Translation elongation factor P (EF-P)

ABSTRACT

Translation elongation factor P (EF-P) is conserved in all three domains of life (called eIF5A and aIF5A in eukaryotes and archaea, respectively) and functions to alleviate ribosome pausing during the translation of specific sequences, including consecutive proline residues. EF-P was identified in 1975 as a factor that stimulated the peptidyltransferase reaction in vitro but its involvement in the translation of tandem proline residues was not uncovered until 2013. Throughout the four decades of EF-P research, perceptions of EF-P function have changed dramatically. In particular, while EF-P was thought to potentiate the formation of the first peptide bond in a protein, it is now broadly accepted to act throughout translation elongation. Further, EF-P was initially reported to be essential, but recent work has shown that the requirement of EF-P for growth is conditional. Finally, it is thought that post-translational modification of EF-P is strictly required for its function but recent studies suggest that EF-P modification may play a more nuanced role in EF-P activity. Here, we review the history of EF-P research, with an emphasis on its initial isolation and characterization as well as the discoveries that altered our perceptions of its function.

Here the authors review how models of elongation factor P have changed over time.

INTRODUCTION

Translation, or the process of decoding mRNA transcripts into amino acid primary protein sequences, is central to all life. Peptide bond formation between amino acids is catalyzed by a highly conserved molecular machine called the ribosome (Roberts 1958). Ribosomes are processive enzymes capable of synthesizing long stretches of polypeptides at a high catalytic rate ranging from 6 (in the case of eukaryotes) to 20 (in the case of bacteria) peptide bonds per second (Pederson 1984; Boström et al. 1986; Liang et al. 2000; Ignolia, Lareau and Weissman 2011). The rate of translation elongation depends, at least in part, on the template and can be slowed by the interaction of the ribosome with particular mRNA sequences, the decoding of rare codons, and the translation of certain consecutive amino acids (Tanner et al. 2009; Ignolia, Lareau and Weissman 2011; Li, Oh and Weissman 2012; Charneski and Hurst 2013; Chevance, Le Guyon and Hughes 2014; Dana and Tuller 2014; Gardin et al. 2014). Rather than being an unfortunate and unavoidable feature of the ribosome, suboptimal translation efficiency has been implicated in the regulation of protein copy number, folding, and secretion (Ude et al. 2013; Fluman et al. 2014; Qi et al. 2018).

The majority of the ribosome is made up of rRNA, which comprises approximately 60% of the structure and is directly responsible for the catalysis of peptide bond formation (Tissiéres et al. 1959; Nissen et al. 2000). Ribosomes also contain ∼50 proteins that serve to stabilize the core rRNA structure as well as assimilate signals from other cellular pathways. In addition to the canonical ribosome structural proteins, elongation factors that transiently interact with the ribosome during translation elongation are also ubiquitous. For example, translation elongation factors EF-Tu and EF-G are responsible for delivering amino-acylated tRNAs to the ribosome and ensuring proper ribosome translocation following peptide bond formation, respectively (Reviewed in Czworkowski and Moore 1996). While association is transient, EF-Tu and EF-G nonetheless interact with the ribosome and participate in each cycle of peptide bond formation. By contrast, another elongation factor, EF-P, only participates in a subset of peptide bond formation events.

EF-P is conserved in all domains of life, where it is called eIF5A in eukaryotes and aIF5A in archaea and was recently shown to alleviate ribosome pausing during the translation of consecutive proline residues (Fig. 1) (Kyrpides and Woese 1998; Doerfel et al. 2013; Ude et al. 2013). Structural analysis indicates that consecutive proline residues impose an unfavorable peptidyl-tRNA geometry that is counteracted by EF-P to promote translation elongation (Huter et al. 2017). EF-P is post-translationally modified in all organisms studied to date and it is thought that post-translational modification is important to alleviate ribosome pausing (Yanagisawa et al. 2010; Doerfel et al. 2013; Lassak et al. 2015; Yanagisawa et al2016). While it is now widely accepted that EF-P acts throughout translation elongation, understanding of its function was misled by the belief that it was primarily involved in the catalysis of the peptide bond between the first two amino acids and was thought to be absolutely essential for growth. While EF-P is essential in many organisms, recent work has shown that the requirement of EF-P for growth in Escherichia coli is conditional and even dispensable in Bacillus subtilis, suggesting a more complex role for EF-P in cellular viability (Rajkovic et al. 2016; Tollerson, Witzky and Ibba 2018 ). Throughout the history of EF-P research, views of its function have quickly arisen and are later called into question. Here, we review the history of EF-P research with a focus on how our understanding of its function in the cell has changed over time. While the eukaryotic and archaeal orthologs eIF5A and aIF5A are not the focus of this review (see Park and Wolff 2018 and Turpaev 2018 for recent reviews), we will also highlight aspects of their history when they influenced our understanding of EF-P function in bacteria.

Figure 1.

EF-P alleviates ribosome pausing at consecutive proline residues. Graphic depicting the current model of the function of EF-P in translation. Ribosomes pause during the translation of particular amino acid sequences such as consecutive proline residues, and pausing results in an empty E-site that allows the entry of EF-P. Upon association with a paused ribosome, EF-P interacts with the P-site tRNA and alleviates ribosome pausing.

EF-P DISCOVERY

Early experiments aimed to interrogate the mechanism of translation utilized relatively few model in vitro systems. In one, the ability of factors to promote the synthesis of poly-phenylalanine peptides was assayed. In these experiments, it was shown that in addition to the ribosome, aminoacylated-tRNAs and an mRNA template, several soluble factors (EF-G, EF-Tu and EF-Ts) were required (Lucas-Lenard and Lipmann 1966). EF-G, EF-Tu and EF-Ts were not required, however, for another model translation system: the formation of a peptide bond between fMet-tRNA^fMet and puromycin (Traut and Monro 1964; Madden, Traut and Monro 1968). Puromycin is an antibiotic that chemically resembles the 3’ end of aminoacylated tRNAs and, upon serving as a substrate in the peptidyl-transferase reaction, it stimulates premature termination of translation (Yarmolinsky and de la Haba 1959; Morris and Schweet 1961). As such, delivery of tRNA's by EF-Tu/EF-Ts and ribosome translocation directed by EF-G are not required for the fMet-tRNA^fMet-puromycin reaction. The puromycin model was instrumental in determining the minimal requirements for the peptidyl-transferase reaction which were shown to be housed in the 50S subunit of the ribosome (Traut and Monro 1964; Madden, Traut and Monro 1968).

While puromycin was critical to early studies of translation, it is a relatively poor peptidyltransferase acceptor, resulting in slow peptide bond formation particularly when puromycin concentrations are low (Glick and Ganoza 1975). Ribosome-mediated peptide bond formation between fMet-tRNA^fMet and puromycin, however, could be improved by adding fractions of whole cell lysates and sub-fractionation identified a soluble factor that accelerated the reaction (Glick and Ganoza 1975). Moreover, the stimulatory factor was not required for 70S ribosome assembly or for binding of fMet-tRNA^fMet and was inhibited by antibiotics that block the peptidyl-transferase center (Glick and Ganoza 1975). Thus, it was proposed that this factor increased the ribosome's inherent ability to catalyze peptide bond formation and, due to its ability to stimulate the peptidyl-transferase reaction, it was named elongation factor P (EF-P) (Glick and Ganoza 1975).

Reaction between fMet-tRNA^fMet and puromycin, however, was insufficient to determine whether or not EF-P stimulated the peptidyl-transferase reaction throughout translation elongation or whether the stimulation was restricted to the first peptide bond. As such, the effect of EF-P on the translation of poly-phenylalanine and poly-lysine was assayed in a subsequent report (Glick and Ganoza 1976). It was observed that addition of EF-P increased poly-lysine synthesis ∼10-fold, consistent with it acting throughout translation elongation (Glick and Ganoza 1976). While EF-P also stimulated poly-phenylalanine production, the degree to which it did so (∼2.5-fold) was lower than that observed for poly-lysine, indicating that the stimulatory effect was perhaps sequence specific (Glick and Ganoza 1976).

To explore the apparent sequence specificity of EF-P, its ability to promote peptide bond formation between fMet-tRNA^fMet and diverse aminoacyl acceptors was tested (Glick, Chládek and Ganoza 1979). Phenylalanine is a high affinity substrate in the peptidyl-transferase reaction and addition of EF-P had little effect on its reaction with fMet-tRNA^fMet (Fig. 2) (Glick, Chládek and Ganoza 1979). Incorporation of lysine, however, proceeded at a slower rate and increased 2.5-fold upon the addition of EF-P (Fig. 2) (Glick, Chládek and Ganoza 1979). The stimulatory effect of EF-P on peptide bond formation was even greater for aminoacyl accepters with less reactivity than lysine (namely leucine, puromycin and glycine) (Fig. 2) (Glick, Chládek and Ganoza 1979). Thus, EF-P stimulated peptide bond formation in a manner that depended on the efficiency of the acceptor and functioned to accelerate polymerization of amino acids that would otherwise be incorporated slowly into proteins (Glick, Chládek and Ganoza 1979).

Figure 2.

EF-P increases the reactivity of poor aminoacyl substrates. The substrate concentration at half of the maximal acceptor activity in the peptidyl-transferase reaction with the donor fMet-tRNA^fMet for a variety of substrates was determined both in the presence and absence of EF-P. Data are a graphical representation of results reported in Table 1 of Glick, Chládek and Ganoza 1979 and are adapted with permission.

These remarkable early studies form the critical and definitive biochemical foundation for understanding the mechanism of EF-P. Our current understanding that EF-P alleviates ribosome pausing primarily at certain consecutive proline residues is well in-line with these early studies, as proline is a particularly poor acceptor in the peptidyl-transferase reaction (Pavlov et al.2009; Johansson et al. 2011). Despite the early successes, however, it was not until 2013 (over 30 years after the initial reports) that the effect of EF-P on poly-proline translation was reported (Doerfel et al. 2013; Ude et al. 2013). There are likely two reasons that over 30 years passed between the identification of EF-P and the discovery of its biological function. First, EF-P was reported to be essential and therefore seemed inconsistent with promoting a reaction at a highly specific sequence in a subset of proteins. Second, the perception of EF-P function erroneously shifted from acting throughout translation elongation to acting specifically during the synthesis of the first peptide bond.

EF-P IS A NON-CANONICAL TRANSLATION ELONGATION FACTOR

It is unclear exactly how or why EF-P began to be viewed as acting primarily during the synthesis of the first peptide bond, but it may be related to the recognition of its homology to the eukaryotic translation factor eIF5A (originally called M2Bα or eIF4D). eIF5A was initially isolated as one of a group of three factors named M₂ that stimulated poly-phenylalanine synthesis at low Mg²⁺ concentration (Shafritz and Anderson 1970; Shafritz et al. 1970). The M₂ group was later shown to be required for translation initiation of hemoglobin, and all three proteins that it contained were therefore classified as initiation factors (Prichard et al. 1970). Subsequent separation of the M₂ group identified its constituents as the bona fide initiation factors eIF5 and eIF1A, as well as eIF5A (Kemper, Berry and Merrick 1976; Anderson et al. 1977). Unlike the other two M₂ group proteins, however, purified eIF5A was not required for the assembly of initiation complexes in vitro and had no effect of globin biosynthesis, arguing that it was not required for initiation (Kemper, Berry and Merrick 1976; Benne and Hershey 1978; Benne, Brown-Luedi and Hershey 1978). One report even speculated that eIF5A acted during translation elongation under the low magnesium conditions in which it was isolated (Schreier, Erni and Staehelin 1977).

The understanding of eIF5A function took a critical turn when another group reported that eIF5A had no effect on initiation complex assembly, elongation, or indeed production of poly-phenylalanine whatsoever, contrary to the previous reports (Benne and Hershey 1978). The inability to detect eIF5A activity during elongation was perhaps due to differences in experimental conditions, especially considering that the eIF5A-dependent stimulation of poly-phenylalanine required low Mg²⁺, but the conditions yielding the negative result were not described and the results were reported anecdotally (Shafritz and Anderson 1970; Shafritz et al. 1970; Benne and Hershey 1978). eIF5A was, however, shown to stimulate peptidyltransferase activity between the initiating Met-tRNA and puromycin and it was proposed that eIF5A acted specifically during the synthesis of the first peptide bond (Benne and Hershey 1978). The model in which the activity of eIF5A was restricted to the synthesis of the first peptide bond persisted in the literature and began to be introduced as fact in the 1990’s (Schnier et al.1991).

The functional similarity between EF-P and eIF5A in the stimulation of the fMet-puromycin reaction was recognized (Aoki et al. 1991) and the two proteins were shown to share weak but significant sequence homology (Kyrpides and Woese 1998). Moreover, the two proteins were similar in three-dimensional structure (Fig. 3A,B). Whereas EF-P was found to have three domains which together strongly resembled a tRNA in shape, size and charge (Fig. 3A,C) (Hanawa-Suetsugu et al. 2004), the structure of eIF5A was similar to two of the EF-P domains but lacked the region that corresponds to the tRNA anticodon (Fig. 3B,C). Thus, similarities in activity, sequence and relative structure suggested that the two proteins were homologous. While it was accepted that both proteins were related to translation, eIF5A was considered an initiation factor and the role of EF-P in protein synthesis was unclear.

Figure 3.

The three dimensional structures of EF-P, eIF5A, and tRNA are similar. A) Three dimensional structure of Thermus thermophilus EF-P (PDB ID 1UEB). Domains I, II and III are labeled accordingly. B) Three dimensional structure of S. cerevisiae eIF5A (PDB ID 3ER0). C) Three dimensional structure of S. cerevisiae tRNA^Phe (PDB ID 1EVV). D) The location of EF-P (blue) bound to a ribosome (grey) stalled at a polyproline sequence solved by cryo-electron microscopy (PDB ID 6ENJ). The positions of the 50S subunit, 30S subunit, P-site tRNA (salmon) and A-site tRNA (green) are labeled accordingly.

In an attempt to reconcile the different functions of the eIF5A and EF-P homologs, the function of EF-P was re-investigated in E. coli by analyzing its presence in polysomes. Polysomes are transcripts with more than one ribosome attached and translation elongation factors were hypothesized to be differentiated from initiation factors based on their relative abundance in polysomes. Canonical elongation factors bind to all elongating ribosomes, and as such their concentration in the polysome fractions increases proportionally with ribosome number. Canonical initiation factors bind to ribosomes only once, during initiation, and thus their concentration depends more on the number of initiation sties on transcripts than the number of ribosomes bound. Thus, polysomes were fractionated from E. coli cells and the ratio of EF-P to ribosomes was measured by western blot analysis. While EF-P indeed bound to polysomes, the amount of binding decreased with increasing ribosome content of the fraction (Aoki et al. 2008). This inverse relationship suggested that EF-P bound to a subset of ribosomes on the transcript, a result seemingly more consistent with the model that it acted only during the synthesis of the first peptide bond. Given our current understanding of EF-P activity, the results described are consistent with EF-P being a sequence specific elongation factor that only binds to ribosomes at particular sites, such as those paused at consecutive proline residues. Nonetheless, at the time, elongation factors were thought to be sequence independent and it was argued that EF-P was involved in the transition from initiation to elongation.

EF-P AND eIF5A ACT THROUGHOUT TRANSLATION ELONGATION

The first in vivo evidence that eIF5A acted during translation elongation rather than initiation was provided by several studies that characterized its activity in Saccharomyces cerevisiae (Jao and Chen 2006; Zanelli et al. 2006; Saini et al. 2009). Immunopreciptation of eIF5A indicated that it was in a complex with ribosomes and elongation factors, but not initiation factors, consistent with it having a post-initiation function (Jao and Chen 2006; Zanelli et al. 2006). Moreover polysomes, rather than monosomes, accumulated at the non-permissive temperature in an eIF5A temperature-sensitive mutant, consistent with a defect in translation elongation (Zanelli et al. 2006; Saini et al. 2009). Finally, the role of eIF5A in elongation was solidified by comparing the ribosome transit time in the presence and absence of eIF5A (Saini et al. 2009). The ribosome transit time was determined by radioactively labeling nascent proteins and determining the amount of time elapsed between when radioactivity was detected in total proteins and when it was detected in completed proteins that had been released from the ribosome. Inactivation of eIF5A increased the ribosome transit time to twice that of wild type, providing conclusive evidence that eIF5A promoted translation at a step following initiation (Fig. 4) (Saini et al. 2009). Thus, it was concluded that eIF5A acted as an elongation factor and it was speculated that EF-P likely functioned in similar manner.

Figure 4.

eIF5A promotes a step in translation following initiation. Analysis of [³⁵S]Met incorporation into total peptides (closed symbols) or completed peptides (open symbols) in wild type S. cerevisiae or the eIF5A temperature sensitive mutant (eIF5A^ts) at the non-permissive temperature. Ribosome transit times, a metric that is unaffected by translation initiation, are indicated. Red text and lines added to original panels. Panel adapted with permission from Saini et al. 2009.

Again, if eIF5A/EF-P functioned as elongation factors, they did so in a manner that had not been previously attributed to this class of protein. Specifically, depletion of eIF5A only resulted in a ∼30% decrease in the rate of protein synthesis in vivo which was not as severe as one would expected for canonical elongation factors (Kang and Hershey 1994). Thus, it was concluded that eIF5A likely only operated on a subset of transcripts. Consistent with conditional activity, the absence of EF-P was found to result in a reduction in a subset of proteins in the outer membrane of Salmonella and in a subset of virulence factors of Agrobacterium (Peng et al. 2001; Zou et al. 2012). Combined with the ability of EF-P to stimulate a subset of peptidyl-transferase reactions (Glick, Chládek and Ganoza 1979), it seemed possible that EF-P and eIF5A activity was dependent on some as-yet-unknown sequence, either at the transcript or protein level.

In 2013, an aspect of the sequence-specificity of EF-P was revealed when two groups simultaneously reported that in E. coli, EF-P alleviated ribosome pausing at poly-proline motifs during translation elongation. In one of these seminal studies, a mutant defective for EF-P activity was found to reduce expression of the Cad acid resistance system (Ude et al. 2013). Experiments using translational reporters in vivo showed that EF-P promoted expression when the reporter gene was introduced downstream of a PPP motif in CadC, the sensor protein responsible for activation of the Cad module (Ude et al. 2013). Moreover, reductions in the number of sequential prolines alleviated EF-P dependency (Fig. 5A) (Ude et al. 2013). The other study revisited the early observations that EF-P promoted peptide bond formation between poor amino-acyl acceptors and tested the reactivity of an array of amino acids in the presence and absence of EF-P (Doerfel et al. 2013). EF-P strongly stimulated peptide bond formation within PPP and PPG tripeptides in vitro (Doerfel et al. 2013). Thus, EF-P promoted translation elongation through poly-proline-containing sequences and the observation was confirmed for eIF5A in S. cerevisiae shortly thereafter (Gutierrez et al. 2013).

Figure 5.

EF-P promotes translation of poly-proline stretches. A) in vivo β-galactosidase activity of translational reporters containing an increasing number of consecutive proline residues in the presence (black bars) or absence (tan bars) of EF-P in E. coli. Panel adapted with permission from Ude et al. 2013. B) Schematic of the chemical structure of proline and a pro-pro peptide bond.

EF-P REGULATES A SUBSET OF XPPX MOTIF-CONTAINING PROTEINS

Systems-level analyses supported the sequence specificity of EF-P. In Salmonella enterica and E. coli, proteomic analyses indicated that indeed, many PPP- and PPG-containing proteins display EF-P-dependence (Hersch et al. 2013; Peil et al. 2013). Not all poly-proline containing proteins, however, exhibited decreased abundance in the absence of EF-P, suggesting that the requirement for EF-P was more complex than previously appreciated (Hersch et al. 2013; Peil et al. 2013). Further analysis of the proteomic data indicated that EF-P more generally promoted translation of X-Pro-Pro-X (XPPX) motifs, with the severity of EF-P-dependency determined by the identity of the residues immediately adjacent to the tandem prolines (Hersch et al. 2013; Peil et al. 2013). Moreover, in vitro translation experiments indicated that reduced protein levels appeared to be due to a decreased rate of synthesis caused by ribosome pausing when attempting to translate XPPX motifs (Doerfel et al. 2013; Peil et al. 2013; Ude et al. 2013). Thus, ribosomes appeared to pause at a subset of consecutive prolines in the primary sequences, and EF-P alleviated the pause.

The context-dependent specificity of ribosome pausing was also supported by ribosome profiling as pausing in the absence of EF-P was only observed at approximately half of all XPPX motifs (Elgamal et al. 2014; Woolstenhulme et al. 2015). Moreover, when ribosome profiles and proteomes were compared, it was noted that even when ribosomes paused at a particular at XPPX motifs, the effects on protein production were quite varied, with only some pauses actually resulting in decreased protein expression (Peil et al. 2013; Elgamal et al. 2014; Woolstenhulme et al. 2015). The combined systems-level approaches lead to a model in which the effect of EF-P on protein levels is a function of three factors: the efficiency of translation initiation, the position of the pause within the open reading frame, and the strength of ribosome pausing (Hersch et al. 2014; Woolstenhulme et al. 2015). It was proposed that genes that exhibited a high rate of translation initiation and contained a strong EF-P-alleviated ribosome pause site towards the 5’ end of the gene were more likely to suffer decreased protein production in the absence of EF-P (Hersch et al. 2014; Woolstenhulme et al. 2015; Lassak, Wilson and Jung 2016).

Ribosome pausing at XPPX motifs not only has the capacity to directly control protein copy number, but has also been reported to indirectly regulate several co-translational processes. Computational analysis of the distribution of XPPX motifs in E. coli proteins found an enrichment of XPPX motifs in inter-domain regions of proteins implicating pausing at these sites in the control of co-translational protein folding (Qi et al. 2018). Similarly, XPPX motifs were found to be enriched approximately 50 residues downstream of transmembrane helices, raising the possibility that XPPX-mediated translational pausing aids in the efficient co-translational insertion of transmembrane helices (Qi et al. 2018). Finally, decreased levels of EF-P observed during macrophage infection by S. enterica has been shown to regulate the expression of two operons through transcriptional attenuation. During conditions in which EF-P levels are high, the proline-rich leader peptides MgtL and MgtP are readily translated, promoting the formation of secondary structures that terminate transcription and thereby decrease expression of their downstream operons (Gall et al. 2016; Nam et al. 2016). Conversely, when cells infect macrophages, expression of EF-P decreases, resulting in increased pausing in the leader peptides and the formation of an anti-terminator hairpin that promotes the transcription of downstream genes (Gall et al. 2016; Nam et al. 2016). Thus, the role of translational pausing and the relief of pausing by EF-P may be regulatory and some proteins may have evolved to include XPPX motifs as a means to control co-translational processes and/or their copy number.

EF-P MECHANISM OF ACTION

To understand how EF-P promotes translation through poly-proline sequences, we must first understand why the translation of prolines is problematic. Proline is the only proteinogenic secondary amino acid, meaning that its nitrogen atom is bound to two organic groups, and rotation about the N-C bond is impossible due to cyclization (Fig. 5B). In vitro translation assays utilizing proline analogs indicated that the geometric, rather than electronic, properties of proline make it a particularly poor peptidyl-transferase substrate (Doerfel et al. 2015). Indeed, unlike other aminoacyl-tRNA analogs, structural analysis revealed that a Pro-tRNA analog bound to the ribosomal A-site had an altered conformation that was predicted to prevent the proper alignment of substrates in the ribosomal active site (Melnikov et al. 2016). Further structural analysis indicated that a tRNA with two prolines adopted an unusual conformation in the ribosome exit tunnel and the activated end of the tRNA lost structural resolution suggesting heightened molecular movement (Huter et al. 2017). Thus, translation may stall in the absence of EF-P because the end of the tRNA is structurally destabilized due to the unique steric constraints imposed by consecutive proline residues.

EF-P and eIF5A appear to stabilize the P-site tRNA by binding to a unique location between the ribosomal P- and E-sites (Fig. 3D) (Blaha, Stanley and Steitz 2009; Schmidt et al. 2016; Huter et al. 2017). While parts of EF-P/eIF5A extend toward the peptidyl transferase center, the protein seems too remote to participate in the peptidyl transferase reaction directly (Blaha, Stanley and Steitz 2009; Schmidt et al. 2016). Instead, it appears that EF-P/eIF5A interacts with the adjacent tRNA in the P-site (Fig. 3D) (Blaha, Stanley and Steitz 2009; Schmidt et al. 2016; Huter et al. 2017). Interaction with the P-site tRNA reduces mobility of its activated end and the structural stabilization is thought to promote peptide bond formation (Huter et al. 2017). In support of this model, calculation of free energy parameters between a proline analog and puromycin showed that EF-P reduced the free energy of the reaction by altering the entropy rather than enthalpy component (Doerfel et al. 2015). Thus, EF-P appears to bind near the E-site and restrict dynamic mobility of the P-site tRNA to stabilize a conformation compatible with the next peptidyl transferase reaction. We note that the requirement of EF-P for growth in E. coli is alleviated at lower temperatures, perhaps consistent with a role of EF-P in reducing molecular motion (Tollerson, Witzky and Ibba 2018).

Recently, mutations that partially suppressed the need for EF-P in B. subtilis were reported, one of which perhaps support the idea that EF-P is involved in controlling space within the ribosome active site (Hummels and Kearns 2019). The peptidyl-transferase reaction is a catalyzed by residues in the 23S rRNA and one residue near the active site (guanosine 2551) is methylated for mechanistically unknown reasons (Lövgren and Wikström 2001). A loss of function mutation in YacO, the enzyme predicted to add the methyl group, reduced the requirement for EF-P (Hummels and Kearns 2019). Thus, methylation of guanosine 2551 may exacerbate constraints imposed on the poly-prolyl-tRNA and its absence may better accommodate the next reaction. Other suppressor mutations that were identified are harder to explain. For example, change-of-function mutations were identified in two ribosomal small subunit proteins, far from where EF-P is thought to either bind or act (Hummels and Kearns 2019). Still other suppressors indicated a role for proteins of poorly understood function but nonetheless are predicted to be ribosome-associated (Hummels and Kearns 2019). At present, it is hard to tell which suppressors directly inform mechanism and which highlight as-yet-unappreciated consequences of EF-P activity.

Evidence suggests that EF-P may do more than alleviate ribosome pausing at a subset of XPPX motifs. A recent study analyzed E. coli EF-P binding dynamics in vivo using super-resolution fluorescence microscopy and showed that EF-P interacted with ribosomes much more often than the frequency of XPPX motifs in transcripts (Mohapatra et al. 2017). It was proposed that EF-P samples empty ribosome E-sites during a majority of translation elongation cycles to minimize the time in which a stalled ribosome awaits rescue (Mohapatra et al. 2017). Alternatively, EF-P may alleviate stalling at additional primary sequences as it does when promoting translation of E. coli PoxB, a protein that does not contain any XPPX motifs (Hersch et al. 2013). EF-P also accelerates the incorporation of non-standard amino acids, such as certain D-amino acids and non-proteinogenic secondary amino acids when they are ligated to tRNA^pro (Doerfel et al. 2015; Katoh et al. 2016; Katoh, Iwane and Suga 2017). Beyond accelerating translation, EF-P has been shown promote reading frame maintenance at CC[C/U]-[C/U] mRNA sequences, particularly those located at the second codon of an open reading frame (Gamper et al. 2015). Finally, suppressors of the efp mutant growth defect in Erwinia amylovora were identified to contain loss-of-function mutations in the DEAH-box RNA helicase HrpA3, raising the possibility that EF-P has previously unappreciated effects on mRNA processing (Klee et al. 2019).

Whatever the mechanism of EF-P, current thinking seems to agree that its function is closely related to its structure, and the size and shape of the EF-P protein resembles that of a tRNA molecule (Fig. 3A,C). EF-P is composed of three domains. Domain III of EF-P resembles the part of tRNA that contains the anti-codon (Fig. 3A,C). While domain III of EF-P is essential and may recognize elements in the E-site codon, eIF5A lacks this domain (Fig. 3A,B) (Hanawa-Suetsugu et al. 2004; Huter et al. 2017). Domain II structurally contorts EF-P into a right angle and recognizes the presence of tRNA^Pro by interaction with a unique structure in the tRNA^Pro D-loop not found in any other tRNA save tRNA^fMet (Fig. 3A) (Katoh et al. 2016). Structural analysis of eIF5A does not support specific interactions with the D-loop tRNA, so while the two proteins appear to have similar functions, they appear to have different recognition elements, perhaps consistent with the ability of eIF5A to relieve pausing at non-proline containing peptides (Pelechano and Alepuz 2017; Schuller et al. 2017). The highest degree of conservation both at the sequence and structural level between EF-P and e/aIF5A is observed in domain I that resembles the activated end of the tRNA (Fig. 3A–C). When bound to ribosomes, domain I lies closest to the peptidyl-transferase center and may constitute the ‘active site’ for the relief of translational pausing (Fig. 3D). Further, another conserved feature of domain I between EF-P, eIF5A and aIF5A is the fact that a residue at the tip of domain I is post-translationally modified.

EF-P POST-TRANSLATIONAL MODIFICATION

In all organisms studied to date, EF-P is post-translationally modified on a conserved lysine or arginine at the position analogous to the aminoacylation site of tRNAs (Katz et al. 2013). Despite the ubiquity of post-translational modification (PTM) of EF-P, eIF5A, and aIF5A, the chemical identity of the modification varies dramatically between species. To date, the following moieties have been identified as EF-P, eIF5A, and aIF5A PTMs: R-β-lysinyl-hydroxylysine (E. coli and S. enterica), rhamnose (Pseudomonas aeruginosa, Neisseria meningitidis, and Shewanella oneidensis), hypusine (eukaryotes, and Sulfolobus solfataricus), deoxyhypusine (Haloferax volcanii) and 5-aminopentanol (B. subtilis) (Cooper et al. 1983; Navarre et al. 2010; Yanagisawa et al. 2010; Peil et al.2012; Lassak et al. 2015; Rajkovic et al. 2015; Prunetti et al. 2016; Rajkovic et al. 2016; Yanagisawa et al. 2016; Bassani et al. 2019) (Fig. 6). In most cases, the enzymes responsible for constructing the PTMs have been identified (reviewed in Lassak, Wilson and Jung 2016 and Rajkovic and Ibba 2017), but the exact biosynthetic pathway for 5-aminopentanol construction in B. subtilis as well as hypusine synthesis in archaea is unclear (Bassani et al. 2019; Witzky et al. 2018). Moreover, bioinformatic analyses have indicated that while EF-P is ubiquitous, many bacterial genomes lack homologs of the currently recognized EF-P PTM enzymes (Fig. 7) (Hummels et al. 2017; Rajkovic and Ibba 2017). Thus, it is likely that additional as-yet-unknown moieties are utilized to post-translationally modify EF-P.

Figure 6.

EF-P, eIF5A and aIF5A are post-translationally modified. Chemical structures of the moieties currently known to post-translationally modify EF-P, eIF5A or aIF5A. The R-group of the conserved lysine or arginine residue that serves as the site of post-translational modification is depicted in black and the modification group is depicted in green.

Figure 7.

Distribution of the EF-P post-translational modification machinery. Presence of homologs to the EF-P post-translational modification enzymes that have been identified to date across the tree of life. Numbers indicate the following clades (1) Flavobacterium-Cytophaga-Bacteroides group, (2) Chlamydiales, (3) Planctomycetes, (4) Spirochaetes, (5) Actinobacteria, (6) Deinococcus-Thermus group and (7) Cyanobacteria. DHS, EarP and EpmA homologs, were identified as reported in Hummels et al. 2017. DOHH, YmfI, YaaO, GsaB and YnbB homologs were identified using a BLAST search (BLAST + version 2.6.0, Camacho et al. 2009) with an E-value threshold of 1e-20. For each group of enzymes, the associated post-translational modification moiety is listed in parentheses. Where multiple enzyme groups are encoded in the same genome, the bar is split accordingly. White space indicates the absence of homologs to any known sets of EF-P modification enzymes.

Despite the chemical diversity of EF-P post-translational modification, the presence of the modification is thought to play an integral role in the EF-P mechanism. A close connection between modification and EF-P activity is supported by genetic analyses indicating that, in many organisms, mutation of the enzymes responsible for PTM resulted in phenotypes similar to that observed upon mutation of EF-P (Yanagisawa et al. 2010; Zou et al. 2012; Ude et al. 2013; Lassak et al. 2015; Yanagisawa et al. 2016; Klee et al. 2019). Moreover, structural analysis of E. coli EF-P bound to the ribosome indicated that the R-β-lysinyl-hydroxylysine moiety extended toward the peptidyl-transferase center and interacted with the aminoacylated end of the P-site tRNA to conformationally stabilize the nascent peptide chain (Huter et al. 2017). Thus, the modification group was in position to participate in the alleviation of ribosome pausing, but how the modification participates per se, and how chemically distinct modification groups serve similar functions, is unclear.

In addition to mechanistically participating in EF-P/eIF5a activity, some evidence suggests that the PTM may be regulatory. Many PTM pathways involve maturation of the modification while it is bound to EF-P, resulting, at least transiently, in the presence of modification intermediates. While the physiological consequences of these intermediates are still unclear, recent work in Fusarium graminearum suggests that they have the ability to alter eIF5A activity (Martinez-Rocha et al. 2016). F. graminearum eIF5A is modified by hypusine, and overexpression of deoxyhypusine synthase (DHS) resulted in a hypervirulent phenotype towards maize and wheat (Martinez-Rocha et al. 2016). Conversely, overexpression of the enzyme deoxyhypusine hydrolase (DOHH) dramatically reduced virulence (Martinez-Rocha et al. 2016). Whereas DHS is capable of both ligating and removing deoxyhypusine from eIF5A, the modification is only stabilized upon its conversion to hypusine by DOHH (Park et al. 2003). As such, overexpression of DOHH resulted in the complete hypusination of eIF5A, with no unmodified or deoxyhypusinated eIF5A detected in cell extracts (Martinez-Rocha et al. 2016). Thus, it was proposed that unmodified and/or partially modified eIF5A played a previously unrecognized role in vivo to promote hypervirulence.

Another regulatory consequence of partial modification was suggested when mutations in the B. subtilis PTM machinery had varied phenotypic effects. Mutation of YmfI, the enzyme responsible for catalyzing the reduction of 5-aminopentanone in the final step of 5-aminopentanol PTM biosynthesis, resulted in a reduction in EF-P activity (Kearns et al. 2004; Hummels et al. 2017). Activity could be enhanced in the absence of YmfI either by mutating synthesis enzymes earlier in the PTM pathway, thereby preventing the accumulation of 5-aminopentanone, or by mutating the post-translationally modified residue on EF-P and abolishing modification altogether (Hummels et al. 2017; Witzky et al. 2018). It was therefore concluded that PTM was not explicitly essential for EF-P activity in B. subtilis and that accumulation of one particular intermediate was instead inhibitory. Moreover, it has been recently shown in E. coli that either artificially modifying EF-P with a rhamnose moiety rather than the native R-β-lysinyl-hydroxylysine or simply overexpressing EF-P in a strain defective in PTM altogether was sufficient to promote the translation of an EF-P-dependent protein (Volkwein et al. 2019). Thus, at least in E. coli, modification per se is more important that the specific shape or chemistry of the modification group, and like that observed in B. subtilis, PTM is not strictly required for EF-P function.

EF-P ESSENTIALITY

EF-P clearly plays a role in the optimal expression of a wide variety of proteins but the biological importance is unclear. Early studies indicated that EF-P was essential for viability in E. coli, consistent with broad effects on translation. To demonstrate essentiality, the gene encoding EF-P, efp, was supplied on a temperature-sensitive plasmid and the chromosomal copy of efp was disrupted (Aoki et al. 1997). At the permissive temperature (33°C), the plasmid was able to replicate, EF-P was expressed, and the cells grew normally (Aoki et al. 1997). Upon shifting to the non-permissive temperature (44°C), plasmid replication failed, EF-P was depleted, the rate of protein synthesis was reduced, and cell growth arrested (Aoki et al. 1997). During the construction of large-scale deletion libraries of E. coli, however, efp deletion mutants were able to be isolated, although they imparted significant growth defects (Baba et al. 2006; Balibar, Iwanowicz and Dean 2013). Thus, EF-P appeared to promote growth but its strict essentiality seemed either inconsistent or conditional.

A recent study supported the conditional requirement for EF-P in the growth of E. coli (Tollerson, Witzky and Ibba 2018). It was found that propagating E. coli either at a lower temperature or in a minimal medium was sufficient to alleviate the growth defect of an EF-P mutant, likely by relieving the translation load (Tollerson, Witzky and Ibba 2018). Consistent with translational load dependency, polysomes accumulated in the absence of EF-P during growth at high but not low temperatures, and it was proposed that at higher temperatures (and higher growth rates) ribosomes became sequestered at XPPX motifs (Tollerson, Witzky and Ibba 2018). Thus, the conclusion that EF-P was essential in early studies can perhaps be attributed to the temperature-dependent nature of the EF-P mutant used. When the culture was shifted to high temperatures to evict the plasmid, the translational burden was increased to levels that required EF-P for the release of ribosomes from XPPX motifs and catastrophic inability to relieve ribosome sequestration led to apparent lethality.

While the majority of reports indicate that EF-P is required for growth, the severity of the requirement varies between organisms. In many organisms, including the eukaryotic systems of yeast, fruit flies and human cells lines, as well as the bacterial pathogens N. meningitidis and Mycobacterium tuberculosis, eIF5A/EF-P has been shown, and is still considered, to be absolutely essential for cell viability (Schnier et al. 1991, Patel et al. 2009; Yanagisawa et al. 2016; and Sassetti, Boyd and Rubin 2003). Likewise, in the archaeal system S. solfataricus, deletion of aIF5A was unsuccessful but knockdown of its expression resulted in a significant growth rate decrease, highlighting its importance for growth (Bassani et al. 2019). In contrast, some bacteria such as A. tumefaciens, Acinetobacter baylyi, S. enterica, and P. aeruginosa do not require strictly EF-P but experience reduced growth rate in its absence, like that observed in E. coli (Peng et al. 2001; de Crécy et al. 2007; Zou et al. 2012; Balibar, Iwanowicz and Dean 2013). Finally, there is at least one example, the bacterium B. subtilis, in which mutation of efp has little to no effect on growth, even during conditions in which the translational burden is presumably high (Rajkovic et al. 2016). Thus, based on the function of EF-P, it stands to reason that growth would most likely be impaired in its absence, but the connection between EF-P and growth is clearly not absolute.

Even if EF-P is required for growth, the reason it is required isn't entirely clear and may be due to pleiotropic effects. The observed pleiotropy in the absence of EF-P may be the direct result of stalling within, and limiting expression of, multiple essential proteins. An example of direct limitation is the stalling that occurs within E. coli ValS, the essential valine aminoacyl-tRNA synthetase (Starosta et al. 2014). Reduction of ValS levels in the absence of EF-P limits the production of charged valine-tRNAs and results in indirect stalling on valine codons in addition to poly-proline stretches, thereby exacerbating pleiotropy (Starosta et al. 2014, Woolstenhulme et al. 2015). Alternatively, pleiotropy may be indirect in that widespread ribosome stalling could sequester ribosomes and reduce overall ribosome availability. Reduced ribosome sequestration may explain why B. subtilis mutants defective in EF-P experience strong ribosome pausing in numerous essential genes, including ValS, but experience little effect on growth, as poly-prolines were found to be under-represented in the B. subtilis genome (Rajkovic and Ibba 2017; Hummels and Kearns 2019). Finally, eIF5A in S. cerevisiae has recently been shown to also promote translation termination and thus, at least in this system, essentiality may be unrelated to the relief of ribosome pausing (Pelechano and Alepuz 2017; Schuller et al. 2017).

Phenotypes that rely on protein complexes with specific subunit stoichiometries may be particularly sensitive to translational pausing in the absence of EF-P. For example, B. subtilis cells defective in EF-P exhibited a reduced number of flagella due to stalling within one particular structural protein (Hummels and Kearns 2019). Flagella are complex nanomachines and insufficient production of even one component restricts motility at the phenotypic level (Kearns et al. 2004; Hummels and Kearns 2019). Consistent with defects in relative stoichiometry, the requirement of EF-P for the activation of the E. coli Cad module was attributed to copy number fluctuations in a complex of two proteins (Ude et al. 2013). Specifically, EF-P alleviates pausing in the transcriptional activator CadC, such that in its absence, CadC levels decrease and become constitutively inhibited by direct interaction with its antagonist LysP (Neely, Dell and Olson 1994; Ude et al. 2013). Thus, EF-P is necessary to maintain a productive ratio of the two proteins to respond to low pH and lysine. Finally, EF-P promotes translation of one component of the F₁F₀ ATPase and a relative deficiency in this protein may result in a failure of the entire machine and contribute to growth rate reduction in E. coli (Elgamal et al. 2014; Hersch et al. 2014). It is difficult to evaluate the importance of EF-P in maintaining subunit stoichiometry within complexes or otherwise because at present few EF-P-related phenotypes have been directly attributed to a single target (Ude et al. 2013; Hummels and Kearns 2019).

CONCLUDING REMARKS

EF-P was first discovered as an elongation factor that promoted peptide bond formation between poor aminoacyl acceptors. For a time, EF-P was misunderstood to be involved in the synthesis of the first peptide bond, but now we recognize it as a bona fide elongation factor, albeit a very unusual one in that it exhibits sequence specificity. Molecular-genetic and systems-level approaches show that the specificity requirement hangs on improving the translation rate through primary sequences encoding particularly poor peptidyl transfer substrates like consecutive prolines. Thus, after nearly 4 decades of study, our understanding of the activity of EF-P has come full circle and the role of EF-P in translation proposed by Glick, Chládek and Ganoza in 1979 appears to be accurate.

While we understand a great deal more about EF-P today than when it was discovered, many aspects of EF-P biology remain mysterious. For example, EF-P promotes translation through many but not all XPPX motifs hinting at still further rules that govern its specificity. Moreover, not all proteins that experience XPPX-mediated stalling actually experience reduced protein copy number (Elgamal et al. 2014; Woolstenhulme et al. 2015). Exactly why some XPPX motifs are particularly difficult to translate isn't entirely clear, nor is the mechanism by which EF-P relieves this problem. The function of post-translational modification, once thought to be essential for EF-P activity, warrants reconsideration as unmodified EF-P and variants modified with a foreign chemical moiety have been recently reported as functional (Hummels et al. 2017; Volkwein et al. 2019). If the modification is part of the mechanism by which EF-P relieves translational pausing, how can so many diverse modifications all accommodate a conserved function? Alternatively, if the modification is regulatory, when and how is the modification added or altered to change EF-P function?

Finally, and perhaps the biggest question, is: how does EF-P promote growth? The absence of EF-P results in reduced levels of a subset of proteins some of which are enzymes involved in essential metabolism. Enzymes are rather resistant to fluctuation in abundance, however, and it is not clear why a reduction in their level(s) would be growth limiting or even inhibitory. Moreover, recent work suggests that the requirement of EF-P, at least in E. coli, can be conditionally relieved simply by inducing slower growth at lower temperatures (Tollerson, Witzky and Ibba 2018). If slow growth and the corresponding reduction in translation rate is sufficient to abolish the need for EF-P, then why does EF-P appear to be essential in slow-growing organisms such as Mycobacterium (Sassetti, Boyd and Rubin 2003)? It is clear that EF-P is conserved in both form and function in all organisms across every domain of life and plays a specific but nonetheless important role in maintaining high rates of translation. Growth essentiality would be a strong selection to ensure such high conservation, but if EF-P is not strictly required for growth, why is it so highly conserved?