Cytochrome P450: The Oxygen Rebound That Breaks Unreactive C–H Bonds

A heme with a sulfur behind it — and why 450 nanometres

A cytochrome P450 is a heme-thiolate monooxygenase. The cofactor is iron protoporphyrin IX — the same heme b that colours blood — but the protein holds it by a cysteine whose deprotonated sulfur binds the iron from below. That proximal thiolate is the whole difference: haemoglobin uses a histidine, cytochrome c a histidine and a methionine, and P450 a negatively charged sulfur that changes what the iron can do to oxygen.

The sixth, distal site is where chemistry happens — water at rest, then O₂, a peroxide, a ferryl oxygen, finally the product's hydroxyl. The net reaction looks trivial: R–H + O₂ + NAD(P)H + H⁺ → R–OH + H₂O + NAD(P)⁺. One oxygen atom enters the substrate, the other is reduced to water — that is what monooxygenase means.

The name is an accident of spectroscopy. Klingenberg and Garfinkel independently found a carbon-monoxide-binding pigment in liver microsomes in 1958; Omura and Sato proved in 1964 that it was a hemoprotein and named it P-450, for pigment, 450 nanometres. An ordinary ferrous heme–CO complex such as carboxymyoglobin absorbs at 420 nm. The thiolate's high-lying sulfur 3p orbitals push electron density onto the iron and mix a sulfur→porphyrin π* charge-transfer state into the porphyrin's own π→π* Soret transition; the band splits into the hyperporphyrin pair and its red-shifted half lands at 450 nm. Protonate that sulfur and it snaps back to 420 nm — the dead form called P420, which is why the 450/420 ratio is checked on every preparation.

The catalytic cycle, step by step

• 1 — Binding and the spin flip. The substrate pushes the axial water off the iron; six-coordinate low-spin Fe(III) (S = 1/2) becomes five-coordinate high-spin Fe(III) (S = 5/2), and the Soret shifts from about 417 nm to about 390 nm — the type I difference spectrum used to titrate binding.
• 2 — The redox gate. That spin change drags the reduction potential with it: in P450cam, camphor binding moves it from −303 mV to −173 mV. The 130 mV are a safety interlock — an empty enzyme cannot be reduced, so it cannot make reactive oxygen for nothing.
• 3 — First electron. From the diflavin NADPH–cytochrome P450 reductase in microsomal (class II) enzymes, or from a ferredoxin — putidaredoxin, E°′ ≈ −196 mV — in bacterial class I systems like P450cam.
• 4 — O₂ binds. The ferrous iron takes dioxygen and immediately hands it an electron, giving ferric-superoxo, Fe(III)–O₂•⁻ — the last intermediate stable enough to sit in a cuvette.
• 5 — Second electron. From the reductase again or from cytochrome b5, giving the ferric-peroxo dianion Fe(III)–O₂²⁻, a nucleophile.
• 6 — Two protons. The first gives ferric-hydroperoxo, Fe(III)–OOH (Compound 0), delivered in P450cam by the conserved Asp251/Thr252 acid–alcohol pair. The second lands on the distal oxygen, making it a leaving group.
• 7 — Compound I. The O–O bond breaks heterolytically, water departs with both electrons, and the iron is left two oxidising equivalents above ferric: an oxo-iron(IV) on a porphyrin π-cation radical, Fe=O ≈ 1.65 Å.
• 8 — Abstraction and rebound. Compound I takes a hydrogen atom — proton and electron together — leaving Fe(IV)–OH (Compound II, Fe–O ≈ 1.82 Å) and a carbon radical, which rebounds onto that hydroxyl. Product leaves, water returns, resting state restored.

Hydrogen peroxide offers a shortcut straight to Compound 0 and on to Compound I. This peroxide shunt is how Compound I is made in the laboratory, and how fungal unspecific peroxygenases run for a living.

The bond-strength budget, worked through

The puzzle is thermodynamic. Compound I must remove a hydrogen from a C–H bond worth 96–105 kcal/mol (401–439 kJ/mol): 105 for methane, 101 primary, 99 secondary, 96 tertiary, 99 for camphor's C5–H. Abstraction is downhill only if the bond formed is at least as strong as the bond broken.

The bond formed is the Fe(IV)O–H of Compound II, and Yosca and Green measured it in 2013 by measuring how basic Compound II is. A Bordwell square scheme turns an acidity and a reduction potential into a bond dissociation free energy:

BDFE(Fe(IV)O–H) = 1.37 × pKa + 23.06 × E° + 57.6 kcal/mol

For CYP119 the measured pKa is ≈ 12 and E°(Compound I/Compound II) ≈ 1.1 V, so 1.37 × 12 = 16.4, 23.06 × 1.1 = 25.4, and the sum with 57.6 gives 99.4 — ≈ 100 kcal/mol (≈418 kJ/mol), the value Green quotes. Abstracting camphor's 99 kcal/mol C–H is near-thermoneutral; even methane is only a few kcal/mol uphill.

Now see what the thiolate buys. Horseradish peroxidase uses a histidine and its Compound II has a pKa below 4 — it never protonates. Each pKa unit is worth 1.37 kcal/mol, so falling from 12 to 4 removes 1.37 × 8 ≈ 11 kcal/mol of driving force. Carried into the barrier, that slows the reaction at 310 K by exp(11 000 / (1.987 × 310)) = e^17.9 ≈ 6 × 10⁷. One amino-acid choice is worth seven orders of magnitude — which is why peroxidases oxidise dyes while P450s oxidise alkanes.

The redox gate is the same arithmetic in millivolts. At 25 °C, RT/F = 25.7 mV, so 130 mV moves an equilibrium constant by exp(130/25.7) = e^5.06 ≈ 160-fold. Against putidaredoxin at −196 mV, electron transfer into substrate-free P450cam (−303 mV) has K ≈ 0.015; with camphor bound (−173 mV), K ≈ 2.4. Nothing signals the enzyme that a substrate has arrived — the arrival is the signal, written in millivolts.

Oxygen rebound: Groves in 1976, Compound I in 2010

John Groves coined oxygen rebound in 1976 for a model iron system, and in 1978 pinned it on the enzyme: purified liver P450 hydroxylating norbornane gave scrambled exo/endo alcohols and an intramolecular kinetic isotope effect near 11. Scrambling proves a freely rotating carbon radical existed; a large isotope effect proves C–H cleavage is the committed step.

How long does that radical live? Newcomb and co-workers used radical clocks — substrates carrying a strained ring that opens at a calibrated rate. With probes rearranging at ~10¹¹ s⁻¹ and almost no rearranged product appearing, the apparent radical lifetime is only ~70–200 femtoseconds — a few C–C stretching periods (one is about 33 fs) and far short of the ~1 picosecond a rotation about a C–C bond takes. Sason Shaik's two-state reactivity explains the residue: the low-spin doublet rebounds with essentially no barrier, the high-spin quartet with a small one, so what little rearranged product appears comes from the minority spin state. Rebound finishes in under 1 picosecond, which is why P450 hydroxylations retain stereochemistry despite running through a radical.

Compound I resisted capture for 34 years because with substrate present it does not last. Rittle and Green trapped it in 2010 in CYP119, a P450 from the thermoacidophile Sulfolobus, by oxidising the resting enzyme with m-chloroperoxybenzoic acid — about 75% Compound I, characterised by Mössbauer, EPR and UV–visible spectroscopy. On lauric acid it worked at 1.1 × 10⁷ M⁻¹s⁻¹ with a kinetic isotope effect of ≈ 13 — at 100 µM substrate, 1 100 s⁻¹, a half-life of 0.6 milliseconds. That isotope effect is itself evidence of tunnelling: zero-point energy alone caps the semiclassical value near 7, so 13 means the hydrogen goes through the barrier, not over it.

How P450 is measured, and how regulators specify it

Counting enzyme. Reduce with sodium dithionite, bubble in carbon monoxide, record the reduced-CO-minus-reduced difference spectrum. The Omura–Sato extinction coefficient for the 450 minus 490 nm difference, 91 mM⁻¹cm⁻¹, still converts absorbance into nanomoles of P450 per milligram of microsomal protein; a shoulder at 420 nm is dead enzyme.

Catching intermediates. Rapid-mix freeze-quench at millisecond resolution, ⁵⁷Fe Mössbauer for the iron's oxidation state and spin, EPR for the porphyrin radical, EXAFS for bond lengths — the Fe–O distance is exactly how Compound I (≈1.65 Å, a double bond) is told from Compound II (≈1.82 Å, a hydroxide). Cryoradiolysis — gamma-irradiating the frozen oxy complex at 77 K — is how the peroxo and hydroperoxo species were seen at all. The first P450 structure was Poulos and colleagues' 2.6 Å map of P450cam in 1985, refined by the same group to 1.63 Å in 1987 (PDB 2CPP).

Drug-development practice. Candidates are profiled against human liver microsomes and recombinant single-enzyme preparations using the index reactions regulators name — diclofenac 4′-hydroxylation for CYP2C9, dextromethorphan O-demethylation for 2D6, midazolam 1′-hydroxylation and testosterone 6β-hydroxylation for 3A4. The FDA's In Vitro Drug Interaction Studies guidance (final, January 2020) and the EMA's drug-interaction guideline set the decision trees. An inhibitor is strong if it raises a sensitive substrate's AUC at least 5-fold, moderate at 2–5-fold, weak at 1.25–2-fold.

Failure modes, famous casualties and deliberate targeting

Uncoupling. The cycle leaks at three points: ferric-superoxo can autoxidise and shed superoxide, Compound 0 can release hydrogen peroxide, and Compound I can be reduced straight to water. P450cam on native camphor is over 95% coupled, but engineered P450s on unnatural substrates often fall below 10% — the central problem of P450 biocatalysis.

Suicide inhibition. If the radical or an epoxide rebounds onto the porphyrin instead of the substrate, the enzyme destroys itself. Terminal alkynes and alkenes alkylate a pyrrole nitrogen; so do 17α-ethinylestradiol and the furanocoumarins of grapefruit juice (bergamottin and 6′,7′-dihydroxybergamottin) on intestinal CYP3A4 — found by accident in 1989, when grapefruit juice was used to mask the taste of ethanol in a felodipine study. One glass can raise some drugs' bioavailability several-fold for a day, because killed enzyme must be resynthesised.

Named casualties. Terfenadine (Seldane) was a safe antihistamine only because CYP3A4 destroyed it completely on first pass; with ketoconazole or erythromycin the parent survived, blocked the hERG potassium channel and caused torsades de pointes — withdrawn in the US in 1997–98 and replaced by its harmless metabolite fexofenadine. Codeine is the mirror image, a prodrug CYP2D6 activates: ultrarapid metabolisers with gene duplications have died of opioid toxicity, bringing an FDA boxed warning in 2013 and a contraindication under age 12 in 2017. CYP2E1 turns 5–10% of a paracetamol dose into NAPQI, harmless until glutathione runs out; CYP3A4 converts aflatoxin B₁ into the 8,9-epoxide that leaves the p53 codon-249 G→T signature in liver tumours.

Targeting it on purpose. Every azole antifungal puts a triazole or imidazole nitrogen onto the iron of CYP51, lanosterol 14α-demethylase; letrozole does the same to aromatase (CYP19A1), abiraterone to CYP17A1. Industry uses the constructive side: Upjohn's 1952 microbial 11α-hydroxylation of progesterone by Rhizopus collapsed the price of cortisone from roughly $200 a gram to a few dollars, and P450 BM3 (CYP102A1), fused to its own reductase and turning over ~17 000 times a minute, is the standard chassis for directed evolution.

What it is not: the look-alikes

Not a cytochrome in the working sense. Cytochromes a, b and c are electron carriers: they shuttle single electrons along a respiratory chain and never bind a substrate. P450 keeps the name only because it was found as a coloured, CO-binding pigment before anyone knew what it did.

Not a peroxidase. Horseradish peroxidase makes the same kind of Compound I, but with a proximal histidine, and oxidises substrates by one-electron transfer at the exposed heme edge — it never transfers its oxygen atom. The thiolate, the buried pocket and the unusually basic Compound II are what make P450 an oxygen-transfer catalyst.

Not a dioxygenase or a flavin monooxygenase. Dioxygenases put both atoms of O₂ into the substrate. Flavin monooxygenases such as FMO3 use a C4a-hydroperoxyflavin on soft nucleophiles — nitrogen and sulfur — and cannot touch a C–H bond; FMO3 deficiency, not any P450 defect, causes trimethylaminuria.

Not a free hydroxyl radical. Free •OH abstracts hydrogen at nearly every collision and would shred the protein around it. Compound I is a hydroxyl equivalent on a leash: comparable power, aimed by a pocket that decides which carbon meets the ferryl oxygen. And the 450 is a wavelength in nanometres — not a mass, a temperature, or a number in a series.

Compound I and its look-alikes: what the proximal ligand decides

Feature	Cytochrome P450	Heme peroxidase (HRP)	Non-heme iron oxygenase (TauD)
Proximal ligand	Cysteine thiolate (S⁻)	Histidine, H-bonded to an aspartate	His₂/carboxylate facial triad, no porphyrin
Reactive intermediate	Fe(IV)=O on a porphyrin cation radical, Fe=O ≈ 1.65 Å	Fe(IV)=O on a porphyrin cation radical	High-spin Fe(IV)=O, S = 2
Compound II basicity	pKa ≈ 12 — exists as Fe(IV)–OH	pKa below 4 — stays as Fe(IV)=O	Not porphyrin-based
Fe(IV)O–H bond strength	≈ 100 kcal/mol (≈418 kJ/mol)	≈ 90 kcal/mol — too weak for an alkane	Comparable, but reached via 2-oxoglutarate
What it does to substrate	Abstracts H from a 96–105 kcal/mol C–H, then oxygen rebound	One-electron oxidation at the exposed heme edge; no O transfer	H-abstraction, then hydroxyl rebound (halogenase cousins rebound Cl instead)
Signature isotope effect	≈ 13 (CYP119 + lauric acid)	Not applicable — no C–H cleavage	≈ 37 (TauD + taurine)