retention. mRNAs become translationally activated as introns are removed, a process that also requires exon junction complex components (Hart and Wolniak, 1999; van der Weele et al., 2007; Boothby et al., 2013).

Transcriptome-wide datasets that contain evidence for celltype specific translation of mRNAs have been collected by ribosome immunoprecipitation. In detail, translational inhibition in response to hypoxia is cell type specific for certain clusters of transcripts (Mustroph et al., 2009). Many additional cases of developmental control of translation may remain hidden in the form of differences in mRNA abundance between cell specific transcriptome and cell specific translatome data (Birnbaum et al., 2003; Mustroph et al., 2009; Jiao and Meyerowitz, 2010).

Are there any specific developmental mechanisms dedicated to the translational regulation of mRNAs? If so, classical forward genetic analysis has not revealed them. Instead, complex information on this topic has surfaced from mutants in the basal translation apparatus, i.e. the ribosome and certain initiation factors. Many but not all mutations in ribosomal proteins tend to cause a defect in expansion growth of the leaf lamina leading to characteristic pointed cotyledons and first leaves (Horiguchi et al., 2012). This defect can probably be attributed to a reduction in bulk translation efficiency and should not be considered evidence for translational regulation of development because it is also observed in mutants of nucleolar ribosome biogenesis proteins (Kojima et al., 2007). On the other hand, some ribosomal protein mutations more specifically affect leaf polarity (Byrne, 2009; Horiguchi et al., 2012) judged by their enhancing the rather mild leaf polarity defects of mutations in the transcription factors AS1 or AS2 (Pinon et al., 2008; Yao et al., 2008), for example mutations in ribosomal proteins RPL10a, RPL5, RPL28, and RPL9 (Pinon et al., 2008; Yao et al., 2008). A related situation has been observed for RPS6Aand B, where double heterozygotes show leaf polarity defects (Creff et al., 2010). It should be noted that not all rpl and rps mutations have these same effects. For example, mutations in RPS10B and RPL27aC appear to be more directed at meristem defects although leaf polarity defects are also evident (Szakonyi and Byrne, 2011; Stirnberg et al., 2012).

Horiguchi and coworkers recently proposed an insightful and detailed framework to interpret the growth and developmental phenotypes of ribosomal protein mutations, outlining three molecular models. Such mutations may reveal the effect of ribosome insufficiency (ribosome function is generally too low), ribosome heterogeneity (an isoform of the ribosome that carries out a specialized function is compromised), or ribosome aberrancy (the mutant ribosome is defective in a specific way (Horiguchi et al., 2012)). Ribosome insufficiency is most likely the reason for the embryo lethality and the pointed leaf phenotypes that are characteristic of severe ribosomal protein mutations. Ribosome heterogeneity at the structural level is a fact of life in the plant cell, given that all ribosomal proteins exist in multiple paralogous isoforms. However, whether the two paralogs confer different functions on the ribosome is generally unclear (Horiguchi et al., 2012). When mutations of two different paralogs display different phenotypes, this could easily be due to different expression characteristics of the two isoforms, rather than different biochemical functions of the isoforms as postulated by the ribosome heterogeneity concept. Ribosome aberrancy can be viewed as a more nuanced form of ribosome insufficiency. Consider that, during translation, the ribosome performs a plethora of individual steps on thousands of mRNAs in order to optimally support cell function. It is self-evident that two given ribosomal mutations will often cause two different biochemical problems, for example a problem during initiation versus elongation. One can postulate that the translation of mRNAs that differ in their nucleotide sequence will be compromised in different ways by the two mutations. Furthermore, given that transcriptomes are cell type specific, the translation defects of the two aberrant mutants may also be cell type specific, and therefore developmentally distinct, as well.

Figure 3.

Localization of ribosomal proteins on the solvent side of the wheat 40S (A) and 60S (B) ribosomal subunits. The images showing ribosomal proteins were created using PyMol (www.pymol.org), and are based on Protein Data Bank codes 31Z6 and 31ZR (Armache et al., 2010a and b). Putative positions of RPS10 and RPS6 have been highlighted based on available yeast and mammalian ribosome structures (Klinge et al., 2012).

In light of the detailed crystal structures of the ribosome, it is reasonable to ask whether the developmentally relevant ribosomal proteins (Horiguchi et al., 2012) colocalize in the ribosome, possibly near the mRNA channel in the top half of the ribosome. However, this is not the case (Klinge et al., 2012). The relevant proteins are distributed over much of the ribosome surface, including the front edge, as defined by the direction of movement along the mRNA (from top to bottom, RPS18, RPS10, RPL9, RPL24, RPS6) and the rear edge (RPL27a, RPL23a, RPS13, RPL38), without any clear pattern (Figure 3). A few mutated proteins reside away from the subunit interfaces on the solvent side (RPL5, RACK1, RPL4, RPL28, RPS15a). Only one mutated protein is clearly located on the subunit interface, RPL10 (Klinge et al., 2012). In summary, the reason why certain ribosomal mutations cause seemingly specific developmental defects remains to be resolved.

3.4. Translational Control by Metabolites

Translational control of gene expression occurs primarily at the level of initiation, but can also take place later, during elongation or termination. An elegant gene-specific case of exceptional elongation control has been worked out through a combination of in vivo and in vitro analyses for the cystathione gamma synthetase mRNA(CGS1), the first committed step in methionine biosynthesis. In the case of CGS1, a 14 amino acid region in the center of the polypeptide renders translation elongation sensitive to the end product of the pathway, S-adenosyl methionine (SAM). An exciting conclusion that emerged from these data is that SAM itself appears to interact with the nascent CGS1 polypeptide in the ribosome exit tunnel. Autoregulated arrest of translation elongation then leads to mRNA degradation at sites defined by stacking of stalled ribosomes (Murota et al., 2011; Onoue et al., 2011).

The remainder of this section focuses on metabolic control of translation initiation. All of the five examples involve uORFs, which often overlap with each other. As we will see, such patterns of overlapping uORFs are emerging as a paradigm for regulatory logic gates that can make gene expression sensitive to small molecules. As the first example, translation efficiency of the phosphatidyl-choline biosynthetic enzyme, phosphoethanolamine N-methyltransferase is dramatically reduced by phosphocholine via a conserved-peptide uORF (CP-uORF) in its 5′ UTR (Tabuchi et al., 2006; AlatorreCobos et al., 2012). A second case exemplifies how uORF-mediated regulation can be linked to nonsense-mediated decay. The Arabidopsis AtMHX1 gene encodes a vacuolar magnesium/zinc - proton antiporter. Translation of its uORF triggers mRNA turnover via nonsense-mediated decay (David-Assael et al., 2005; Nyiko et al., 2009; Saul et al., 2009). Third, many of the mRNAs that are responsible for polyamine synthesis or function are subject to translational regulation by uORFs. Polyamines are small organic polycations that bind RNA, affect start codon recognition, and serve as substrate for certain post-translational modifications in the cell (Ivanov et al., 2010). S-adenosylmethionine decarboxylase (AdoMetDC) catalyzes a commitment step in the pathway leading to the end product, polyamine. Importantly, polyamines cause end product repression of AdoMetDC at the translation level (Hanfrey et al., 2002; Hanfrey et al., 2003; Hanfrey et al., 2005), arguably the best understood model system for metabolic control of translation. The AdoMetDC mRNA harbors two conserved overlapping uORFs in its 5′ leader, a highly conserved ‘short’ uORF of 48–55 codons and a tiny uORF of 2 to 3 codons that overlaps the AUG of the short uORF. This pair of uORFs forms the basis of a negative feedback model for AdoMetDC translation. Under conditions of low polyamine concentration, the scanning ribosome recognizes the AUG of the tiny uORF, translates past the AUG of the short uORF, and, after translation termination, reinitiates efficiently at the downstream AdoMetDC ORF, thereby increasing AdoMetDC activity and raising the polyamine level in the cell. When polyamine levels are sufficiently high, the scanning ribosome ignores the AUG of the tiny uORF, which is in a weak context, and therefore recognizes the stronger AUG of the short uORF. Because the short uORF does not allow reinitiation, the downstream AdoMetDC ORF is not translated (Hanfrey et al., 2005).

Another case study of uORF-mediated regulation, likewise related to polyamine regulation, was discovered around a conserved-peptide CP-uORF in the bHLH transcription factor SAC51, which supports stem elongation in Arabidopsis. In this case, polyamine is needed to suppress the uORF and allow SAC51 translation. Under a physiological polyamine level, this uORF is suppressed, allowing expression of SAC51. This case was dissected by classical forward genetic analysis. For example, a premature stop codon in the SAC51 uORF was discovered as a suppressor mutation of the acaulis5 polyamine biosynthesis mutant (lmai et al., 2006). The shortening of the uORF is expected to make it less inhibitory for SAC51 expression. The second suppressor of acauMs5 is a mutation in ribosomal protein RPL10A; the mechanism of its action is less clear (lmai et al., 2008).

Finally, uORF-mediated sucrose sensing has come forth as another translational regulatory mechanism for metabolite signaling (Hanson et al., 2008; Hummel et al., 2009; Thalor et al., 2012). The Arabidopsis AtbZip11 mRNA harbors four uORFs in its 5′ leader (Rook et al., 1998b; Rook et al., 1998a; Wese et al., 2004). The translation of AtbZip11 is repressed in response to sucrose by virtue of the peptide encoded by uORF2 (Wiese et al., 2004; Rahmani et al., 2009). AtbZip11 regulates amino acid and sugar metabolism, indicating that translational control by the transport form of sugar, sucrose, remodels carbon and nitrogen pathways (Hanson et al., 2008).

Taken together, a variety of small molecule metabolites can regulate the translation of specific mRNAs, and in most cases a uORF or uORF cluster is required. With the exception of CGS1, the biochemical mechanism of metabolite recognition, which may be indirect or direct, remains to be elucidated.

3.5. Sequence Elements that Drive Translational Control

The mRNA sequence elements that govern translational regulation of plant mRNAs are generally not well defined. Why is this so? On those occasions when sequence elements are known to function at the post-transcriptional level, it is often not clear whether the element affects translation, RNA stability, or RNA localization. Elements known to drive RNA localization (zipcodes) affect translation indirectly. Many translational control elements have been inferred on theoretical grounds, because they contain non-canonical start codons (Kobayashi et al., 2001; Wamboldt et al., 2009), a short coding region (uORF), or because they are conserved (Hayden and Jorgensen, 2007; Tran et al., 2008; Jorgensen and DorantesAcosta, 2012; Vaughn et al., 2012). In contrast, rather few plant translational control elements have been identified de novo by experimental structure-function analysis. RNA structures that regulate translation are very well appreciated in viruses (Zuo et al., 2010; Nicholson and White, 2011; Stupina et al., 2011), but barely characterized for plant mRNAs (Niepel et al., 1999; Wang and Wessler, 2001; Hulzink et al., 2002; Li et al., 2012).

In the following, we summarize pertinent concepts of translational control by uORFs, alternative initiation codons, motifs for cap-independent initiation, and miRNAs, emphasizing recent findings (Kawaguchi and Bailey-Serres, 2005) (Figure 4).

3.5.1 uORFs

uORFs are prevalent in 5′ UTRs of mRNAs for transcription factors and kinases, among others, suggesting they play an important role in regulating gene expression, growth and development. For example, several auxin response transcription factors (ARFs) harbor uORFs in their 5′ UTR, which render the translation dependent on the translation reinitiation machinery (Nishimura et al., 2005; Zhou et al., 2010; Schepetilnikov et al., 2013). uORFs fall into three classes; (i) the uORF encoded peptide may act in trans; (ii) it may have a conserved peptide and act in cis; (iii) the uORF may have a non-conserved, presumably inert, coding region that simply acts as a translational barrier before the main ORF.

(i) Trans-acting uORF peptides are those that can affect expression of an RNA molecule different from the one that gave rise to it. They are rare. The best example comes from the Medicago truncatula MtHAP2 mRNA, which encodes a transcription factor of the CCAAT enhancer binding protein class. A uORF peptide encoded by the MtHAP2 mRNA binds to the 5′ leader, which causes the subsequent degradation of its transcript in the meristematic zone of the Medicago nodule (Combier et al., 2006; Combier et al., 2008). It is debatable whether this case should even be regarded as an example for a trans-acting uORF peptide, since the peptide regulates its own mRNA.

(ii) Cis-acting conserved peptide uORFs (CP-uORFs) are uncommon. The uORFs in this class, some of which were discussed in detail in the section on metabolic regulation, are typically conserved at the peptide level across the dicot and monocot plant lineages (Hayden and Jorgensen, 2007; Tran et al., 2008; Jorgensen and Dorantes-Acosta, 2012; Vaughn et al., 2012). Most conserved peptides that have been examined strongly inhibit translation. By inference from prior studies on CP-uORF peptides in animals and fungi, these peptides are hypothesized to stall the ribosome, thus blocking the progression of upstream ribosomes or suppressing reinitiation. The number of CP-uORFs is estimated at only a few dozen per genome.

Figure 4.

Schematic of post-transcriptionally active sequence elements in a generic mRNA. Not all elements will be present in every mRNA. (1) The 5′ end of the mRNA consists of the 7-methyl-guanosine cap bound by the elF4 complex. (2–5) The 5′ leader may possess an RNAstem loop and stem loop binding protein (2), an upstream open reading frame (uORF) (3), or a translation enhancer motif (4), or a miRNA binding site (5). (6–8) The main open reading frame may commence with a noncanonical start codon (6) or a canonical AUG start codon (7) and may bear an exon junction complex (8). The open reading frame depicted here terminates with UGA as a stop codon. (9–12) The 3′ untranslated region (3′UTR) may harbor a miRNA target site (9), a zipcode element (10) for mRNA localization, or a structural element bound by RNA binding protein (11). The poly(A) tail is bound with poly(A) binding protein (12). Ribosomal subunits are indicated by tan ovals.

(iii) uORFs with non-conserved peptides represent the vast majority of uORFs. About 35% of Arabidopsis genes give rise to a uORF-containing mRNA, and about half of these have multiple uORFs (Kim et al., 2007). This fraction is similar in all other plant genomes examined (Vaughn, Jia, and von Arnim, unpublished). Thus, the Arabidopsis genome encodes an estimated 23,463 uORFs, a number in the same league as the number of main ORFs. It is difficult to assess whether the position or sequence of these uORFs is actively conserved. Over long evolutionary timescales, the peptides appear unconstrained by selection as judged by synonymous to non-synonymous codon bias, but short timescales have yet to be examined. However, the presence of non-CP-uORFs is often conserved (Vaughn et al., 2012).

Upstream ORFs must be translated regularly, since AUG triplets in any context can be recognized by the scanning ribosome most of the time (Lukaszewicz et al., 1998; Sugio et al., 2010), and recent ribosome footprinting studies in yeast and metazoan mRNAs have proven that existing uORFs regularly engage ribosomes, sometimes even initiating on non-AUG triplets (lngolia et al., 2009; lngolia et al., 2011). Non-AUG initiation is probably fairly common in plants as well (Gordon et al., 1992; Kobayashi et al., 2002). Where examined, uORFs suppress translation in a length dependent manner (Wese et al., 2004; Nyiko et al., 2009; Roy et al., 2010). In detail, uORFs of more than 25 codons strongly inhibit translation while uORFs shorter than 5 often cause no or barely noticeable inhibition.

Generally speaking, it is rare for mutations in uORFs to emerge from genetic screens (lmai et al., 2006). This is most likely because the uORF presents a small target, and premature stop codons are usually needed for a dramatic effect, and the resulting gain-of function phenotypes may be mild. Vice versa, mutations that generate uORFs must create an AUG in the 5′ UTR, a rare event. Most such cases will generate uORFs that are fairly short and therefore not highly inhibitory.

3.6. Translational Control by RNA Binding Proteins, RNA Localization, and via Nonsense-Mediated Decay

3.7. Signaling Pathways Implicated in Translational Control-TOR, S6K, GCN2, RACK1

In light of our sparse mechanistic knowledge about translational control in plants it is appropriate to take a leaf from other eukaryotes and assess whether the hypotheses founded on metazoan and yeast research are valid in plants. One major pathway that controls the translation machinery in metazoans is stimulated by extracellular signals such as insulin, and proceeds via a kinase cascade, including PI-3-Kinase, Akt kinase, TOR kinase, and S6 kinase (S6K), eventually phosphorylating a variety of components of the translation initiation machinery (Laplante and Sabatini, 2012). Key elements of this pathway, namely TOR and S6K, exist in Arabidopsis, and the pathway must play a critical role in the regulation of plant growth and development. As seen below, however, the upstream signals that stimulate this pathway in plants and the precise downstream targets at the translation level are only beginning to come into focus.

While several components of the translation machinery are phosphorylated, among others, poly(A) binding protein and elF4E (Le et al., 2000; Pierrat et al., 2007), the responsible kinases are known in only a few cases. Casein kinase 2 (CK2) is one such exception. In contrast to other pathways in this section, phosphorylation by CK2, specifically of elF5, has the potential to regulate translation directly, namely by regulating interaction between components of the multifactor complex (Dennis and Browning, 2009). This final section of the review will elaborate on TOR/S6K, briefly revisit GCN2 kinase, and conclude with a summary of another emerging translational regulator, the RACK1 protein.