RuBisCO: The Slow Enzyme That Feeds the Planet

Switching it on: a carbamate, a magnesium, a carbanion

RuBisCO (ribulose-1,5-bisphosphate carboxylase/oxygenase, EC 4.1.1.39) cannot work straight out of the ribosome. It must first be carbamylated: the neutral ε-amino group of Lys201 (spinach large-subunit numbering) attacks a CO₂ molecule that will never be fixed, giving a carbamate, −NH−COO⁻. That new negative charge, together with the carboxylates of Asp203 and Glu204, chelates a single Mg²⁺ ion. The result is the activated ECM complex — Enzyme·CO₂·Mg.

This is the light switch. Carbamylation needs a deprotonated lysine and free magnesium, and illuminated chloroplasts supply both: protons pumped into the thylakoid lumen push stromal pH from ~7.2 to ~8.0 while Mg²⁺ counter-flows out of the lumen into the stroma. In the dark the enzyme quietly decarbamylates and stops.

Ribulose-1,5-bisphosphate (RuBP) then binds, its C2 and C3 oxygens completing the magnesium's octahedral coordination. Lys175 acts as the base and abstracts the proton from C3, and the resulting 2,3-enediolate — a carbanion smeared over two carbons and stabilised by the Mg²⁺ — is the whole story of this enzyme. Everything good and everything bad that RuBisCO does happens to that one intermediate. A classic labelling result makes the subtlety concrete: the carbamylating CO₂ and the fixed CO₂ are different molecules, proven with ¹⁴C by Lorimer and Miziorko around 1980.

Carboxylation — and the identical route to disaster

With the enediolate formed, mobile loop 6 of the large subunit's (β/α)₈ barrel folds shut over the site, bringing Lys334 in to grip the incoming carboxyl group, and a C-terminal tail latches the lid. Then:

• Carboxylation. CO₂ adds to C2 of the enediolate, giving 2-carboxy-3-ketoarabinitol-1,5-bisphosphate (CKABP).
• Hydration. Water adds at C3 to make a gem-diol.
• Cleavage. The C2–C3 bond breaks. The C3–C5 fragment leaves directly as 3-phosphoglycerate; the fragment carrying C1, C2 and the newly added carboxyl carbon departs as a carbanion (aci-acid) that Lys175 stereospecifically protonates to give the second 3-phosphoglycerate.

Net: 1 RuBP + 1 CO₂ + 1 H₂O → 2 × 3-PGA. Five reaction steps, one active site, no cofactors beyond magnesium.

The oxygenase reaction is the same chemistry with the wrong gas. Triplet O₂ attacks the same C2 of the same enediolate, forming a 2-peroxy-3-keto intermediate; hydration and cleavage then release one 3-PGA and one molecule of 2-phosphoglycolate — a two-carbon dead end.

Salvaging it is photorespiration, and it crosses three organelles: a chloroplast phosphatase gives glycolate; in the peroxisome, glycolate oxidase makes glyoxylate plus H₂O₂ (destroyed by catalase) and transamination gives glycine; in the mitochondrion, the glycine decarboxylase complex working with serine hydroxymethyltransferase turns two glycines into one serine and vents one carbon as CO₂ plus ammonia; serine returns via the peroxisome as glycerate and is re-phosphorylated to 3-PGA. The bill: ATP for glycerate kinase, ATP and reductant to re-fix the ammonia through GS/GOGAT, and a further 3 ATP + 2 NADPH to re-fix the CO₂ just thrown away.

The governing relation: why one turnover in four is a mistake

RuBisCO cannot tell CO₂ from O₂ by chemistry. There is no redox metal in the site — Mg²⁺ is redox-inert — so it has no way to activate one gas and ignore the other. Discrimination is geometric and electrostatic: the site merely stabilises the carboxylation transition state slightly better than the oxygenation one.

The relevant figure of merit is the specificity factor:

• S_c/o = (V_c·K_o) / (V_o·K_c), which gives the working relation v_c/v_o = S_c/o × [CO₂]/[O₂].

Worked example. Take spinach, S_c/o ≈ 90. At 25 °C, air-equilibrated water holds ~14 µM CO₂ (Henry's law: 4.2 × 10⁻⁴ atm × 0.034 mol L⁻¹ atm⁻¹) and ~250 µM O₂. Stomatal and mesophyll resistances drop the chloroplast CO₂ to roughly 60% of ambient, about 8.5 µM, while O₂ is if anything enriched near photosystem II. So v_c/v_o = 90 × 8.5/250 ≈ 3. One oxygenation for every three carboxylations, straight out of the arithmetic.

Cost that out: two oxygenations vent one CO₂, so one oxygenation per three carboxylations throws away 0.5/3 ≈ 17% of gross fixed carbon directly. Add the ATP and reducing power burned on salvage and nitrogen re-assimilation and the accounting lands near 25% of fixed carbon in a temperate C3 crop. Warming makes it worse — CO₂ solubility falls faster than O₂ solubility and S_c/o itself declines — so the photorespiratory compensation point Γ* climbs from ~25 µmol mol⁻¹ at 15 °C to ~43 at 25 °C and ~70 at 35 °C.

Why not evolve a faster enzyme? Tcherkez, Farquhar and Andrews argued in PNAS (2006) that you cannot: improving specificity means binding the carboxylation transition state more tightly, but that transition state closely resembles the carboxyketone intermediate which must leave, so tighter binding necessarily slows turnover. Measured RuBisCOs fall along that trade-off line, each near-optimal for its own CO₂ environment.

The hardware, and why a leaf carries grams of it

Plant RuBisCO is Form I: an L8S8 hexadecamer of ~540 kDa, eight ~53 kDa large subunits and eight ~15 kDa small subunits, with 8 active sites. Each site straddles a large-subunit dimer, built from the N-terminal domain of one subunit and the TIM-barrel of its partner — which is why the monomer is catalytically dead.

Assembly is a chimera of two genomes: rbcL is encoded in the chloroplast, rbcS in the nucleus, and folding needs chaperonin-60/20 plus dedicated assembly chaperones (RbcX, Raf1, Raf2, Bsd2) — which is why plant RuBisCO resisted expression in E. coli for three decades. A second protein, RuBisCO activase, an AAA+ ATPase, is mandatory in plants: it wrenches tightly bound dead-end sugars out of occupied sites so the enzyme can re-carbamylate. It was found through an Arabidopsis mutant (rca) that grew only in high CO₂ — the plant made perfectly good RuBisCO that it could not switch back on.

Worked example — the bulk. A healthy C3 leaf has V_cmax ≈ 100 µmol CO₂ m⁻² s⁻¹. At k_cat = 3 s⁻¹ that needs 100/3 ≈ 33 µmol of active sites per m². Each site carries 540 kDa / 8 = 67,500 g per mol of sites, so 33 × 10⁻⁶ × 67,500 ≈ 2.2 g of RuBisCO per square metre of leaf. Total soluble leaf protein is ~4–5 g m⁻², which is exactly where the famous "up to 50% of soluble protein" comes from. The same numbers give a specific activity of ~2.7 µmol CO₂ min⁻¹ mg⁻¹, matching the published spinach figure, and put active sites in the stroma near 4 mM — four hundred times the concentration of the CO₂ they are hunting.

How it is measured and specified

The canonical assay activates the enzyme first — pH 8.0, ~20 mM MgCl₂, ~10 mM NaHCO₃, several minutes of pre-incubation — then starts with RuBP and traps fixed carbon as acid-stable ¹⁴C. Skipping activation is the standard way to under-report k_cat several fold.

Because k_cat is defined per active site, sites must be counted, not assumed from protein mass. That is done by titration with 2-carboxyarabinitol-1,5-bisphosphate (CABP), a transition-state analogue that binds essentially irreversibly (K_d around 10⁻¹¹ M). The same molecule freezes the enzyme for crystallography: the activated spinach complex with CABP has been refined to ~1.6 Å (PDB 8RUC).

The specificity factor is measured by membrane-inlet mass spectrometry, following CO₂ and O₂ consumption simultaneously on one sample. Beware the units: S_c/o ≈ 80–100 is on a dissolved-concentration basis; quoted against partial pressures the same enzyme scores ~2,100–2,650, because CO₂ is roughly 25–30× more soluble than O₂. Mixing the two is the most common error in the literature.

In the field, RuBisCO is specified in planta through the Farquhar–von Caemmerer–Berry model (1980) fitted to gas-exchange A/C_i curves: V_cmax, K_c ≈ 405 µmol mol⁻¹, K_o ≈ 278 mmol mol⁻¹ and Γ* ≈ 42.75 µmol mol⁻¹ at 25 °C. Those hang together: Γ* = 0.5 × O/S, and 0.5 × 210,000 / 2,450 ≈ 43 µmol mol⁻¹.

A fourth measurement leaves the plant entirely. RuBisCO discriminates against ¹³C by roughly 29‰, which is why C3 plants sit near δ¹³C = −27‰ and C4 plants near −13‰; isotope-ratio mass spectrometry uses that gap to expose honey and fruit juice cut with cane sugar or corn syrup.

A century of arguments

The problem was found before the enzyme was. In 1920 Otto Warburg showed that oxygen inhibits photosynthesis in Chlorella — an effect with no explanation for fifty years.

In 1947 Sam Wildman and Jim Bonner isolated the dominant protein of spinach leaf and, having no idea what it did, called it Fraction I protein. (Wildman's celebrated crystals of the same protein, grown from tobacco, came only in 1971.) In 1954 Quayle, Fuller, Benson and Calvin identified the carboxylating activity of the Calvin–Benson cycle, then named carboxydismutase — and the two turned out to be the same molecule. The acronym RuBisCO was only coined in 1979, by David Eisenberg, at a symposium honouring Wildman.

The decisive result came in 1971, when Bowes, Ogren and Hageman showed the enzyme also acts as an oxygenase, producing phosphoglycolate. Warburg's fifty-year-old inhibition suddenly had a mechanism, and photorespiration had a source.

Structures followed: the simple L2 enzyme from Rhodospirillum rubrum in 1986, then activated plant L8S8 with CABP by the end of that decade. Activase emerged from Somerville and Ogren's rca mutant in the early 1980s, and Tcherkez, Farquhar and Andrews delivered the near-optimality argument in 2006. Bar-On and Milo put a number on the bulk in 2019: ~0.7 Gt of RuBisCO on Earth, about 90 kg for every living person. Engineering has so far had more luck going around the enzyme than improving it: a synthetic glycolate bypass that short-circuits photorespiration raised field-grown tobacco biomass by roughly 40% (South et al., Science, 2019).

How it fails, and the three things it is not

Failure modes are all about blocked sites. If RuBP binds the decarbamylated enzyme, it locks it shut — substrate acting as inhibitor. Some species make 2-carboxyarabinitol-1-phosphate (CA1P) overnight, a natural tight-binding inhibitor that parks the enzyme until dawn. And roughly once every few hundred turnovers the enediolate misfires, yielding xylulose-1,5-bisphosphate and related dead-end sugars that jam the site. All three are cleared by activase — which is itself thermolabile, denaturing above roughly 35 °C. That is a major reason photosynthesis collapses in a heatwave before the leaf is damaged: the catalyst is fine, its mechanic has quit. A separate cost is nitrogen — ~20–25% of leaf nitrogen is locked in this one protein.

It is not the Calvin cycle. RuBisCO is step one of thirteen. The cycle also needs phosphoglycerate kinase and GAPDH for reduction, and a rearrangement sequence — aldolase, transketolase, SBPase, phosphoribulokinase — to regenerate RuBP. Three RuBisCO turnovers are needed to net one exported triose phosphate.

It is not PEP carboxylase. PEPC takes bicarbonate, not CO₂, runs roughly 10× faster, and its enolate intermediate is too short-lived and too buried in the site for O₂ to attack. C4 and CAM plants exploit that: PEPC fixes carbon in the mesophyll, then decarboxylation dumps CO₂ into the bundle sheath at ~1,000–2,000 µmol mol⁻¹, where oxygenation is all but abolished. The pump costs 2 extra ATP per CO₂, which is why C3 plants still win in cool and shaded conditions.

And photorespiration is not pure waste. 2-phosphoglycolate is genuinely toxic — it inhibits triose-phosphate isomerase and sedoheptulose-1,7-bisphosphatase — and mutants that cannot run the salvage pathway die in ordinary air. Calling it a design flaw mistakes the ambulance for the accident.

RuBisCO against the enzymes and forms it gets confused with

Enzyme	What it captures	Turnover per site	Reacts with O2?
RuBisCO Form I (C3 plant, L8S8)	Dissolved CO₂ onto RuBP	~3 s⁻¹ (spinach, 25 °C)	Yes — roughly 1 turnover in 4; S_c/o ≈ 80–100
RuBisCO Form II (Rhodospirillum rubrum, L2)	Dissolved CO₂ onto RuBP	~5–10 s⁻¹	Yes, badly — S_c/o ≈ 10–15; evolved in low-O₂ niches
RuBisCO Form IV (RuBisCO-like protein)	Nothing — a methionine-salvage enolase	Not a carboxylase	No
PEP carboxylase (EC 4.1.1.31)	HCO₃⁻ onto phosphoenolpyruvate	~20–40 s⁻¹ (roughly 10× RuBisCO)	No — its enolate intermediate is short-lived and shielded from O₂
Carbonic anhydrase II (EC 4.2.1.1)	Only interconverts CO₂ and HCO₃⁻ — fixes no carbon	~10⁶ s⁻¹	No