Gut colonization by Bacteroides requires translation by an EF‐G paralog lacking GTPase activity

Abstract

Protein synthesis is crucial for cell growth and survival yet one of the most energy‐consuming cellular processes. How, then, do cells sustain protein synthesis under starvation conditions when energy is limited? To accelerate the translocation of mRNA–tRNAs through the ribosome, bacterial elongation factor G (EF‐G) hydrolyzes energy‐rich guanosine triphosphate (GTP) for every amino acid incorporated into a protein. Here, we identify an EF‐G paralog—EF‐G2—that supports translocation without hydrolyzing GTP in the gut commensal bacterium Bacteroides thetaiotaomicron. EF‐G2's singular ability to sustain protein synthesis, albeit at slow rates, is crucial for bacterial gut colonization. EF‐G2 is ~10‐fold more abundant than canonical EF‐G1 in bacteria harvested from murine ceca and, unlike EF‐G1, specifically accumulates during carbon starvation. Moreover, we uncover a 26‐residue region unique to EF‐G2 that is essential for protein synthesis, EF‐G2 dissociation from the ribosome, and responsible for the absence of GTPase activity. Our findings reveal how cells curb energy consumption while maintaining protein synthesis to advance fitness in nutrient‐fluctuating environments.

Bacterial elongation factor paralog EF‐G2 is abundant during carbon starvation and maintains slow translation during nutrient‐limiting conditions.

Introduction

Protein synthesis is carried out by the ribosome. Following incorporation of each amino acid into a growing polypeptide chain, the ribosome moves by three nucleotides along the mRNA. The two tRNAs bound to the A and P sites of the ribosome move to the P and E sites, respectively, exposing a new codon in the A site. This movement—referred to as translocation—can happen spontaneously but is accelerated by four orders of magnitude in the presence of specific translation factors essential in all living cells. These translation factors—elongation factor G (EF‐G) in prokaryotes and elongation factor 2 (EF2) in eukaryotes—are guanosine triphosphate (GTP)‐hydrolases (GTPases) that use the energy derived from GTP hydrolysis to accelerate translocation (Rodnina et al, 2019). EF‐G also functions with ribosome recycling factor (RRF) to promote the rapid dissociation of the ribosomal complex into the small and large subunits after release of the synthetized protein, also at the cost of GTP hydrolysis (Hirashima & Kaji, 1973; Peske et al, 2005). Here, we present the first example of an EF‐G that sustains protein synthesis without consuming GTP and establish that this ability is critical for bacterial colonization of the mammalian gut.

Members of the Bacteroidetes (Bacteroidota) phylum, especially the Bacteroides genus, make up a major portion of the human gut microbiota and are widespread across human populations (Wexler & Goodman, 2017). Bacteroides species are obligate anaerobes highly adapted to life in the gut as they live and grow exclusively in the gastrointestinal tracts of mammals (Ley et al, 2008). They are the foundation of the microbial food webs in the gut because they can break down a wide range of complex sugars (including dietary plant polysaccharides, host glycans, and milk oligosaccharides) that are major carbon sources (Martens et al, 2014; Schwalm III & Groisman, 2017; La Rosa et al, 2022). Nevertheless, Bacteroides species appear to experience carbon limitation in the gut because proteins and signaling molecules that accumulate upon carbon starvation are required for gut colonization (Schofield et al, 2018; Townsend II et al, 2020). This raises the question: How do Bacteroides species sustain protein synthesis under energy‐limiting conditions?

Bacteroides thetaiotaomicron harbors two EF‐G‐like proteins: the essential EF‐G1 (BT2729), which shares 57% amino acid identity (73% similarity, Fig EV1A) with the canonical Escherichia coli EF‐G, and EF‐G2 (BT2167), which shares 34% amino acid identity (53% similarity, Fig EV1A) with E. coli EF‐G. EF‐G1 and EF‐G2 share 31% amino acid identity and 51% amino acid similarity (Fig EV1A). EF‐G2 is required for successful colonization of the murine gut (Wu et al, 2015; Townsend II et al, 2020) but dispensable under laboratory conditions. When B. thetaiotaomicron experiences carbon limitation, the mRNA amounts of the canonical EF‐G1 (BT2729) gene decrease ~10‐fold (Townsend II et al, 2020), which is consistent with translational shut‐down under starvation conditions and makes sense given that protein synthesis is the most energy‐demanding cellular activity and that carbon starvation dramatically reduces energy production. By contrast, the mRNA amounts of the EF‐G2 (BT2167) gene increase > 200‐fold during carbon limitation (Townsend II et al, 2020). That EF‐G1 is essential but EF‐G2 is not distinguishes B. thetaiotaomicron from bacterial species that harbor two essential EF‐G‐like proteins, one responsible for translocation and the other responsible for ribosome recycling (Suematsu et al, 2010; Margus et al, 2011).

Figure EV1. Sequence alignments of different EF‐G‐like proteins.

• A
  Alignment of the deduced amino acid sequences of EF‐G1 (EF‐G), EF‐G2, spdEF‐G1 and spdEF‐G2 proteins. spdEF‐G1 spdEF‐G2 are EF‐G paralogs found only in Spirochaetes, Planctomycetes and Delta‐proteobacteria, which are the only bacteria that do not harbor EF‐G orthologs (Margus et al, 2011). Borrelia burgdorferi (Bbu) spdEF‐G1 is only active in translocation but not ribosome recycling; by contrast, Bbu spdEF‐G2 was found to function specifically in ribosome recycling without translocation activity (Suematsu et al, 2010). Black triangle indicates the position of S84 in B. thetaiotaomicron EF‐G2 and H91 in E. coli EF‐G. Black diamonds indicate the conserved glutamine in loop I of domain IV, including Q507 in E. coli EF‐G and Q491 in B. thetaiotaomicron EF‐G2, and conserved histidine in loop II of domain IV, including H583 in E. coli EF‐G and H593 in B. thetaiotaomicron EF‐G2. Msm, Mycobacterium smegmatis; Tth, T. thermophilus; Bbu, Borrelia burgdorferi; Bth, B. thetaiotaomicron; Eco, E. coli; Pae, Pseudomonas aeruginosa.
• B
  Alignment of the deduced amino acid sequences of EF‐G2 proteins from the Bacteroidetes phylum showing the S84 region and the region including the Bacteroidetes EF‐G2‐specific insert.

Here, we identify B. thetaiotaomicron EF‐G2 as the first natural EF‐G protein that promotes translation elongation without GTP hydrolysis. We determine that EF‐G2 does not hydrolyze GTP in the presence of pre‐translocation ribosome complexes or ribosomes without bound tRNAs or mRNA (vacant ribosomes) even though EF‐G2 binds GTP more tightly than the canonical B. thetaiotaomicron EF‐G1. Cryo‐electron microscopy analysis reveals that the GTP‐binding pocket of EF‐G2 is located further away from the sarcin‐ricin loop of the ribosome than in complexes with canonical EF‐G proteins, which may account for the lack of ribosome‐stimulated GTPase activity in EF‐G2. An engineered B. thetaiotaomicron strain with a translocation‐deficient EF‐G2 variant replacing wild‐type EF‐G2 is as defective in gut colonization as an EF‐G2 null mutant. Taken together with the high abundance of EF‐G2 in bacteria harvested from murine ceca that contrasts with the exceedingly low amounts of EF‐G1, this indicates that protein synthesis mediated by EF‐G2 is essential for gut colonization. Our findings reveal how paralogous translation factors enable commensal bacteria to switch from rapid energy‐consuming protein synthesis to a slower energy‐efficient process, advancing bacterial fitness in nutrient‐fluctuating environments.

Results

EF‐G2 is well conserved across the Bacteroides genus and displays unique sequence signatures

Phylogenetic analysis of the EF‐G1‐ and EF‐G2‐encoding genes suggests that they originated from ancestral duplicated genes that likely existed in the last common ancestor of bacteria (Atkinson, 2015). EF‐G2‐encoding genes have been identified in every bacterial phylum but only in about one fourth of the species (Margus et al, 2011). Notably, EF‐G2 never exists as the sole EF‐G‐like protein in any organism, being always accompanied by either an ortholog of canonical EF‐G or both spdEF‐G1 and spdEF‐G2, which are EF‐G‐paralogs specialized in translocation and ribosome recycling, respectively, and present in Spirochaetes, Planctomycetes, and Delta‐proteobacteria in place of canonical EF‐G (Suematsu et al, 2010; Margus et al, 2011). The scattered distribution of EF‐G2‐specifying genes in currently available genomes may have resulted from gene loss in some lineages during evolution. The deduced amino acid sequences of the extant EF‐G2 proteins reveal more divergence than EF‐G proteins, harboring lineage‐specific motifs and insertions/deletions (Margus et al, 2011) that may confer lineage‐specific functions. One such function(s) is likely responsible for EF‐G2 being required for colonization of the mammalian gut by three different Bacteroides species: B. thetaiotaomicron, B. cellulosilyticus, and B. ovatus (Wu et al, 2015). This prompted us to investigate the conservation of EF‐G2 in the Bacteroides lineage.

Genome analysis revealed that genes specifying EF‐G2 are present in all 71 analyzed fully sequenced genomes from the Bacteroides genus, with the corresponding amino acid sequences sharing 75–100% identity with the B. thetaiotaomicron EF‐G2 protein (Fig 1A). Outside the Bacteroides genus, EF‐G2‐specifying genes share lower deduced amino acid sequence identity (60–80% in Bacteroidaceae and 48–60% in the rest of Bacteroidetes) with the B. thetaiotaomicron EF‐G2 (Fig 1A). Of the 146 genomes from the Bacteroidia class analyzed, only two lack an EF‐G2‐specifying gene (Fig 1A). By contrast, an EF‐G2‐specifying gene is found in a smaller portion of Bacteroidetes outside the Bacteroidia class (10/197 Flavobacteriia, 2/30 Sphingobacteriia, 3/100 Cytophagia, 19/29 Chitinophagia, 61/81 Saprospiria genomes analyzed).

Figure 1. EF‐G2 is well conserved across the Bacteroides genus.

• A
  Phylogenetic tree of 149 whole‐genome sequenced strains from the Bacteroidetes phylum, including 71 Bacteroides spp. strains (light blue background), with the blue‐yellow heatmap showing the presence of EF‐G1 or EF‐G2 sequelog, and the identity between each homolog and the corresponding B. thetaiotaomicron VPI‐5482 protein; the ring of letters (aa84) shows amino acid at the position 84 (B. thetaiotaomicron EF‐G2 numbering) of corresponding EF‐G2 proteins; and the outermost bar‐charts [Insert (bp)] shows the length of the insert corresponding to residues 514–539 in B. thetaiotaomicron EF‐G2. Background colors: blue—Bacteroides genus; purple—Bacteroidaceae family; gray—Bacteroidales order, Bacteroidia class. The three listed Bacteroidetes lacking EF‐G2 sequelogs are: Flavobacterium johnsoniae, which is a soil bacterium; Candidatus Azobacteroides pseudotrichonymphae, which was isolated from a single cell of the protist Pseudotrichonympha grassii, which resides in the termite gut; Porphyromonas sp. KLE 1280, which is a human oral bacterium.
• B, C
  Sequence logos of the aligned deduced amino acid sequences of the EF‐G1 and EF‐G2 proteins from Bacteroides species showing the regions corresponding to Loop I (B) and Loop II (C) at the tip of domain IV in canonical EF‐G. Only unique (non‐redundant) protein sequences were used for the alignment.
• D
  Sequence logos of aligned deduced amino acid sequences of EF‐G1 and EF‐G2 proteins from Bacteroides species showing the region surrounding the EF‐G2 insert with amino acids that form beta‐strand and beta‐hairpin in B. thetaiotaomicron EF‐G2 shown by arrows below the logos.
• E
  Sequence logos of aligned deduced amino acid sequences of EF‐G1 and EF‐G2 proteins from Bacteroides species showing the G3‐box (DTPG in canonical EF‐G) followed by the conserved histidine in Bacteroides EF‐G1s and the corresponding serine/alanine in Bacteroides EF‐G2s. The B. thetaiotaomicron EF‐G2 has a serine at this position (S84).

All Bacteroides EF‐G2 proteins harbor the conserved glutamine Q507 and histidine H583 residues (E. coli EF‐G numbering; Figs 1B and C, and EV1A), which contact tRNA and are critical for translocation by EF‐G (Savelsbergh et al, 2000; Gao et al, 2009). These residues are conserved in EF‐G orthologs, EF‐G2 proteins, and spdEF‐G1 but not in the translocation‐incompetent spdEF‐G2 (Fig EV1A; Margus et al, 2011).

A distinguishing feature of Bacteroides EF‐G2 is the presence of a 26‐amino acid long insert in domain IV found neither in EF‐G proteins nor in EF‐G2 proteins from bacteria outside the Bacteroidetes phylum (Fig EV1A; Margus et al, 2011). The insert is shared by all Bacteroides EF‐G2 proteins, having the same length and well‐conserved amino acid sequence (Fig 1D). By contrast, EF‐G2 proteins from Bacteroidetes outside the Bacteroidia class harbor a shorter 17 residue‐long insert (Figs 1A and EV1B).

Curiously, histidine H91 (E. coli EF‐G numbering), which is conserved in EF‐G orthologs (Fig EV1A) as well as in other translational GTPases, including EF‐Tu, RF3, LepA, and IF2, is not conserved in EF‐G2 proteins, being substituted by either a serine (S84 in B. thetaiotaomicron EF‐G2) or an alanine in Bacteroides EF‐G2 proteins (Fig 1E). An earlier study of a more limited number of genomes reported that the consensus amino acid at this position of EF‐G2 is serine in the Bacteroidetes phylum and Alpha‐proteobacteria and cyanobacteria groups, leucine in Chlorobi, and phenylalanine or tyrosine in other groups of bacteria (Margus et al, 2011). We noticed that EF‐G2s from the Bacteroidia class harbor serine, alanine, or methionine at this position and that EF‐G2s from the other classes of Bacteroidetes harbor tyrosine, phenylalanine, or leucine (Figs 1A and EV1B). Despite harboring a tyrosine at this position, the Thermus thermophilus EF‐G2 is active in GTP‐hydrolysis and translation elongation (Connell et al, 2007). By contrast, an E. coli EF‐G variant with the H91A substitution is inactive in both GTP hydrolysis and translation elongation but can support a single round of slow tRNA–mRNA translocation (Cunha et al, 2013). This prompted us to determine whether EF‐G2 from B. thetaiotaomicron is a functional translation factor.

EF‐G2 is a bona fide translation elongation factor that promotes slow translocation

To determine whether EF‐G2 can support protein synthesis, we used an in vitro reconstituted custom‐made protein synthesis system (PURExpress, New England Biolabs) containing a defined mix of purified E. coli transcription and translation components except for EF‐G. Recombinant B. thetaiotaomicron EF‐G2 purified from E. coli promoted synthesis of a reporter protein specified by the provided DNA template (Fig 2A), albeit more slowly than recombinant B. thetaiotaomicron EF‐G1 purified under the same conditions (Fig 2A). EF‐G1 exhibited similar activity to E. coli EF‐G (Fig 2A), validating the approach of using the E. coli‐based system to study B. thetaiotaomicron EF‐G proteins.

Figure 2. B. thetaiotaomicron EF‐G2 supports slower translation elongation than B. thetaiotaomicron EF‐G1 and E. coli EF‐G.

• A
  Western blot analysis of reporter HslO‐FLAG protein produced with a custom‐made PURExpress® coupled in vitro transcription‐translation system supplemented with the indicated EF‐G proteins and incubated for the indicated times. Blot was developed using anti‐FLAG antibodies. Shown is a representative from at least two independent experiments. Eco EF‐G: E. coli EF‐G.
• B
  Ribosome translocation determined as the fluorescence change of the fluoresceine‐labeled mRNA in stopped‐flow experiment supplemented with the indicated proteins and nucleotides. Translocation rates are as follows: EF‐G1 with GTP, 6.9 ± 0.1 s⁻¹; EF‐G1 with GDPNP, 0.64 9 ± 0.01 s⁻¹; EF‐G2 with GTP, 0.7 ± 0.01 s⁻¹; EF‐G2 with GDPNP, 0.62 ± 0.01 s⁻¹. See Materials and Methods for details. Shown are averages of 5–7 technical replicates.
• C
  Tripeptide formation assay using translation initiation complex programed with the tripeptide‐encoding mRNA, corresponding aminoacyl‐tRNA:EF‐Tu:GTP ternary complexes and GTP, and the indicated EF‐G proteins following incubation for 1 or 5 min. Shown are the results from two independent experiments and their average. Insert: tripeptide formation with EF‐G2 following incubation for 0, 1, 5, and 10 min.

Control experiments verified that the observed protein synthesis is mediated by the added EF‐G proteins. First, no reporter protein was produced in the absence of E. coli EF‐G, EF‐G1, or EF‐G2 (Fig 2A), indicating that the protein synthesis system was free of endogenous E. coli EF‐G. Second, the B. thetaiotaomicron EF‐G2 (H593K) mutant protein, purified using the same method as for purification of wild‐type EF‐G2, was inactive in protein synthesis (Fig 2A). H593 in B. thetaiotaomicron EF‐G2 corresponds to H583 in E. coli EF‐G (Fig EV1A), a histidine residue which upon substitution to lysine renders EF‐G unable to support rapid translocation (Savelsbergh et al, 2000). The inability of the purified EF‐G2 (H593K) variant to sustain protein synthesis indicates that the protein synthesis activity conferred by recombinant wild‐type EF‐G2 is unlikely to result from contamination with endogenous E. coli EF‐G from the bacterial host used to overproduce the B. thetaiotaomicron EF‐G2 protein.

We then tested whether the slower rate of protein synthesis conferred by EF‐G2 relative to EF‐G1 (Fig 2A) is due to slower tRNA–mRNA translocation in the presence of EF‐G2. We measured translocation kinetics in a fully reconstituted translation system using an established stopped‐flow assay that monitors the fluorescence change of fluorescein attached to the 3′ end of a short mRNA upon translocation (Savelsbergh et al, 2003). The experiments were carried out in the presence of excess EF‐G, which allowed us to capture the effect on translocation independently of the subsequent EF‐G dissociation step. In the presence of GTP, EF‐G2‐mediated translocation was significantly slower than that mediated by EF‐G1 (0.7 ± 0.1 s⁻¹ vs. 6.9 ± 0.2 s⁻¹; Fig 2B), which was as fast as that supported by E. coli EF‐G (Fig EV2A). EF‐G2's behavior is reminiscent of that exhibited by the GTPase‐deficient E. coli EF‐G (H91A) variant (Cunha et al, 2013). In agreement with this notion, translocation carried out with EF‐G1 in the presence of the non‐hydrolysable GTP analog GDPNP (guanosine 5′‐[β,γ‐imido]triphosphate; Fig 2B) was much slower than in the presence of GTP (Fig 2B) and resembled the kinetics of EF‐G2‐promoted translocation in the presence of either GDPNP or GTP (Fig 2B). Moreover, the fluorescence change in the reaction with EF‐G2 reached a lower amplitude than that in reactions with EF‐G1 or E. coli EF‐G (Figs 2B and EV2A). The lower fluorescence amplitude, which is also observed with E. coli EF‐G in the presence of other non‐hydrolysable GTP analogs (Belardinelli et al, 2016), is due to the inhibition of 30S ribosomal subunit movements, namely the impaired swivel motion of the head domain of the 30S ribosomal subunit back to the ground state (Belardinelli et al, 2016). The ribosomes are blocked in an intermediate state (termed chimeric) with the mRNA fully translocated relative to the body domain of the 30S subunit, whereas translocation on the head domain is incomplete, which leads to an incomplete fluorescence amplitude with mRNA reporters.

Figure EV2. EF‐G2 supports translocation but not ribosome splitting.

• A
  Ribosome translocation measured by monitoring the fluorescence change of the fluoresceine‐labeled mRNA in a stopped‐flow experiment supplemented with the indicated proteins. Translocation rates are as follows: E. coli EF‐G, 8.4 ± 0.1 s⁻¹; EF‐G1, 8.3 ± 0.2 s⁻¹; EF‐G2, 0.7 ± 0.1 s⁻¹. See Materials and Methods for details. Shown are averages of 5–7 technical replicates.
• B
  Ribosome translocation determined by monitoring the fluorescence change of the fluoresceine‐labeled mRNA in a stopped‐flow experiment supplemented with the indicated proteins and nucleotides. See Materials and Methods for details. Shown are averages of 5–7 technical replicates.
• C
  Western blot analysis of polysome fractions prepared from strain WH407, which expresses EF‐G1‐FLAG and EF‐G2‐HA from their normal chromosomal locations and promoters. Blots were developed using anti‐FLAG and anti‐HA antibodies.
• D
  Ribosome splitting effect of EF‐G1, EF‐G2, or E. coli EF‐G in the presence of RRF from the same species were measured by light scattering in a stopped‐flow apparatus. See Materials and Methods for details.

To further characterize translocation in the presence of EF‐G2, we monitored formation of a tripeptide upon addition of EF‐G to ribosome initiation complexes, aminoacyl‐tRNA:EF‐Tu:GTP ternary complexes, and GTP (Peng et al, 2019). After formation of the first peptide bond (which is independent of translocation), binding of the second aminoacyl‐tRNA:EF‐Tu:GTP can occur only if tRNA–mRNA translocation is completed and EF‐G has dissociated from the ribosome. EF‐G2 promoted complete translocation to an extent comparable to that with EF‐G1 or E. coli EF‐G (Fig 2C), but with a much lower rate, about 0.004 s⁻¹ (Fig 2C, inset). This is much slower than the dissociation of E. coli EF‐G, about 4 s⁻¹ (Belardinelli et al, 2016), or the mRNA translocation in the presence of EF‐G2 measured with the translocation assay in Fig 2B (0.7 s⁻¹). Because formation of the tripeptide requires that EF‐G dissociate, the very slow tripeptide reaction reflects the completion of the translocation reaction (see Fig 2B) and EF‐G2 dissociation. Tripeptide formation with EF‐G1 (Fig 2B) was too rapid to estimate the dissociation rate. These results demonstrate that EF‐G2 is able to support translation, albeit at much slower rates than the conventional EF‐G species.

We considered the possibility of the alarmone ppGpp impacting translocation by EF‐G1 and/or EF‐G2 because: (i) it inhibits various GTPases involved in mRNA translation and ribosome biogenesis (Rojas et al, 1984; Mitkevich et al, 2010; Corrigan et al, 2016; Pausch et al, 2018; Diez et al, 2020; Vinogradova et al, 2020); and (ii) carbon starvation triggers accumulation of both ppGpp (Schofield et al, 2018) and the mRNA corresponding to the EF‐G2‐encoding BT2167 gene (Townsend II et al, 2020) in B. thetaiotaomicron. However, ppGpp failed to inhibit translocation by either EF‐G1 or EF‐G2 when present at equimolar amounts as GTP (Fig EV2B).

In agreement with the results of the in vitro protein synthesis experiments discussed above, both EF‐G1 and EF‐G2 co‐sedimented with translating polysomes when a B. thetaiotaomicron cell lysate was applied to a sucrose gradient and subjected to centrifugation (Fig EV2C), indicating that both EF‐G1 and EF‐G2 interact with translating ribosomes in vivo.

We next examined if EF‐G2 functions during ribosome recycling. When incubated with IF3 and monitored ribosomal complex disassembly by light scattering, B. thetaiotaomicron EF‐G1 and B. thetaiotaomicron RRF promoted splitting of vacant E. coli 70S ribosomes, although the effect was small when compared to reactions catalyzed by E. coli EF‐G and E. coli RRF (Fig EV2D). By contrast, no ribosome splitting was observed with B. thetaiotaomicron EF‐G2 and B. thetaiotaomicron RRF, which appeared to stabilize the ribosomal complex (Fig EV2D).

Cumulatively, the results in this section establish that EF‐G2 is a bona fide translation elongation factor that mediates translocation, albeit at a slower rate than canonical EF‐G proteins. Moreover, they raised the possibility that EF‐G2 lacks the ability to bind or hydrolyze GTP.

EF‐G2 lacks ribosome‐stimulated GTPase activity

Canonical EF‐G hydrolyses GTP in the presence of vacant or pre‐translocation ribosomes (Rodnina et al, 1997). The B. thetaiotaomicron EF‐G1 exhibited similar GTPase activity to E. coli EF‐G (Figs 3A and EV3A). By contrast, EF‐G2 did not hydrolyze GTP in the presence of vacant E. coli ribosomes (Fig EV3A) or crude ribosomes purified from B. thetaiotaomicron (Fig 3B). EF‐G2 also lacked GTPase activity in the presence of pre‐translocation E. coli ribosomes (Fig 3A), indicating that it does not hydrolyze GTP when mediating translocation. Control experiments demonstrated that, like E. coli EF‐G (Kuriki et al, 1970), neither EF‐G1 nor EF‐G2 display intrinsic GTPase activity (Fig EV3B).

Figure 3. EF‐G2 binds GTP but lacks ribosome‐stimulated GTPase activity.

• A
  GTP hydrolysis exhibited by the indicated EF‐G proteins in the presence of pre‐translocation ribosomes. Shown are the results from three independent experiments.
• B
  GTP hydrolysis exhibited by the indicated EF‐G proteins in the presence of a B. thetaiotaomicron ribosome preparation. Shown are the results from three technical replicates and their average, error bars correspond to SD. P‐values are from two‐tailed Student's t‐test between each protein and the blank, ns indicate P > 0.05.
• C
  GTP hydrolysis by EF‐G1 (1 μM) measured in the presence of a fixed amount of vacant E. coli ribosomes and varied amounts of EF‐G2 (1–10 μM, 1×–10×). Shown are the results from three independent experiments and their average, error bars correspond to SD.
• D
  GTP binding by EF‐G1 and EF‐G2 measured using differential radial capillary action of ligand assay (DraCALA; Roelofs et al, 2011). Shown are results of three technical replicates. The DraCALA blot of one representative replicate is shown in Fig EV3C. See Materials and Methods for details.

Figure EV3. EF‐G2 preferably binds GTP over GDP but lacks GTPase activity.

• A
  GTP hydrolysis exhibited by the indicated EF‐G proteins in the presence of vacant E. coli ribosomes. Mean values and SD for three independent experiments are shown.
• B
  No GTP hydrolysis is exhibited by the EF‐G1 and EF‐G2 proteins in the absence of ribosomes.
• C
  GTP binding of the EF‐G1 and EF‐G2 proteins (at the indicated concentrations) by differential radial capillary action of ligand assay (DraCALA; Roelofs et al, 2011). A representative from three replicates is shown.
• D
  Binding of radioactive GTP to the EF‐G1 and EF‐G2 proteins in the presence of the indicated concentrations non‐radioactive (cold) GTP or GDP (assayed via DraCALA). A representative from three replicates is shown.

EF‐G2 inhibited ribosome‐stimulated GTP hydrolysis by EF‐G1 (Fig 3C), suggesting that EF‐G2 and EF‐G1 bind to the ribosome at overlapping sites. However, EF‐G2 appears to have lower affinity for the ribosome than EF‐G1 because ~2.5 times more EF‐G2 than EF‐G1 was necessary to reduce EF‐G1‐stimulated GTP hydrolysis by half (Fig 3C).

Using differential radial capillary action of ligand assay (DraCALA; Roelofs et al, 2011), we determined that EF‐G2 binds GTP with higher affinity than EF‐G1 (Figs 3D and EV3C). Additionally, unlabeled GTP was better than GDP at competing with radiolabeled GTP for binding to EF‐G2 (Fig EV3D), suggesting that EF‐G2 has higher affinity for GTP than for GDP.

These results demonstrate that EF‐G2 lacks ribosome‐stimulated GTPase activity, despite binding GTP and the ribosome. To our knowledge, EF‐G2 is the first example of an EF‐G protein that sustains translation elongation despite lacking ribosome‐stimulated GTPase activity.

An EF‐G2‐specific region required for protein synthesis

A salient property of the EF‐G2 subfamily of proteins is the presence of insertions and deletions, which is uncommon in other EF‐G subfamilies (Margus et al, 2011). The B. thetaiotaomicron EF‐G2 protein harbors a 26‐amino acid long insert (positions 514 to 539; Figs 4A and EV1A; Margus et al, 2011) that is absent from translocation‐competent EF‐G proteins, such as E. coli EF‐G, T. thermophilus EF‐G, and T. thermophilus EF‐G2 (Connell et al, 2007). An EF‐G2 variant lacking the insert is unable to sustain protein synthesis (Fig 4B) even though it exhibits low ribosome‐stimulated GTPase activity, unlike wild‐type EF‐G2, albeit much lower than that exhibited by EF‐G1 (Fig 4C).

Figure 4. Biochemical features of B. thetaiotaomicron EF‐G2.

• A
  Structure of EF‐G2 domain IV highlighting the insert (residues 514–539) and hairpin (residues 516–527) regions.
• B
  Western blot analysis of reporter HslO‐FLAG protein synthesized in vitro using a custom‐made PURExpress® system with EF‐G1, EF‐G2, and two engineered EF‐G2 variants: one missing the 26‐amino acid insert, EF‐G2 (∆514–539) and one missing the beta hairpin in the insert, EF‐G2 (∆516–527). Shown is a representative experiment from at least two independent experiments.
• C
  GTP hydrolysis by EF‐G1, EF‐G2 and indicated variants incubated in the presence of vacant E. coli ribosomes. Shown are results from three technical replicates and their average, error bars correspond to SD, and P‐values derived from two‐tailed Student's t‐test comparing each protein to the blank.
• D
  SDS–PAGE analysis of co‐sedimentation assays monitoring factor (EF‐G1 EF‐G2 or EF‐G2 variant EF‐G2 (∆516–527) or EF‐G2 (∆514–539)) binding to E. coli 70 S ribosomes. Binding reactions were centrifuged through a sucrose cushion to recover ribosome and associated factor in the pellets. The positions of EF‐G proteins and ribosomal proteins and molecular weight markers are indicated.
• E
  Western blot analysis of reporter HisC‐FLAG protein synthesized in vitro using a custom‐made PURExpress® system with the EF‐G1, EF‐G2 or EF‐G2 (S84H) proteins. Shown is a representative from at least two independent experiments.

In the cryo‐EM structure (discussed in detail in the following section), the insert is resolved as a protrusion consisting of a beta‐hairpin and an additional beta‐strand on the beta‐sheet of domain IV of EF‐G2 (Fig 4A). A comparison of the insert‐less variant EF‐G2 (∆514–539) to the variant EF‐G2 (∆516–527) lacking only the beta‐hairpin revealed distinct biochemical properties. While both EF‐G2 variants failed to sustain protein synthesis (Fig 4B), the EF‐G2 (∆516–527) variant had no ribosome‐stimulated GTPase activity (Fig 4C), like to the wild‐type EF‐G2 protein, whereas the EF‐G2 (∆514–539) variant had some low GTPase activity (Fig 4C). Notably, inserts are shorter in EF‐G2 proteins from Bacteroidetes outside the Bacteroidia class (the one to which Bacteroides species belong), lacking most of the hairpin region (Fig EV1B).

We next asked if the insert is required for EF‐G2 to dissociate from the ribosome. This property enables the naturally GTPase‐deficient EF‐G2 to continue protein synthesis (Fig 2A and C), which sets it apart from the GTPase‐deficient variant of E. coli EF‐G with the H91A substitution (Cunha et al, 2013). We determined that large amounts of the EF‐G2 (∆516–527) and EF‐G2 (∆514–539) variants co‐sedimented with ribosomes when incubated with GTP, centrifuged on a sucrose cushion, and analyzed in the pelleted ribosomal complex by SDS–PAGE (Fig 4D). These EF‐G2 variants were recovered at near stoichiometric amounts with the ribosomal proteins, indicating that they form a stable complex with the ribosome. By contrast, little wild‐type EF‐G2 or EF‐G1 was detected in the pellet fraction with ribosomal proteins (Fig 4D).

We explored the possibility that EF‐G2 lacks GTPase activity due to the presence of a serine instead of histidine at position 84 (corresponding to H91 in E. coli EF‐G). However, the EF‐G2 (S84H) variant exhibited marginally higher GTPase activity than wild‐type EF‐G2 (Fig 4C) that is not statistically significant (P = 0.12 two‐tailed Student's t‐test). Curiously, the EF‐G2 (S84H) variant supported faster protein synthesis than wild‐type EF‐G2 (Fig 4E).

Collectively, the results in this section identify a region of EF‐G2 necessary for protein synthesis despite being absent from other EF‐G proteins, including members of the EF‐G2 subfamily. In addition, they indicate that the identified region contributes to both the lack of GTPase activity and to EF‐G2's dissociation from the ribosome.

Ribosome‐bound EF‐G2 in the GTP‐bound conformation reveals GTPase domain shifted away from the sarcin‐ricin loop

To gain insights into EF‐G2's unique biochemical properties, we examined the structure of EF‐G2 bound to the E. coli ribosome using cryo‐EM (see Materials and Methods for details about the cryo‐EM data collection and structure solving). From a sample containing ribosomes, EF‐G2, and GTP in a buffer with 10 mM Mg²⁺, we obtained density maps for vacant ribosomes (70S) and ribosome‐EF‐G2 complexes (70S‐EF‐G2; Appendix Fig S1) at a global resolution of 2.7 and 2.9 Å, respectively (Appendix Fig S2A and B). After 3D classification, the two best classes of 70S‐EF‐G2 particles (Class 1 and Class 2) were further refined to a global resolution of 3.2 and 3.0 Å, respectively (Appendix Figs S1, and S2C and D). Class 1 corresponds to EF‐G2 bound to a non‐rotated ribosome with ~2° of 30S head domain swiveling (Fig EV4A). Class 2 corresponds to EF‐G2 bound to the ribosome with ~4° of inter‐subunit rotation and ~16° of 30S head swiveling (Fig EV4A).

Figure EV4. Two classes of EF‐G2‐ribosome structures with different angles of inter‐subunit rotation and 30S head swiveling.

• A
  Overview of Class 1 (left) and Class 2 (right) structures derived from workflow shown in Appendix Fig S1. The degrees of inter‐subunit rotation and 30S head swiveling (as compared to PDB: 7ST6) are shown below 30S subunit head (dark slate gray) and body (light slate gray). 50S subunit is shown in light blue, EF‐G2 in red and EF‐G2 insert in dark blue.
• B
  Close‐up view of the EF‐G2 insert region in domain IV in Class 1 (left) and Class 2 (right) structures. Same color scheme as in panel (A).

Our analysis revealed that EF‐G2 binds the ribosome at the same site used by the canonical EF‐G protein (Fig 5A). The overall five‐domain structure of E. coli EF‐G is preserved in EF‐G2 (Fig 5B and C), except for the 26‐amino acid insert in domain IV of EF‐G2 forming a protrusion consisting of a beta‐hairpin and an additional beta‐strand on the domain IV beta‐sheet (Fig 5D). On the non‐rotated ribosome (Class 1), the insert is near the 30S head (Fig EV4B), being closest to the C‐terminus of ribosomal protein S19, helix 31, and ~U1502 of 16S rRNA. (Please note that the resolution in this region is ~6–7 Å due to the dynamic nature of 30S head domain.)

Figure 5. Structure of the EF‐G2‐ribosome complex revealed by Cryo‐EM.

• A
  Overview of the EF‐G2‐ribosome complex structure. The ribosomal 50S subunit is in gray and the 30S subunit in light blue. The color of the five domains of EF‐G2 are as follows: blue—I, pink—II, orange—III, green—IV, light purple—V. The same color scheme for the EF‐G2 domains is used in other panels.
• B
  Overview of the five‐domain structure of the EF‐G2 protein.
• C
  Comparison of the EF‐G2 structure with that of E. coli EF‐G (PDB: 7N2V) in light gray. The structures are aligned by their domain I.
• D
  Comparison of domain IV of EF‐G2 and E. coli EF‐G (PDB: 7N2V) in light gray, highlighting the 26‐amino acid insert present in EF‐G2 (residues 514–539) and absent from EF‐G1 and E. coli EF‐G (Fig EV1).
• E
  Local density and model fitting of the resolved switch I and II regions and GTP in the ligand‐binding pocket of EF‐G2. The switch I (residues K37‐L67) region is colored orange, the switch II (D80‐V97) region blue, the P‐loop (G16‐T23) purple and the G4‐box (N134‐D137) region green. The same color scheme is used in other panels for EF‐G2.
• F
  Fitting GTP or GDP‐Pi in the density of the ligand showing GTP fits better. The models of T. thermophilus EF‐G2‐GTP (PDB: 2DY1) or E. coli EF‐G‐GDP‐Pi (PDB: 7SSL) were rigid‐body fitted in focused refined EF‐G2 map, and density corresponding to the ligand were extracted as shown in the upper and lower panels, respectively.
• G
  Superimposed structures of the GTP‐binding pocket including the switch I and II regions of EF‐G2 and the GTP‐bound conformation of E. coli EF‐G (PDB: 7N2V, light gray) and T. thermophilus EF‐G2 (PDB: 2DY1, dark gray).
• H, I
  Superimposed structures showing position of the ribosomal sarcin‐ricin loop (SRL, yellow) relative to EF‐G2 (colored) or GDPCP‐bound E. coli EF‐G on vacant ribosome (PDB: 4V9O) (dark gray) (H), and GDP‐bound E. coli EF‐G on chimeric translocation intermediate (PDB: 7SSD) (light gray) (I). Structures were aligned by the 23S rRNA. Red arrows indicate the rotation of EF‐G2 domains I and II relative to E. coli EF‐G structures. Same color scheme for EF‐G2 as in previous panels.
• J
  Superimposed structures showing position of the SRL (yellow) relative to the GTP binding pocket of EF‐G2 (colored) or GDPCP‐bound E. coli EF‐G on vacant ribosome (PDB: 4V9O) (light gray).
• K
  Distance (in Å) between SRL (Cα of A2662) and GTP‐binding pocket (Cα of T23 in the P‐loop and Pβ of the ligand GTP/analog/GDP/GDP‐Pi) in the EF‐G2 structures and in published structures of ribosome‐bound E. coli or T. thermophilus EF‐G with PDB identifiers indicated in parenthesis.

Ligand density consistent with a GTP can be seen in the GTP‐binding pocket of domain I in EF‐G2 (Fig 5E and F), with local resolution around 3.7 Å (Appendix Fig S2F). The EF‐G2 switch I and switch II regions are fully resolved and structured like those of canonical EF‐G and T. thermophilus EF‐G2 in their active (GTP/GDP‐Pi/GTP‐analog binding) forms (Fig 5G; Connell et al, 2007; Pulk & Cate, 2013; Tourigny et al, 2013; Zhou et al, 2013; Carbone et al, 2021; Petrychenko et al, 2021; Rundlet et al, 2021).

In the complex with the ribosome, domains I and II of EF‐G2 shift away from the sarcin‐ricin loop (SRL) of the 23S rRNA, adopting an orientation not previously observed in ribosomes complexed with canonical EF‐G proteins (Fig 5H and I; Gao et al, 2009; Pulk & Cate, 2013; Tourigny et al, 2013; Zhou et al, 2013; Carbone et al, 2021; Petrychenko et al, 2021; Rundlet et al, 2021). Meanwhile, the distance between the GTP‐binding pocket of EF‐G2 and the SRL in the 70S‐EF‐G2 complexes (local resolution ~3.7 Å, 11.3–12.4 Å between Cα of A2662 and Pβ of the ligand) is much longer than that in previously reported structures of canonical EF‐G‐ribosome complexes (8.8–10.3 Å between Cα of A2662 and Pβ of the ligand; Fig 5J and K), in which the SRL is closely packed with the GTP‐binding pocket and essential for GTPase activation and EF‐G binding (Clementi et al, 2010; Voorhees et al, 2010; Shi et al, 2012; Petrychenko et al, 2021). The GTP‐binding pocket is closely packed with the SRL even in the structures of the GTPase‐defective E. coli EF‐G (H91A) bound to the ribosome (Fig 5K; Li et al, 2015). As the release of the E. coli EF‐G domain I from the SRL initiates the dissociation of the factor from the ribosome (Carbone et al, 2021), the longer separation of EF‐G2 from the SRL may facilitate EF‐G2 clearance of the ribosome, thereby enabling turnover translation.

Domains III, IV, and V of EF‐G2 adopt an orientation and position similar to the corresponding domains of E. coli EF‐G (in the GTP‐analog‐bound form) on ribosomes without mRNA and tRNA (Fig 5H; Pulk & Cate, 2013), on ribosomal complexes without A‐site tRNA (Tourigny et al, 2013; Zhou et al, 2013), in GDP‐bound form on a post‐translocation ribosome (Gao et al, 2009), or in GDP‐bound form on translocation intermediates (Fig 5I) in which the peptidyl‐tRNA has moved to the ap/P state, vacating the A site on the 30S body domain (Carbone et al, 2021; Petrychenko et al, 2021; Rundlet et al, 2021), with domain IV of EF‐G2 reaching into the A site of the 30S body. The ap/P (also called chimeric) state is a bona‐fide intermediate of translocation, suggesting that EF‐G2 promotes tRNA–mRNA movement along a similar trajectory.

Cumulatively, this analysis revealed EF‐G2 in GTP‐bound conformation interacting with the ribosome but not in close contact with the SRL, a critical element of the GTPase center formed by translational GTPases and the ribosome.

EF‐G2 is expressed under starvation conditions and is the dominant EF‐G protein in B. thetaiotaomicron harvested from the murine gut

To understand the role of EF‐G2 in B. thetaiotaomicron's lifestyle, we compared its abundance to that of EF‐G1 in strains expressing epitope‐tagged versions of these proteins from their normal promoters and chromosomal locations. Western blot analysis of bacterial crude extracts revealed that EF‐G2 abundance is exceedingly low during exponential growth in complex medium but increases dramatically upon entry into stationary phase (Fig 6B–D). EF‐G2 also accumulated upon carbon starvation (Fig EV5B and C), reflecting the increase in the corresponding mRNA (Townsend II et al, 2020), but not in response to nitrogen starvation (Fig EV5B and C). By contrast, EF‐G1 is highly abundant during all phases of growth in complex medium, reaching its peak values during early exponential phase and gradually decreasing as the culture reaches stationary phase (Fig 6A, C and D). EF‐G1 abundance was also high in bacteria grown in minimal medium or upon a shift to carbon‐ or nitrogen‐starvation conditions, without a significant decrease within 60 min of carbon‐ or nitrogen‐starvation (Fig EV5A and C). In sum, EF‐G1 is highly abundant under all examined laboratory conditions, whereas EF‐G2 is not expressed under fast‐growing conditions and accumulates upon carbon starvation.

Figure 6. Differential expression of the EF‐G1 and EF‐G2 protein in laboratory medium and mouse cecum.

• A–D
  Western blot analysis of B. thetaiotaomicron strains expressing C‐terminally FLAG‐tagged EF‐G1 (WH405) (A), or EF‐G2 (GT1301) (B), from their normal promoters and chromosomal locations, grown in Tryptone Yeast Extract Glucose liquid medium (TYG) and sampled at the indicated times. Bacterial growth (OD₆₀₀) is shown in (C) and Western blot quantifications (FLAG signal normalized to loading control GroEL) are show in (D). Blots were developed with anti‐FLAG antibodies; anti‐GroEL antibodies were used as loading controls. Shown is a representative from at least two independent experiments.
• E, F
  Western blot of crude extracts from wild‐type B. thetaiotaomicron harvested from the ceca of mice (representative blot for two out of four mice tested) or corresponding to stationary phase cultures harvested following growth in TYG for 24 h or purified EF‐G1 and EF‐G2 proteins at the indicated amounts (E). Blots were developed with anti‐EF‐G1 or EF‐G2 polyclonal antibodies; anti‐GroEL antibodies were used as loading controls. Western blot quantifications of all four biological replicates and their average are shown in (F), where EF‐G1 and EF‐G2 abundances were estimated by normalizing the signal of bacterial sample to the closest signal of known amount of purified protein, and the derived absolute amount (pmol) was normalized to the signal of GroEL. Error bars represent SD. Please note log scale of y axis.

Figure EV5. EF‐G2 is highly abundant under carbon‐starvation conditions whereas EF‐G1 abundance stays consistently high during logarithmic growth and starvation conditions.

• A–C
  Western blot analysis of crude extracts prepared from (B). thetaiotaomicron strains encoding a C‐terminal FLAG‐tagged EF‐G1 (WH405) (A), or a C‐terminal FLAG‐tagged EF‐G2 (GT1301) (B), expressed from their normal promoters and chromosomal locations during exponential growth in minimal medium with glucose as carbon source (MM, time point: −5 min), and after shifted to the same minimal medium (MM) or to carbon‐ (No C) or nitrogen‐ (No N) limited minimal media for 15 or 60 min. Blots were developed with anti‐FLAG antibodies; anti‐GroEL antibodies were used as loading controls. A representative biological replicate from duplicates is shown. Western blot quantifications (FLAG signal normalized to GroEL signal) for the biological duplicates and their averages are shown in (C).

We next investigated the abundance of EF‐G1 and EF‐G2 in wild‐type B. thetaiotaomicron harvested from the ceca of mono‐colonized mice 4 days after oral gavage, using antisera raised against recombinant EF‐G1 and EF‐G2 proteins, which showed excellent specificity to their corresponding antigens and little cross‐reactivity (Appendix Fig S3). First, EF‐G1 was barely detected in bacteria harvested from the cecum (Fig 6E), which was unexpected given that EF‐G1 amounts are higher in bacteria grown in laboratory media (Fig 6A) and that the BT2729 gene encoding EF‐G1 is essential (Goodman et al, 2009). Second, EF‐G2 amounts were as high in bacteria harvested from the cecum (Fig 6E) as in bacteria from stationary phase laboratory cultures (Fig 6E and F). By normalizing the signals from bacteria harvested from mice to those of known amounts of purified EF‐G1 and EF‐G2, we estimate that EF‐G2 abundance surpasses EF‐G1's by at least 10‐fold in the murine gut (Fig 6F).

In sum, B. thetaiotaomicron increases EF‐G2 amounts while decreasing the abundance of the essential EF‐G1 in the murine cecum, suggesting that EF‐G2‐mediated protein synthesis is necessary for gut colonization.

Gut colonization by B. thetaiotaomicron requires EF‐G2's ability to support translation elongation

An EF‐G2 null mutant is defective in murine gut fitness (Townsend II et al, 2020). This defect appears to be largely due to EF‐G2's role in protein synthesis because a B. thetaiotaomicron strain harboring the EF‐G2 (H593K) variant inactive in protein synthesis (Fig 2A) was as defective for gut colonization as an isogenic EF‐G2 null mutant (Fig 7): when the four isogenic B. thetaiotaomicron strains were inoculated in germ‐free mice at nearly identical amounts, the BT2167 null mutant (lacking EF‐G2) and the BT2167 null mutant complemented in trans with a mutant BT2167 gene specifying the variant EF‐G2 (H593K) were out‐competed by wild‐type B. thetaiotaomicron and the isogenic BT2167 null mutant complemented with the wild‐type BT2167 gene after 1 week of colonization (Fig 7). This result argues that EF‐G2's role in translocation is required for B. thetaiotaomicron fitness in the murine gut.

Figure 7. EF‐G2‐mediated protein synthesis is required for murine gut colonization.

Abundances of barcoded wild‐type (GT478) and three isogenic B. thetaiotaomicron strains: one lacking the EF‐G2‐specifying gene BT2167 (∆EF‐G2, WH148), one lacking BT2167 and expressing the wild‐type BT2167 gene in trans (∆EF‐G2 + EFG2, WH160), and one lacking BT2167 and expressing a variant of BT2167 specifying an EF‐G2 with the H593K substitution (∆EF‐G2 + EFG2 (H593K), WH514) in gnotobiotic mice (N = 5) at the indicated times following gavage. Values for day 0 are calculated using the number of colony‐forming units (CFU) in the inoculum, and for day 1–14 are based on qPCR measurement of each barcode in mouse fecal sample. See Materials and Methods for details. Mean values and SD of five biological replicates are shown.

Discussion

In all kingdoms of life, protein synthesis requires an essential GTPase to promote ribosome translocation. Designated EF‐G in E. coli, this canonical GTPase hydrolyzes GTP on the ribosome, accelerating ribosome translocation by four orders of magnitude during protein synthesis (Noller et al, 2017; Rodnina et al, 2019). We have now identified a natural paralog of this GTPase that promotes slow ribosome translocation and supports slow protein synthesis (Fig 2) without hydrolyzing GTP (Figs 2B and 3). Designated EF‐G2 in B. thetaiotaomicron, this protein binds GTP (Figs 3D and 5E) but lacks the ribosome‐stimulated GTPase activity (Figs 3A and EV3A) of canonical EF‐G proteins (Rodnina et al, 1997).

Specifically induced under carbon starvation laboratory conditions (Figs 6B and EV5B; Townsend II et al, 2020), EF‐G2 enables the prominent commensal B. thetaiotaomicron to colonize the mammalian gut by supporting essential protein synthesis (Figs 2A and 7) in a slow but energy‐saving manner (Fig 3A). B. thetaiotaomicron also harbors EF‐G1, an ortholog of E. coli EF‐G that is highly abundant under all tested laboratory conditions (Figs 6A and EV5A) but largely reduced in the murine gut, being surpassed by EF‐G2 by > 10‐fold (Fig 6F). These findings suggest that paralogous EF‐Gs enable B. thetaiotaomicron to switch between rapid and energy‐efficient modes of protein synthesis in nutrient‐fluctuating environments.

EF‐G2 supports translation elongation without hydrolyzing GTP

GTP‐hydrolysis by canonical EF‐G accelerates ribosome translocation and allows the GDP‐bound form to dissociate from the ribosome (Inoue‐Yokosawa et al, 1974; Belitsina et al, 1975, 1976; Kaziro, 1978; Rodnina et al, 1997; Cunha et al, 2013; Belardinelli et al, 2016). When GTP hydrolysis is blocked (e.g., by a non‐hydrolysable GTP analog or mutation of a catalytic His91 residue in the protein), EF‐G is trapped on the ribosome following a single round of slow ribosome translocation (Inoue‐Yokosawa et al, 1974; Cunha et al, 2013; Salsi et al, 2016), thereby blocking subsequent binding of aminoacyl‐tRNA.

Naturally lacking GTPase activity (Figs 3A and EV3A), B. thetaiotaomicron EF‐G2 promotes slower ribosome translocation than canonical EF‐G proteins (Figs 2B and EV2A), with kinetics similar to translocation mediated by EF‐G1 with a non‐hydrolysable GTP‐analog (Fig 2B), reminiscent of the GTPase‐deficient E. coli EF‐G (H91A) variant (Cunha et al, 2013; Holtkamp et al, 2014). However, unlike the latter variant, EF‐G2 dissociates spontaneously from the ribosome after promoting translocation, allowing protein synthesis to continue (Fig 2A and C). The ability of EF‐G2 to dissociate from the ribosome appears to require the 26‐amino acid insert in domain IV (Fig 4D), which is exclusively found in Bacteroidia (Figs 1A and EV1B; Margus et al, 2011). EF‐G2 proteins from bacteria outside the Bacteroidetes phylum lack this insert, while EF‐G2 proteins from Bacteroidetes outside the Bacteroidia class have shorter (17‐amino acid) inserts lacking the hairpin region critical for EF‐G2 dissociation (Fig 4). Therefore, the property of mediating turnover translocation without GTP hydrolysis is probably limited to EF‐G2 members of the Bacteroides genus and their close relatives.

We propose that EF‐G2 binds the ribosome and promotes translocation when bound to GTP because: (i) EF‐G2 binds GTP with a micromolar affinity (Fig 3D) and has a higher affinity for GTP than GDP (Fig EV3D). Thus, EF‐G2 should be in GTP‐bound form in cells because the cellular GTP concentration is at millimolar level and usually higher than the GDP concentration. (Please note that the GTP/GDP ratio does not change significantly before or after 1 h of carbon starvation in B. thetaiotaomicron (Schofield et al, 2018).) (ii) Density of a ligand consistent with GTP is visualized in the cryo‐EM structure of EF‐G2 bound to the ribosome (Fig 5E and F). (iii) There are similarities between the structure of EF‐G2 when bound to the vacant ribosome and that of the canonical EF‐G bound to ribosomes without tRNA or mRNA or to ribosome complexes in the intermediate state of translocation (Fig 5H and I). Therefore, the EF‐G2 structure resembles those of canonical EF‐G proteins that promote translocation when bound to GTP, GDP‐Pi, or non‐hydrolysable GTP‐analogs but not to GDP or in ligand‐free form (Inoue‐Yokosawa et al, 1974; Belitsina et al, 1975, 1976; Rodnina et al, 1997; Salsi et al, 2016).

Ribosome recycling strictly requires EF‐G‐dependent GTP hydrolysis and phosphate release (Savelsbergh et al, 2009). This contrasts with EF‐G‐stimulated ribosome translocation, which is still supported by EF‐G in the absence of GTP hydrolysis (Inoue‐Yokosawa et al, 1974; Rodnina et al, 1997; Cunha et al, 2013; Salsi et al, 2016). In agreement with this notion, both EF‐G1 and EF‐G2 support translocation in E. coli ribosomes (Fig 2), but only the GTPase‐competent EF‐G1 exhibited ribosome recycling activity (Fig EV2D); EF‐G2 was inactive in ribosome disassembly and actually stabilized the ribosomal complex (Fig EV2D). The inability of EF‐G2 to support ribosome recycling may also explain the essentiality of the EF‐G1‐encoding BT2729 gene (Goodman et al, 2009).

Colonization of the mammalian gut requires energy‐efficient protein synthesis

The mammalian gut environment is rich in dietary and host‐derived carbohydrates (or glycans) that provide sufficient carbon and energy to support over 10¹¹ microbial cells per milliliter (Donaldson et al, 2016; Sender et al, 2016). However, systems that respond to nutrient limitation in vitro are critical for intestinal colonization (Schofield et al, 2018; Townsend II et al, 2020), indicating that microbes experience nutrient limitation in the gut, likely due to fluctuations in available nutrients and competition with other gut residents. The prevalent commensal B. thetaiotaomicron requires the alarmone (p)ppGpp to colonize the gut and produces (p)ppGpp in response to carbon starvation (Schofield et al, 2018) rather than amino acid limitation, as enteric bacteria do (Cashel & Gallant, 1969; Haseltine et al, 1972). Moreover, carbon starvation activates transcription factor BT4338, a master regulator of carbohydrate utilization and gut colonization (Wu et al, 2015; Schwalm III et al, 2016; Townsend II et al, 2020), which activates transcription of the EF‐G2‐specifying BT2167 gene > 100‐fold and of genes involved in energy generation (Townsend II et al, 2020).

EF‐G2 appears to be largely responsible for protein synthesis some of the time B. thetaiotaomicron is in the mammalian gut because EF‐G2 is > 10‐fold more abundant than EF‐G1 in bacteria recovered from the mouse cecum (Fig 6E and F) and because the ability of EF‐G2 to support protein synthesis (Fig 2A) is necessary for B. thetaiotaomicron fitness in the gut (Fig 7). Although slower than EF‐G1‐mediated protein synthesis (Fig 2A), EF‐G2‐mediated protein synthesis confers the unique advantage of not consuming GTP at each ribosome translocation step (Figs 2C and 3A), unlike protein synthesis supported by canonical EF‐G (Rodnina et al, 1997). That is, by employing EF‐G2, B. thetaiotaomicron saves energy at the costliest step of protein synthesis. In addition, EF‐G1, like canonical EF‐G, can hydrolyze GTP non‐productively when interacting with ribosomes not engaged in translation (Fig EV3A), unlike EF‐Tu, the other factor involved in translation elongation, that stimulates GTP hydrolysis only when engaged in translation. That EF‐G2 abundance is 10‐fold higher than EF‐G1's (Fig 6E and F) would also prevent unnecessary energy (GTP) consumption upon EF‐G1 binding to non‐translating vacant ribosomes (Fig EV3A), which may accumulate during carbon starvation (Li et al, 2018).

Concluding remarks

Gene duplication plays a key role in the evolution of new cellular abilities because it provides the raw material for mutation and selection as well as redundant genes with reduced functional constraints (Zhang, 2003). Thus, random mutations in one of the copies of the duplicated gene, being under relaxed purifying selection, can change the function of that gene, providing an advantage under particular environmental conditions or genetic backgrounds (Dykhuizen & Hartl, 1980; Kimura, 1983).

We have now uncovered a novel function of an ancient gene duplicate in a mammalian gut commensal. We determined that the EF‐G2‐specifying paralog of the essential EF‐G‐specifying gene encodes a starvation‐induced translation factor that mediates energy‐saving slow protein synthesis, in contrast to that mediated by the GTP‐hydrolyzing ancestral EF‐G1. Significantly, a single amino acid substitution in EF‐G2 can improve its protein synthesis activity (Fig 4E) but has not been selected for in B. thetaiotaomicron, suggesting that a reduction in the energetic costs of protein synthesis is required for the organism's lifestyle. Moreover, deploying EF‐G2 under starvation conditions may provide the additional benefit of preventing ribosome collision and unnecessary abortion of translation (Subramaniam et al, 2014; Saito et al, 2022) by matching translation elongation speed to the reduced aminoacyl‐tRNA substrate pool present under nutrient‐limited conditions.

Finally, Bacteroides incorporated the energy‐efficient EF‐G2‐specifying paralog into the regulon of the master regulator of carbohydrate utilization, energy generation, and gut colonization that is activated in response to carbon starvation (Townsend II et al, 2020). This evolutionary event enables bacteria to calibrate both metabolism and protein synthesis to the availability of nutrient and energy resources in the gut.

Materials and Methods

Reagents and resources

Details about key reagent, resources, software, and tools used in this work are provided in Appendix Table S1.

Bacterial strains and growth conditions

Bacteroides thetaiotaomicron VPI‐5482 tdk (Koropatkin et al, 2008) and derived strains (Appendix Table S1) were cultured anaerobically at 37°C in liquid Tryptone Yeast Extract Glucose (TYG) medium (Holdeman et al, 1977), glucose minimal medium (MM) [100 mM KH₂PO₄ (pH 7.2), 15 mM NaCl, 8.5 mM (NH₄)₂SO₄, 0.5 μg ml⁻¹ L‐cysteine, 1.9 μM hematin, 200 μM L‐histidine, 100 μM MgCl₂, 1.4 μM FeSO₄, 50 μM CaCl₂, 1 μg ml⁻¹ vitamin K₃, 5 ng ml⁻¹ vitamin B₁₂, plus 0.5% (wt/v) glucose] (Martens et al, 2008), or on Brain Heart Infusion Agar containing 5% defibrinated horse blood. When appropriate, antibiotics were added at the following final concentrations: tetracycline 2 μg ml⁻¹, erythromycin 10 μg ml⁻¹, or gentamicin 200 μg ml⁻¹. An anaerobic chamber (Coy Laboratory Products) containing 20% CO₂, 10% H₂, and 70% N₂ was used for all anaerobic microbiology procedures. E. coli S17‐1 λpir and E. coli BL21(DE3) were cultured at 37°C in Luria Bertani broth (BD), containing 100 μg ml⁻¹ ampicillin when appropriate.

Gnotobiotic animal experiments

All experiments using mice were performed using protocols approved by the Yale University Institutional Animal Care and Use Committee. Germ‐free C57BL/6J mice were maintained in flexible plastic gnotobiotic isolators with a 12‐h light/dark cycle. Individually caged animals (N = 5 per group, littermates of mixed sex were randomly assigned to experimental groups) were fed a standard, autoclaved mouse chow (5 K67 LabDiet, Purina) ad libitum. Mice were 12–16 weeks of age at the time of gavage. Within a given experiment, mice were age‐matched within 2 weeks.

Construction of strains and plasmids

Genetic engineering methods

DNA purification, PCR, and cloning were performed using standard methods. All strains and plasmids (listed in Appendix Table S1) were constructed using oligonucleotides with sequences provided in Appendix Table S2. All B. thetaiotaomicron strains were derived from strain VPI‐5482 tdk (Koropatkin et al, 2008). Introduction of plasmid into B. thetaiotaomicron was achieved by conjugation with E. coli S17‐1 λpir harboring the corresponding plasmids. All plasmids were verified by Sanger sequencing across the insert before their introduction into B. thetaiotaomicron. Plasmids derived from pNBU2‐tetQ and oligonucleotide barcodes encoded pNBU2‐tetQ vectors were introduced into the B. thetaiotaomicron genome (NBU2 att‐1 site) in single copy as described (Martens et al, 2008). Introduction of the pKNOCK‐tetQ suicide vector into B. thetaiotaomicron genome by homologous recombination was carried out as described (Raghavan et al, 2014). In‐frame, unmarked, nonpolar deletions were generated using a counter‐selectable allelic exchange procedure as described (Koropatkin et al, 2008) and confirmed by PCR and Sanger sequencing across the chromosomal region of interest.

Construction of a strain specifying BT2729 with a C‐terminal tag from its native chromosomal location

We used plasmid pKNOCK‐tetQ, cloning the 750 bp sequence at the 3′ end of the target gene including additional nucleotide sequences encoding the FLAG‐tag and a stop codon as described (Townsend II et al, 2020), except that the BT2731 promoter (p _BT2731 , which transcribes the BT2729 operon) was added in front of the 750 bp sequence, so that after plasmid integration, the genes downstream of BT2729 were transcribed from this promoter.

Preparation of proteins and ribosomes

Protein overexpression and purification

The BT2729, BT2167, and BT2167(H593K) genes without the corresponding stop codons were PCR amplified using oligonucleotides with sequences provided in the Appendix Table S2 and cloned (separately) between the NdeI and XhoI sites of plasmid pET22b(+) by Gibson assembly using the NEBuilder Assembly Master Mix (NEB). Thus, the resulting genes harbor at the 3′‐end a DNA sequence encoding an in‐frame C‐terminal His tag. The nucleotide sequences of the cloned DNAs were confirmed by Sanger sequencing and the corresponding plasmids were individually transformed into E. coli strain BL21 (DE3).

For protein overexpression, bacteria were grown at 37°C in LB medium supplemented with 100 μg ml⁻¹ ampicillin. Gene expression was induced by addition of isopropyl b‐D‐1‐thiogalactopyranoside (IPTG; 1 mM final concentration), and cultures were further grown for 3 h. Cells were harvested and washed in Tris‐buffered saline (TBS ‐ 50 mM Tris, 138 mM NaCl, 2.7 mM KCl; pH 8.0), and stored at −80°C until protein purification. Cells were lysed by re‐suspension in lysis buffer containing 20 mM Tris–HCl, pH 7.4, 300 mM NaCl, 15% glycerol, 1× BugBuster^® (Millipore), 1 mg ml⁻¹ lysozyme, 25 U ml⁻¹ Benzonase^® Nuclease (Sigma) and 1× cOmplete™ protease inhibitor (Roche) and incubation on ice for 15 min. The lysate was centrifuged at 7,000 × g at 4°C for 30 min. The supernatant was applied to a column containing TALON® metal affinity resin (TaKaRa) for affinity purification of the His‐tagged proteins. The column was washed with buffer containing 20 mM Tris–HCl, pH 7.4, 300 mM NaCl and 15% glycerol, and proteins were eluted using the same buffer now containing 250 mM imidazole. The eluted protein was concentrated and buffer‐changed to 2× TAKM₇ buffer [100 mM Tris–HCl, pH 7.5, 140 mM NH₄Cl, 60 mM KCl, 14 mM MgCl₂] by centrifugal ultrafiltration using Amicon Ultra‐15 Centrifugal Filters with MW cutoff of 50 K (Millipore); one volume of glycerol was added before protein storage at −20°C.

Preparation of crude B. thetaiotaomicron ribosomes

Wild‐type B. thetaiotaomicron (GT23) was grown in TYG medium to OD₆₀₀ = ~0.6, harvested by centrifugation and washed in cold Ribosome Extraction Buffer [20 mM Tris–HCl, pH 7.5, 50 mM Mg(OAc)₂, 100 mM NH₄Cl, 1.0 mM DTT, 0.5 mM ethylenediaminetetraacetic acid (EDTA)] (Rivera et al, 2015). The cell pellet was resuspended in Ribosome Extraction Buffer supplemented with 100 μl 10× BugBuster^® (Millipore), 1 mg ml⁻¹ lysozyme, 1,000 U·ml⁻¹ SUPERase·In™ and 10 U ml⁻¹ RNase‐free DNase I (Roche) and lysed by three cycles of freeze–thaw. The lysate was cleared of cell debris by centrifugation at 20,000 × g at 4°C for 30 min. The supernatant was transferred to ultracentrifuge compatible tubes and centrifuged at 100,000 × g at 4°C for 1 h to pellet the ribosomes. After removing the supernatant, the crude ribosome pellet was resuspended in buffer containing 20 mM Tris–HCl (pH 7.5), 10 mM Mg(OAc)₂, 10 mM KCl and 20 mM NH₄Cl and stored at −80°C. The prepared crude ribosome had poor rRNA integrity likely due to endogenous B. thetaiotaomicron RNases. Despite its low in vitro protein synthesis activity, the ribosome preparation stimulated GTP hydrolysis by EF‐G1 (Fig 2B).

Components of the translation machinery

E. coli ribosomes, f[³H]Met‐tRNA^fMet, [¹⁴C]Phe‐tRNA^Phe, [¹⁴C]Lys‐tRNA^Lys, Phe‐tRNA^Phe, initiation factors, EF‐Tu and E. coli EF‐G were prepared as described (Peng et al, 2019). mRNAs were synthesized by IBA (Goettingen, Germany). The following mRNA sequences were used (start codons are in bold face): mRNA(MKF): 5′‐GUUAACAGGUAUACAUACU AUG AAAUUCAUUAC‐3′; mRNA(MF): 5′‐GUUAACAGGUAUACAUACU AUG UUUGUUAUUAC‐3′.

Ribosome complexes

To prepare initiation complexes (IC), 70S ribosomes were incubated with a 2‐fold excess of mRNA, 1.7‐fold excess of initiation factors, 3‐fold excess of f[³H]Met‐tRNA^fMet, and 1 mM GTP in TAKM₇ buffer (50 mM Tris–HCl pH 7.5 at 37°C, 70 mM NH₄Cl, 30 mM KCl and 7 mM MgCl₂) at 37°C for 30 min. Ternary complexes (TC) were prepared by incubating EF‐Tu (3‐fold excess over tRNA) with 1 mM GTP, 3 mM phosphoenolpyruvate, 0.5% pyruvate kinase in TAKM₇ buffer at 37°C for 15 min and subsequent addition of aminoacyl‐tRNAs cognate to the mRNA coding sequence. Pre‐translocation complex (PRE) was formed by mixing IC and TC (2–5‐fold excess over IC). Purification of IC and PRE were performed by centrifugation through a 1.1 M sucrose cushion in TAKM₂₁ buffer (50 mM Tris–HCl pH 7.5 at 37°C, 70 mM NH₄Cl, 30 mM KCl and 21 mM MgCl₂). Pellets were dissolved in TAKM₂₁ buffer and the concentration of purified complex was determined by filtration through a nitrocellulose membrane (Peng et al, 2019). The magnesium concentration of purified complexes was adjusted to working (7 mM) concentration before use. Experiments were carried out in TAKM₇ unless stated otherwise.

Tripeptide synthesis assays

Tripeptide formation was examined by incubation of initiation complex (IC) programmed with mRNA(MKF) (0.1 μM) with TC(Phe‐tRNA^Phe, [¹⁴C]Lys‐tRNA^Lys; 0.2 μM each) and EF‐G (2 μM) in TAKM₇ buffer at 37°C. Samples were taken at 1 and 5 min and quenched by addition of KOH (0.5 μM). Products released from the ribosomes following incubation at 37°C for 30 min were neutralized with glacial acetic acid and analyzed by reversed‐phase HPLC (Chromolith®RP‐8 e column, Merck) using a 0–65% acetonitrile gradient in 0.1% trifluoroacetic acid. The amounts of fMKF tripeptide was quantified by [³H]Met and [¹⁴C]Lys radioactivity counting.

Stopped‐Flow translocation experiments

Translocation was measured as described (Peng et al, 2019). Pre‐translocation complexes programmed with mRNA (mMF) labeled with fluoresceine attached at position + 14 (MF+14Flu; 0.05 μM) and purified by ultracentrifugation through 1.1 M sucrose cushion and mixed with different EF‐G proteins (2 μM) along with GTP (200 μM), GDPNP (400 μM) or 200 μM GTP plus 200 μM ppGpp in TAKM₇ buffer in a stopped‐flow apparatus (SX 20, Applied Photophysics) at 37°C (all concentrations after mixing). Fluorescence was excited at 470 nm and emission detected after passing a KV500 cut‐off filter. Changes in fluorescence were recorded with time and 5–7 technical replicates were averaged (n = 5–7). Date was evaluated by double exponential fitting using TableCurve software (Systat Software Inc).

Ribosome recycling experiments

Subunit splitting was measured as described (Savelsbergh et al, 2009). Vacant ribosomes (0.05 μM) were rapidly mixed with EF‐G (2 μM), RRF (5 μM), IF3 (1 μM), and GTP (1 mM) in TAKM₇ buffer in a stopped‐flow apparatus (SX 20, Applied Photophysics) at 37°C and light‐scattering was monitored at 430 nm (all concentrations after mixing).

Factor‐ribosome co‐sedimentation assays

Co‐sedimentation assays were performed as described (Cunha et al, 2013; Wieland et al, 2022). Specifically, 0.4 μM of purified E. coli 70S were mixed with 1 μM indicated EF‐G protein, 500 μM GTP in TAKM₇ buffer (50 mM Tris–HCl pH 7.5 at 37°C, 70 mM NH₄Cl, 30 mM KCl and 7 mM MgCl₂). Reactions were incubated for 5 min at 20°C, and subsequently loaded on a 10% (w/v) sucrose cushion in TAKM₇ and centrifuged for 35 min at 212,911 x g in a TLA100 rotor at 4°C. The pre‐centrifugation reactions and 5 pmol of the ribosome pellet (quantified by A260), were applied to NuPAGE 4–12% Bis‐Tris protein gel (ThermoFisher) and fractionated at 180 V in 1× MOPS running buffer (ThermoFisher) for 45 min and stained with GelCode Blue Stain (ThermoFisher).

GTP binding assays

Differential radial capillary action of ligand assay (DraCALA; Roelofs et al, 2011) was used to detect GTP binding by the EF‐G1 and EF‐G2 proteins. Purified proteins of specified concentrations (0.02–25 μM) in binding buffer [50 mM Tris–HCl, (pH 7.5), 10 mM MgCl₂, 70 mM NH₄Cl, 1 mM dithiothreitol (DTT)] were mixed with ∼1.6 nM [α‐³²P]‐labeled GTP and incubated at room temperature for 5 min before spotting 2.5 μl on a dry nitrocellulose membrane (General Electric), where protein and bound ligand are immobilized at the site of contact (forming an “darker” inner circle in the final radiograph, if ligand is bound to protein), whereas free ligand is mobilized by capillary action with the liquid phase (forming a larger outer circle; Roelofs et al, 2011). For competition assays, cold nucleotides (GTP or GDP) of specified concentrations (1 mM, 100 μM and 10 μM) were added in the initial mixture containing 25 μM of EF‐G1 or EF‐G2 and ∼1.6 nM [α‐³²P]‐labeled GTP. Spots were air‐dried, and radioactivity signals were detected using a Typhoon FLA9000 PhosphorImager. The fraction of ligand bound was calculated as described (Roelofs et al, 2011) using the areas and signal intensities of the inner circle (containing bound ligand) and of the outer circle (entire sample) quantified by the ImageJ software. Data were fitted and dissociation constant K_d was estimated using the one site‐specific binding model in GraphPad Prism version 9.3.1.

GTPase assays

Ribosome‐stimulated GTPase activity assays

Multiple turnover GTPase activity of EF‐G proteins was investigated by incubating vacant E. coli 70S ribosomes or pre‐translocation complexes programmed with mRNA(MF) (0.5 μM) and EF‐G (1 μM) together with 1 mM GTP with a trace amount of [γ‐³²P] GTP at room temperature. At time points indicated in the figures, reactions were quenched by adding the same volume of 40% formic acid. Samples were analyzed by thin‐layer chromatography (Polygram CEL 300, Macherey‐Nagel) using 0.5 M potassium phosphate (pH 3.5) as mobile phase. Radioactivity was detected using the phosphor screen and analyzed by phosphorimager (Peng et al, 2019). For GTPase competition experiments, EF‐G1 and EF‐G2 were used at the concentrations indicated in the figure legend, and reactions were incubated at room temperature for 5 min, quenched and analyzed.

B. thetaiotaomicron ribosome‐stimulated and intrinsic GTPase activity assays

GTPase assays were carried out as described (Palmer et al, 2013) in 50 μl reactions containing: 50 mM Tris–HCl, (pH 7.5), 10 mM MgCl₂, 70 mM NH₄Cl, 1 mM dithiothreitol (DTT) and 1.8 mM GTP. Different concentrations of the EF‐G1 or EF‐G2 (between 0 and 0.05 μM) were tested for both intrinsic GTPase activity and vacant ribosome‐dependent GTPase activity upon addition of E. coli ribosomes (0.2 μM). To test the ability of B. thetaiotaomicron ribosomes to stimulate the GTPase activity, reactions were carried out with EF‐G1 or EF‐G2 (0.2 μM) and crude B. thetaiotaomicron ribosomes (0.1 μM as estimated by A260). When testing EF‐G1 and EF‐G2 variants, we used 0.05 μM protein and 0.2 μM E. coli ribosomes. Incubations were carried out at 37°C for 30 min and reactions were stopped by addition of 150 μl of 50 mM EDTA. The amount of GTP hydrolyzed was determined by measuring the amount of the inorganic phosphate liberated using a colorimetric GTPase assay kit (Novus Biologicals) per the manufacturer's directions.

In vitro protein synthesis assays

In vitro protein synthesis

In vitro protein synthesis was carried out using a custom‐made PURExpress® system (New England Biolabs) lacking the E. coli EF‐G protein, which allowed us to test the behavior of the purified E. coli EF‐G, B. thetaiotaomicron EF‐G1 and EF‐G2 proteins, and engineered variants. As template, we used linear DNA fragments corresponding to a gene encoding the HslO‐FLAG protein or a gene encoding the His‐FLAG protein driven by a T7 promoter, generated by PCR using primers listed in Appendix Table S2 with genomic DNA from Salmonella enterica serovar Typhimurium strain 14,028 as template (Gao et al, 2019). Reaction mixtures taken at different times were added to NuPAGE™ LDS sample buffer (ThermoFisher) containing 100 mM dithiothreitol (DTT) and kept on ice until heated for SDS–PAGE.

Western blot analyses of in vitro synthesized proteins

Samples were heated at 95°C for 5 min, loaded onto a NuPAGE 4–12% Bis‐Tris protein gel (ThermoFisher), and fractionated at 180 V in 1× MOPS running buffer (ThermoFisher) for 60 min. Fractionated proteins were transferred to a nitrocellulose membrane using an iBlot device (Invitrogen) and the resulting membrane was blocked in TBS containing 3% skim milk for 1 h. The FLAG‐tagged proteins were detected using a 1:5,000 dilution of mouse anti‐FLAG antibody (Sigma) followed by a 1:5,000 dilution of an HRP‐conjugated anti‐mouse antibody (Promega). Between primary and secondary antibody incubations, and after secondary antibody incubation, membranes were washed with TBS containing 0.05% Tween‐20 and rinsed with TBS. Blots were developed with SuperSignal West Femto Maximum Sensitivity Substrate (ThermoFisher) and imaged using a LAS‐4000 imager (General Electric).

Polysome profiling and fraction analysis

Polysome profiling was carried out as reported (Becker et al, 2013). Briefly, B. thetaiotaomicron strain WH407 (encoding EF‐G1‐FLAG and EF‐G2‐HA) was grown in TYG to OD₆₀₀ = ~0.45, when 100 μg ml⁻¹ chloramphenicol was added. Cultures were shaken for 1 min and poured onto crushed chloramphenicol‐ice (ice made from ultrapure water with 100 μg ml⁻¹ chloramphenicol). Cells were collected by centrifugation at 7,000 × g at 4°C for 5 min (Aretakis et al, 2018). (The chloramphenicol and centrifugation method was chosen because the B. thetaiotaomicron culture rapidly clogged the membrane filter used for ribosome profiling experiments with Bacillus subtilis or E. coli cells.) Cell pellets were washed in lysis buffer once followed by centrifugation at 7,000 × g at 4°C for 5 min and resuspended in lysis buffer before added 100 U ml⁻¹ RNase‐free DNase I (Roche). The cell suspension was slowly dripped into and frozen in liquid nitrogen, stored at −80°C before lysis. Cells were lysed by cryogenic milling in liquid nitrogen chilled stainless‐steel jars using a mixer mill (Retsch, MM400) for five times at 15 Hz for 3 min each time. Jars were chilled in liquid nitrogen in between. Pulverized cells were kept frozen at −80°C before sucrose gradient ultracentrifugation analysis. The lysate was thawed on ice, centrifuged at 20,000 × g at 4°C for 10 min to remove insoluble cell debris. The cleared cell lysate (~16 unit of A260) was layered on top of a 10–40% sucrose gradient and subjected to ultracentrifugation at 209,490 x g at 4°C for 3 h in a SW41 rotor (Beckman). The gradient was fractionated and the polysome profile (A260) was recorded using a density gradient fractionation system with a UV detector (BRANDEL). Eleven fractions total were collected from each gradient, and proteins from each fraction were extracted using TCA precipitation and analyzed by Western blotting using antibodies directed to the FLAG and HA tags.

Subjecting B. thetaiotaomicron to carbon or nitrogen limitation

B. thetaiotaomicron strains were grown in TYG medium anaerobically overnight before being sub‐cultured into glucose minimal media (MM; see Bacterial Strains and Growth Conditions). The resulting stationary phase culture was diluted 1:50 into identical pre‐reduced medium and grown to mid‐exponential phase (OD ~0.5) at which time an aliquot was collected by centrifugation at 7,000 × g at room temperature for 1 min. After decanting, the cell pellet was immediate placed on dry ice until the end of the experiment (samples denoted “‐5 min"). The remaining culture was centrifuged at 7,000 × g at room temperature for 3 min in sealed tubes, re‐introduced into the anaerobic chamber where the supernatants were decanted. Cell pellets were resuspended in an equivalent volume of pre‐warmed, pre‐reduced MM or MM without glucose or (NH₄)₂SO₄ (“No C” or “No N” samples in the figure, respectively) and incubated at 37°C anaerobically (Townsend II et al, 2020). Following incubation for 15 and 60 min, aliquots were collected by centrifugation, the supernatant was decanted before the pellet was placed on dry ice until storage at −80°C.

Western blot for bacterial extracts

Whole cell lysate preparation

Frozen pellets from 10 ml of exponential phase cells grown in minimal media or equivalent amounts of TYG‐cultured cells from different growth stages were used for whole cell lysate preparation. Pellets were thawed and lysed in 300 μl Tris‐buffered saline (TBS – 50 mM Tris, 138 mM NaCl, 2.7 mM KCl; pH 8.0) containing 1× BugBuster® (Millipore), 1 mM EDTA, 0.5 mg ml⁻¹ chicken egg lysozyme and 25 U·ml⁻¹ Benzonase® Nuclease (Sigma), by incubating with constant nutating at room temperature for 20 min. Samples were centrifuged at 20,000 × g at 4°C for 10 min to remove cell debris. Protein concentrations were estimated by measuring absorbance at 280 nm using a Nanodrop 8000 (ThermoFisher).

Western blot analysis to detect FLAG‐tagged EF‐G1 or EF‐G2 proteins

A volume corresponding to 100 μg of protein from each sample was combined with 5 μl of 4× LDS Buffer (ThermoFisher) containing 100 mM dithiothreitol and subjected to heating at 95°C for 5 min. Samples were loaded onto a NuPAGE 4–12% Bis‐Tris protein gel (ThermoFisher) and fractionated at 180 V in 1× MOPS running buffer (ThermoFisher) for 60 min. Fractionated proteins were transferred to a nitrocellulose membrane using an iBlot device (Invitrogen) and the resulting membrane was cut below the 65 kDa marker and both portions blocked in TBS containing 3% skim milk for 1 h. FLAG‐tagged EF‐G2 was detected on the top portion of the membrane using a 1:5,000 dilution of a mouse anti‐FLAG antibody (Sigma) followed by incubation with a 1:5,000 dilution of an HRP‐conjugated anti‐mouse antibody (Promega). GroEL was detected on the bottom portion of the membrane using a 1:5,000 dilution of rabbit anti‐GroEL antibody (Sigma) followed by incubation with a 1:5,000 dilution of an HRP‐conjugated anti‐rabbit antibody (General Electric). Between primary and secondary antibody incubations, and after secondary antibody incubation, membranes were washed with TBS containing 0.05% Tween‐20 and rinsed with TBS. Blots were developed with SuperSignal West Femto Maximum Sensitivity Substrate (ThermoFisher) and imaged using a LAS‐4000 imager (General Electric).

Mouse gut colonization experiments

Inoculum preparation

Overnight cultures of barcoded B. thetaiotaomicron strains grown separately in TYG medium anaerobically overnight were diluted 1:50 into fresh TYG and cultured individually for 8 h. Bacteria were collected by centrifugation and resuspended in TYG containing 20% glycerol and divided into identical aliquots before storage at −80°C. Density of each strain stock was measured by thawing one aliquot and plating for CFUs after serial dilution.

Mono‐colonization experiments

10⁸ CFUs were suspended in 200 μl phosphate‐buffered saline and administered to each animal by oral gavage. Mice were euthanized 4 days after gavage and cecal contents were collected as ~100 mg aliquots, flash frozen in liquid nitrogen, and stored at −80°C before RNA or protein extraction.

In vivo competition experiments

10⁸ CFUs of each strain (wild‐type, GT478; ΔBT2167, WH148; ΔBT2167 + BT2167, WH160; ΔBT2167 + BT2167(H593K), WH514) were combined and suspended in 200 μl phosphate‐buffered saline before administering to each animal by oral gavage. Input (day 0) abundance of each strain was determined by plating and counting the CFUs of each strain stock after serial dilution. Fecal pellets were collected on the indicated days and genomic DNA was extracted as described (Martens et al, 2008). The abundance of each strain was measured by qPCR, using barcode‐specific primers (wild‐type, primers W1701 and W1713; ΔBT2167, primers W1702 and W1713; ΔBT2167 + BT2167, primers W1712 and W1713; or ΔBT2167 + BT2167(H593K), primers W1711 and W1713) as described (Martens et al, 2008).

Western blot for bacteria from murine gut

Cecal bacterial lysate preparation

~100 mg of cecal content was well‐suspended in 10 ml cold TBS and centrifuged at 300 × g at 4°C for 1 min. The supernatant was transferred to a new tube and bacteria were pelleted by centrifuging at 7,000 × g at 4°C for 5 min. Cell pellets were then washed with 1 ml cold TBS, re‐suspended and lysed in 400 μl cold TBS containing 1× BugBuster® (Millipore), 1 mM EDTA, 0.5 mg ml⁻¹ chicken egg lysozyme, 25 U ml⁻¹ Benzonase® nuclease (Sigma) and 1× cOmplete™ protease inhibitor, by incubating with constant nutating at 4°C for 10 min. Samples were centrifuged at 20,000 × g at 4°C for 10 min to remove cell debris. Protein concentrations were estimated by measuring absorbance at 280 nm using a Nanodrop 8000 (ThermoFisher).

Western blot detection of wild‐type (i.e., untagged) EF‐G1 and EF‐G2 proteins

A volume corresponding to 50 μg of protein from each cecal sample was combined with 5 μl 4 × LDS Buffer (ThermoFisher) containing 100 mM dithiothreitol and incubated at 95°C for 5 min. Samples were loaded onto a NuPAGE 4–12% Bis‐Tris protein gel (ThermoFisher) and fractionated at 180 V in 1× MOPS running buffer (ThermoFisher) for 60 min. Fractionated proteins were transferred to a nitrocellulose membrane using an iBlot device (Invitrogen) and the resulting membrane was cut below the 65 kD marker and both portions blocked in TBS containing 3% skim milk for 1 h. The EF‐G1 and EF‐G2 proteins were detected at the top portion of the membrane using a 1:1,000 dilution of the rabbit anti‐EF‐G1 antiserum (day‐56 bleed) or 1:5,000 dilution of the rabbit anti‐EF‐G2 antiserum (day‐56 bleed), followed by a 1:5,000 dilution of an HRP‐conjugated anti‐rabbit antibody (General Electric). GroEL was detected on the bottom portion of the membrane using a 1:5,000 dilution of anti‐GroEL (Sigma) followed by incubation with a 1:5,000 dilution of an HRP‐conjugated anti‐rabbit antibody (General Electric). Between primary and secondary antibody incubations, and after secondary antibody incubation, membranes were washed with TBS containing 0.05% Tween‐20 and rinsed with TBS. Blots were developed with SuperSignal West Femto Maximum Sensitivity Substrate (ThermoFisher) and imaged using a LAS‐4000 imager (General Electric).

Cryo‐EM structure analysis

Cryo‐EM sample preparation

For Cryo‐EM grid preparation, we mixed E. coli ribosomes (1.6 μM), EF‐G2 (5.2 μM), and GTP (2.5 mM) in buffer containing 50 mM Tris–HCl pH 7.5, 10 mM MgCl₂ and 70 mM NH₄Cl at room temperature. A total of 5 μl of the samples was applied onto glow‐discharged (40 s) 200‐mesh R2/1 Quanti‐foil Cu grids with 2 nm thick continuous carbon film. The grids were blotted for 3 s manually and rapidly frozen in liquid ethane using a home‐made plunger.

Cryo‐EM single‐particle data acquisition

The frozen grids were loaded into a Titan Krios (ThermoFisher) operated at 300 kV, condenser lens aperture 50 μm, spot size 7, parallel beam with illuminated area of 1.25 μm in diameter. Microscope magnification was at 81,000× (corresponding to a calibrated sampling of 1.068 Å per physical pixel). Movie stacks were collected automatically using Serial‐EM software on a K3 direct electron camera equipped with a GIF quantum energy filter (Gatan) with a slit width of 20 eV to remove inelastically scattered electrons. Images are collected in super‐resolution mode at a recording rate of 40 raw frames per second and a total exposure time of 2 s, yielding 40 frames per image stack, and a total dose of 30 e−/Ų. Totally, ~10,720 micrographs were collected automatically using Serial‐EM (Mastronarde, 2005) with defocus values ranging between −0.6 and −2 μm.

Cryo‐EM data processing and model building

Movie stacks were motion‐corrected using Motioncor2 (Zheng et al, 2017). After CTF correction by Gctf (Zhang, 2016), Relion (Scheres, 2016) were used to pick particles. After 2d classification and 3d classification, ~342,233 particles contributing to the 70S ribosome reconstruction were sorted out. This group of particles were further classified into two classes: 70S ribosome without and with EF‐G2 associated. To improve the resolution of the EF‐G2 associated with the ribosome, we subtracted the EF‐G2 density map from the raw images and carried out focused refinement, which generated a density map with a global resolution of 4 Å estimated using 0.143FSC standard. To analyze the different conformations of the 70S‐EF‐G2, we performed 3d classification on this group of particles.

After 3d classification, 40,595 and 93,541 particles contributing to the two best‐resolved classes from the 70S‐EF‐G2 group were subjected to further 3d refinement, which generates the two density maps with a global resolution of 3.2 and 3.0 Å, respectively.

To better visualize the interaction between the EF‐G2 and SRL, we applied local mask around this region for 3d refinement. After 3d classification, the best class contributed by 63,782 particles was subjected to further refinement, which generates a density map with local resolution of the EF‐G2‐SRL region around 3.7 Å. To build atomic model of the EF‐G2 associated with the ribosome, we first ran AlphaFold2 (Jumper et al, 2021) to obtain predicated model, which was fitted into the focused refined EF‐G2 density map first and then refined in Phenix (Afonine et al, 2018). The refined model was manually adjusted based on our density map in Coot (Emsley et al, 2010). To build the models of the two refined classes from the 70S‐EF‐G2 complexes, we extracted the models of the 70S from the released models (PDB: 7ST6 and 7SSD) and combined them with our refined EF‐G2 model respectively. The composed models were then fitted into the two density maps and refined in real space using Phenix respectively (Appendix Fig S1). All structures were visualized using Chimera (Pettersen et al, 2004) or Chimera X (Pettersen et al, 2021).

Bioinformatic analyses

Bacteroidetes genome analysis was carried out using PATRIC (Davis et al, 2020); phylogenetic tree and data were visualized using iTOL (Letunic & Bork, 2021). The 149 Bacteroidetes genomes available for building the Bacteroidale order level pre‐built tree in PATRIC were first analyzed for the presence and conservation of EF‐G1 and EF‐G2 by tBLASTn, using B. thetaiotaomicron EF‐G1 or EF‐G2 as query sequences. The resulted hits that share more identity with EF‐G1 or EF‐G2 were considered as EF‐G1‐ or EF‐G2‐ orthologs, respectively. More fully sequenced genomes from Bacteroidetes outside the Bacteroidia class were randomly picked for analyzing the presence of EF‐G1 and EF‐G2 using the same tBLASTn method.

Protein sequence alignments were carried out by MUSCLE (Edgar, 2004) and Clustal Omega (Sievers et al, 2011). For analyzing sequence conservation in non‐redundant Bacteroides EF‐G1/EF‐G2 proteins, all proteins entries with annotations containing key words ‘elongation factor G’, ‘EF‐G’ or ‘fusA’, and ‘Bacteroides’ were retrieved from NCBI identical protein groups, aligned using MUSCLE and built distance tree using PHYLIP (Felsenstein, 2005), proteins on the same clade as B. thetaiotaomicron EF‐G1 or EF‐G2 were considered as EF‐G1‐ or EF‐G2‐ ortholog proteins, respectively, and sequence logos were built using WebLogo 3 (Crooks et al, 2004).

Author contributions

Weiwei Han: Conceptualization; formal analysis; investigation; writing – original draft; writing – review and editing. Bee‐Zen Peng: Formal analysis; investigation. Chunyan Wang: Formal analysis; investigation; visualization. Guy E Townsend II: Investigation; writing – review and editing. Natasha A Barry: Resources. Frank Peske: Conceptualization; resources; formal analysis; investigation; methodology. Andrew L Goodman: Resources; funding acquisition. Jun Liu: Supervision; funding acquisition. Marina V Rodnina: Resources; formal analysis; supervision; funding acquisition; methodology; writing – review and editing. Eduardo A Groisman: Conceptualization; formal analysis; funding acquisition; writing – original draft; project administration; writing – review and editing.

Disclosure and competing interests statement

The authors declare that they have no conflict of interest.

Supporting information

Appendix

Expanded View Figures PDF

PDF+

The EMBO Journal (2023) 42: e112372

Data availability

The atomic coordinates and cryo‐EM 3D maps for protein structures produced in this study are available through the Protein Data Bank (PDB) and the Electron Microscopy Data Bank (EMDB), respectively, as follows:

• focused refined EF‐G2: PDB 8DMF (http://www.rcsb.org/pdb/explore/explore.do?structureId=8DMF);

• focused refined EF‐G2: EMD‐27535

• 70 S‐EF‐G2 complex: EMD‐27561

• Class 1 of 70 S‐EF‐G2 complex: EMD‐27543

• Class 2 of 70 S‐EF‐G2 complex: EMD‐27546

• Local refined density map including SRL‐EF‐G2 region: EMD‐27547