Elongation and Codon Recognition

Elongation is the phase of translation during which the ribosome reads codons sequentially and adds amino acids to a growing polypeptide chain, following initiation at the start codon. Each elongation cycle has three steps: codon recognition (selection of the correct aminoacyl-tRNA), peptide bond formation (catalyzed at the peptidyl transferase center), and translocation (movement of the ribosome by one codon). This article isolates the first step, the moment at which the ribosome must decide, molecule by molecule, whether an incoming tRNA correctly matches the mRNA codon exposed in the aminoacyl (A) site. The central molecular actor is elongation factor Tu (EF-Tu) in bacteria (its eukaryotic and archaeal counterpart is eEF1A). EF-Tu is a small GTPase, roughly 43 kDa, that binds GTP and an aminoacyl-tRNA simultaneously to form a ternary complex. This complex, not free aminoacyl-tRNA, is the actual substrate delivered to the ribosome. EF-Tu serves two purposes at once: it protects the labile ester bond linking the amino acid to the tRNA from spontaneous hydrolysis, and it converts the act of codon recognition into a kinetic proofreading problem, using the energy of GTP hydrolysis to discriminate correct from incorrect codon-anticodon pairs far more sharply than binding affinity alone would allow. The key conceptual leap, first articulated by John Hopfield and Jacques Ninio independently in 1974-1975, is that a single binding-and-release equilibrium cannot generate very high fidelity because the free-energy difference between a correct (Watson-Crick) and a near-cognate (single-mismatch) codon-anticodon pair is small — on the order of 1-3 kcal/mol, corresponding to at most 10- to 30-fold discrimination at equilibrium. Yet observed error rates are far lower than this. The resolution is that the ribosome does not simply let the ternary complex bind and dissociate at equilibrium; it interposes an irreversible step (GTP hydrolysis, then EF-Tu release) between an initial selection and a later proofreading check, so that the small discrimination is applied twice, multiplying rather than adding.

Step 1 — Ternary complex formation. In the cytoplasm, aminoacyl-tRNA synthetases charge tRNAs with their cognate amino acids (a fidelity step in its own right, not covered here). EF-Tu·GTP binds the aminoacylated 3′ end of the tRNA, forming EF-Tu·GTP·aa-tRNA. EF-Tu binds essentially all elongator aminoacyl-tRNAs with comparable affinity regardless of which amino acid is attached — a property called the "adaptor" or thermodynamic compensation function, ensuring that bulky or small amino acids do not bias delivery rates. Only initiator fMet-tRNA and fully deacylated tRNA are excluded from this complex.

Step 2 — Delivery and initial codon sampling. The ternary complex diffuses into the ribosomal A site, docking first through contacts between EF-Tu and the GTPase-associated center of the large subunit (sarcin-ricin loop region of 23S rRNA) and the L7/L12 stalk. The anticodon loop of the tRNA, still tethered to EF-Tu, samples the mRNA codon displayed in the A site of the small (30S) subunit. At this stage the tRNA is not yet fully accommodated into the A site; it is held in a bent, "A/T" (A site, tRNA-bound-to-EF-Tu) configuration.

Step 3 — Codon-anticodon monitoring. Three universally conserved bases of 16S rRNA — A1492, A1493, and G530 (Escherichia coli numbering) — flip out of helix 44 and probe the minor groove of the codon-anticodon helix. These bases hydrogen-bond to the sugar-phosphate backbone in a geometry that is satisfied only when Watson-Crick base pairing is present at all three codon positions (with the wobble-position allowance discussed under the Wobble Hypothesis). A correct fit closes the 30S subunit around the mRNA-tRNA duplex, a domain closure that is transmitted, through the tRNA body and EF-Tu, to the GTPase center of EF-Tu itself.

Step 4 — GTPase activation. Domain closure and correct codon pairing accelerate the intrinsic GTPase activity of EF-Tu roughly 10⁵-fold, allowing the catalytic His84 of EF-Tu (assisted by the sarcin-ricin loop of 23S rRNA, which is not itself catalytic but organizes the active site) to promote GTP hydrolysis. Near-cognate tRNAs, which fail to satisfy the decoding-center geometry, activate GTP hydrolysis far more slowly and are usually rejected (dissociate) before hydrolysis occurs. This is the initial selection step.

Step 5 — EF-Tu release and proofreading. GTP hydrolysis converts EF-Tu to its GDP-bound form, which has low affinity for aminoacyl-tRNA. EF-Tu·GDP undergoes a large conformational change (domains 1, 2, and 3 splay apart) and dissociates from the ribosome and the tRNA. The tRNA, now freed from EF-Tu, must still travel from the A/T state to the fully accommodated A/A state, threading its acceptor end into the peptidyl transferase center. This accommodation step is itself a second, independent check: because GTP has already been irreversibly hydrolyzed, a near-cognate tRNA that survived initial selection (a rare event) is still likely to dissociate here rather than proceed, since its codon-anticodon interaction is intrinsically less stable. This second chance is the proofreading step. Only a correctly paired, accommodated tRNA reaches peptidyl transferase for peptide bond formation.

Step 6 — Consequence for overall fidelity. Because rejection can happen either before or after the irreversible GTP hydrolysis step, and because both checks use the same underlying (and imperfect) discrimination based on codon-anticodon geometry, the two probabilities of "correct tRNA proceeds" versus "near-cognate tRNA proceeds" are multiplied rather than added. If each check alone discriminates by a factor of ~200, the combined process can discriminate by a factor approaching 200 × 200 = 40,000, consistent with observed in vivo error rates near 10⁻⁴ to 10⁻⁵ per codon. The energetic price is that every ternary complex that binds — correct or incorrect — typically hydrolyzes one GTP, so the cell pays a fixed GTP cost per elongation cycle regardless of outcome, plus extra GTP for near-cognate complexes that are hydrolyzed and rejected.

Once EF-Tu·GDP has departed, it must be recharged. A separate protein, EF-Ts, catalyzes nucleotide exchange on EF-Tu·GDP, displacing GDP so that a fresh GTP can bind; the regenerated EF-Tu·GTP is then free to pick up another aminoacyl-tRNA and re-enter the cycle. This regeneration step is essential because cellular GDP/GTP ratios alone would not release GDP fast enough to sustain the observed elongation rate of ~20 amino acids per second per ribosome.

Ninio and Hopfield's theoretical proposal (1974-1975). John Hopfield and, independently, Jacques Ninio noted a puzzle: measured translational error rates (roughly 10⁻³ to 10⁻⁴) were far lower than what the free-energy difference between correct and incorrect codon-anticodon pairs (a few kcal/mol) could explain via simple equilibrium binding. Hopfield proposed "kinetic proofreading": if a discrimination step is followed by an irreversible reaction (fueled by NTP hydrolysis) before the product is used, and if a second, independent discrimination step follows, the two small biases multiply. This was a prediction from first-principles enzyme kinetics, made before the detailed structural mechanism was known — a rare case in molecular biology of theory substantially preceding structural confirmation.

Biochemical kinetics of EF-Tu-dependent binding (Kurland, Ehrenberg, and colleagues, 1970s-1990s). Using purified E. coli translation components in reconstituted, defined in vitro systems, Mans Ehrenberg, Charles Kurland, and coworkers measured the rate constants for each substep of ternary complex binding, GTP hydrolysis, and tRNA rejection, for both cognate and near-cognate codon-anticodon pairs. They showed directly that (1) GTP hydrolysis by EF-Tu is much faster on cognate than on near-cognate complexes (consistent with an initial selection step), and (2) even after GTP hydrolysis, near-cognate tRNAs dissociate from the ribosome at a much higher rate than cognate ones during the accommodation step (the proofreading step), while cognate tRNAs proceed efficiently to peptide bond formation. Fitting these rate constants into a full kinetic scheme reproduced the observed in vivo error frequency of about 10⁻⁴, providing quantitative, mechanistic confirmation of Hopfield's model — not just a plausibility argument but measured numbers that add up.

Kirromycin as a mechanistic probe. The antibiotic kirromycin binds EF-Tu and blocks the conformational change and dissociation that normally follow GTP hydrolysis, effectively freezing EF-Tu on the ribosome after GTP is hydrolyzed. In the presence of kirromycin, GTP hydrolysis still occurs (showing hydrolysis is triggered by codon recognition, not by subsequent conformational release), but the aminoacyl-tRNA remains delivered without normal proofreading rejection, and error rates rise measurably. This drug experiment separated the "GTP hydrolysis trigger" step from the "EF-Tu release and accommodation" step, confirming that these are mechanistically and temporally distinct — exactly as the kinetic proofreading model requires two separable checkpoints.

Structural confirmation: the decoding center caught in the act. In the early 2000s, X-ray crystallography of the 30S ribosomal subunit bound to cognate and near-cognate codon-anticodon complexes — notably work from Venki Ramakrishnan's laboratory (building on earlier crystallographic work by the groups of Ada Yonath, Thomas Steitz, and Harry Noller) — directly visualized bases A1492, A1493, and G530 of 16S rRNA flipping out to monitor the minor groove of the codon-anticodon helix. Crucially, these structures showed that this monitoring geometry is satisfied for Watson-Crick pairs at the first two codon positions but not for mismatches, giving a direct structural explanation for the "initial selection" discrimination step that biochemical kinetics had already inferred indirectly. Later cryo-EM studies (Harry Noller, Joachim Frank, Rachel Green, and others through the 2010s) captured the ribosome-bound ternary complex in multiple intermediate states — A/T binding, GTPase activation, and accommodation — visually confirming the sequence of conformational events proposed from kinetics decades earlier.

Single-molecule fluorescence studies (2000s-2010s). Using FRET-labeled tRNAs and ribosomes, groups led by Ruben Gonzalez, Joachim Frank, and others tracked individual ternary complex binding events in real time. These experiments directly observed that most binding events, including many cognate ones, end in rapid, reversible sampling and dissociation without GTP hydrolysis (consistent with an initial, low-cost screening step), while a smaller fraction proceed to GTP hydrolysis and, more rarely, full accommodation — providing direct, single-molecule temporal confirmation of the multi-step selection process inferred from bulk kinetics. Together these lines of evidence — theoretical prediction, bulk enzyme kinetics, a mechanistic antibiotic dissection, atomic-resolution structures, and single-molecule real-time observation — form an unusually complete case in molecular biology where a quantitative model was proposed, tested, and then directly visualized.

Example 1 — Quantifying the fidelity gain from two-step proofreading. Suppose the intrinsic thermodynamic discrimination between a cognate and a near-cognate codon-anticodon pair (a single G·U wobble-type mismatch, say) corresponds to a free energy difference of ΔΔG = 2 kcal/mol at 37 °C (310 K). Equilibrium discrimination factor (single check) is given by: K = e^(ΔΔG / RT) With R = 1.987 × 10⁻³ kcal/mol·K: RT = 1.987 × 10⁻³ × 310 ≈ 0.616 kcal/mol K = e^(2 / 0.616) = e^3.25 ≈ 26 So a single equilibrium binding step discriminates only about 26-fold — nowhere near the roughly 10,000-fold discrimination implied by an in vivo error rate of 10⁻⁴ (i.e., 1/10,000). If the ribosome applies this same discrimination twice, at two kinetically separated checkpoints (initial selection and proofreading), the combined discrimination is approximately: 26 × 26 ≈ 676 This is closer to, though still short of, the observed fidelity — showing that real ΔΔG values (which vary by codon position and mismatch type, and are enhanced by the decoding-center's induced-fit tightening around correct pairs) must be somewhat larger than 2 kcal/mol at at least one step, or that additional factors (differential accommodation rates, differential dissociation rates that are not purely thermodynamic) contribute extra discrimination beyond the simple equilibrium estimate. The qualitative lesson holds regardless: two multiplicative checks vastly outperform one, converting hundreds-fold reliability into thousands-fold reliability — the entire point of kinetic proofreading.

Example 2 — GTP cost of protein synthesis with a realistic mis-selection load. Consider a bacterial ribosome synthesizing a 300-residue protein. Assume:

• Each correct incorporation event consumes 1 GTP (via EF-Tu) plus 1 GTP for translocation (via EF-G, foreshadowing translocation) — but here we isolate only the EF-Tu/codon-recognition GTP cost.
• For every correct ternary complex that is accepted, on average 4 near-cognate or non-cognate ternary complexes sample the A site and are rejected, but importantly, only a fraction of these rejected complexes proceed far enough to trigger GTP hydrolysis before being rejected at the initial selection step; assume 50% of sampled near-cognate complexes trigger hydrolysis before rejection (i.e., they are rejected during proofreading, having already spent a GTP), and 50% are rejected before hydrolysis (no GTP spent).

For one correct amino acid added: GTP consumed by the correct complex = 1 GTP consumed by rejected near-cognate complexes = 4 × 0.5 × 1 = 2 Total GTP per residue (EF-Tu cycle only) = 1 + 2 = 3 GTP For the full 300-residue protein: Total EF-Tu-associated GTP = 300 × 3 = 900 GTP molecules This is a substantial "overhead": two-thirds of the GTP spent in this step of translation is consumed rejecting incorrect tRNAs rather than advancing correct ones. This calculation illustrates concretely why fidelity is not free — the cell pays continuously in GTP to maintain an error rate low enough (about 1 wrong amino acid per 10,000 correctly incorporated) that a 300-residue protein has, on average, roughly a 3% chance of containing at least one missense error (1 − 0.9999^300 ≈ 0.03), a tolerable but non-negligible burden that helps explain why proofreading exists at all: without it, a 300-residue protein made at ~1/26 fidelity per residue would almost always contain a mistake.

The theoretical breakthrough behind this article's central idea came from a physicist turned molecular biologist rather than a classical biochemist, illustrating how deeply quantitative reasoning has shaped modern molecular biology.

Antibiotics that hijack the decoding center or EF-Tu. Several clinically important antibiotic classes act directly on the machinery described here. Aminoglycosides (such as streptomycin and gentamicin) bind near the 16S rRNA decoding bases (A1492/A1493) and stabilize the closed, "accepting" conformation of the decoding center even when codon-anticodon pairing is incorrect, dramatically increasing bacterial misreading and inducing lethal levels of mistranslation — a direct pharmacological exploitation of the fidelity mechanism described in this article. Kirromycin and the related pulvomycin/GE2270A class target EF-Tu itself, either freezing it on the ribosome after GTP hydrolysis (kirromycin) or blocking ternary complex formation altogether, and remain of interest as leads for new antibacterials given the essential and bacteria-specific nature of EF-Tu's cycle (its eukaryotic counterpart eEF1A differs enough structurally to allow selective drug targeting).

Engineering fidelity for synthetic biology and genetic code expansion. Efforts to incorporate unnatural or non-canonical amino acids into proteins (for protein engineering, biopharmaceuticals, and click-chemistry labeling) depend on engineering orthogonal EF-Tu variants, or bypassing wild-type EF-Tu's proofreading tolerances, so that a modified aminoacyl-tRNA carrying a non-standard amino acid can still be efficiently and faithfully delivered to reprogrammed codons. Understanding exactly how EF-Tu's ternary complex affinity and GTPase triggering work has allowed rational engineering of EF-Tu mutants that better accommodate bulky, non-natural side chains without sacrificing delivery efficiency.

Mistranslation, aging, and disease. Mutations that compromise ribosomal decoding fidelity (in ribosomal proteins, rRNA, or translation factors) or aminoacylation fidelity are linked in model organisms to increased proteotoxic stress, mitochondrial dysfunction, and reduced lifespan; in humans, some mitochondrial translation factor mutations that impair analogous fidelity mechanisms in mitochondrial ribosomes are associated with mitochondrial diseases affecting high-energy-demand tissues (muscle, brain). Conversely, cancer cells sometimes show altered translational fidelity control, an active area of research into whether deliberately perturbing elongation fidelity could be therapeutically exploited.

Basic quantitative biology and biophysics education. Beyond direct applications, this mechanism is one of the most frequently used teaching examples of how thermodynamics alone is insufficient to explain a biological outcome, and how coupling a discrimination step to an energy-consuming, irreversible reaction is a general strategy cells use whenever high fidelity is required — a principle that recurs in DNA replication proofreading and immune receptor signaling far beyond translation itself.

Ask the tutor: How does the wobble position at codon position 3 avoid becoming a major source of translational errors if it is monitored less strictly than positions 1 and 2?