Retinal Photoisomerization: The 200-Femtosecond Twist That Starts Vision

One Chromophore, Seven Helices, One Salt Bridge

Rhodopsin is a 348-residue class-A G-protein-coupled receptor. Its seven transmembrane helices cage a single buried chromophore: 11-cis-retinal, covalently joined to Lys296 on helix 7 as a Schiff base (an imine, C15=NH⁺–Lys). That imine nitrogen is protonated, and its positive charge is held by the carboxylate of Glu113 on helix 3. That buried ion pair sets the colour and gates the reaction.

The counterion is why rhodopsin is pink rather than yellow. The same protonated retinal Schiff base free in methanol absorbs near 440 nm; inside opsin it absorbs at 500 nm. In wavenumbers that is 22,700 cm⁻¹ against 20,000 cm⁻¹ — a red shift of about 2,700 cm⁻¹, or 7.7 kcal/mol, called the opsin shift. The protein produces it by loosening the electrostatic grip on the positive charge: Glu113 sits a few ångströms away rather than in contact, and polarisable residues line the polyene chain. The π system delocalises, the HOMO–LUMO gap narrows, and absorption lands in the blue-green, exactly where twilight is brightest. Change only the residues around the chromophore and you get the cone pigments — the identical retinal molecule tuned from 420 nm to 560 nm. Colour vision is electrostatic spectral tuning, not chemical variety.

Two Hundred Femtoseconds, Step by Step

Absorption is vertical: the Franck–Condon principle says electrons rearrange while the nuclei are effectively frozen, so the molecule arrives on the excited surface with its 11-cis geometry intact but the wrong electronic structure for it.

• t = 0. A 500 nm photon drives a π→π* excitation. The excitation inverts bond-order alternation along the polyene: formal single bonds gain double-bond character and C11=C12 loses it. The torsional barrier that held the cis kink rigid simply disappears.
• 0–80 fs. The molecule slides down an essentially barrierless S1 surface. Torsion about C11–C12 couples to hydrogen-out-of-plane (HOOP) wagging of the C11–H and C12–H pair, and the vibrational wavepacket stays compact rather than dephasing.
• ~80 fs. The wavepacket reaches an S1/S0 conical intersection near a 90° twist (Polli et al., Nature, 2010). There the two potential energy surfaces touch, the Born–Oppenheimer separation collapses, and population funnels to the ground state. This is the trapdoor: no fluorescence, no intersystem crossing, no time for anything else.
• ~200 fs. The all-trans photoproduct — photorhodopsin — is complete (Schoenlein, Peteanu, Mathies and Shank, Science, 1991). It is born vibrationally coherent: the product absorption is still modulated at roughly 60 cm⁻¹, a wavepacket that survived the reaction.
• Picoseconds. It relaxes into bathorhodopsin: a strained, twisted all-trans chromophore still trapped in an 11-cis-shaped pocket, flagged by an intense resonance-Raman HOOP band near 921 cm⁻¹.

For scale: a 1,500 cm⁻¹ C=C stretch has a 22 fs period, and the low-frequency torsion that carries the reaction (~100 cm⁻¹) has a period near 330 fs. The reaction finishes in less than one full swing of its own reaction coordinate — which is why coherence survives at all.

The Energy Ledger, Worked Through

Start with the photon. E = hc/λ, and hc = 1,239.8 eV·nm, so a 500 nm photon carries 1,239.8/500 = 2.48 eV. Multiply by 96.485 kJ mol⁻¹ per eV and you get 239 kJ/mol, or 57 kcal/mol per einstein (one mole of photons) — the whole budget for everything downstream.

Bathorhodopsin banks ~32 kcal/mol, about 56% of the photon's energy — photocalorimetric and photoacoustic measurements cluster at 32–35 kcal/mol out of the 57 kcal/mol absorbed. It is stored as torsional strain in the twisted polyene plus charge separation across the stretched Schiff base–Glu113 ion pair, and no further light is needed: the rest of the cascade spends that spring thermally.

The protein also buys efficiency. Free in solution, a retinal protonated Schiff base takes ~3 ps to isomerize and does so with a quantum yield of only ~0.2, scrambling into several isomers. In rhodopsin the reaction is fifteen times faster and the quantum yield is 0.65 — about 2 of every 3 absorbed photons isomerize — with essentially one product. Warshel's bicycle-pedal model (1976) explains why: the pocket enforces a volume-conserving, concerted counter-rotation about adjacent double bonds, so the two ends of the molecule barely translate. The β-ionone ring stays clamped in its hydrophobic slot against Trp265 while the C12–H and the C20 methyl on C13 swing through. A naive one-bond flip would have to sweep half the chromophore through the protein wall, which is why the pocket makes the reaction faster than free solution rather than slower.

From Bathorhodopsin to Metarhodopsin II

The protein then relaxes through a well-catalogued series of intermediates, each with its own absorption maximum. Yoshizawa and Wald (1963) mapped them by freezing the pigment in glycerol glasses and warming it in stages — bathorhodopsin survives at liquid-nitrogen temperature, lumirhodopsin below about −40 °C, metarhodopsin I below about −15 °C.

• Bathorhodopsin (~529 nm), formed in picoseconds — strained all-trans, Schiff base still protonated.
• Lumirhodopsin (~492 nm), ~150 ns — the chromophore relaxes, strain begins to move into the protein.
• Metarhodopsin I (~478 nm), microseconds (tens of µs at physiological temperature) — helices start to rearrange.
• Metarhodopsin II (λmax 380 nm), ~1 ms at 37 °C — the Schiff base deprotonates onto Glu113. Losing the positive charge collapses the conjugation, the pigment turns colourless to the eye, and the receptor opens: helix 6 swings outward by roughly 5–6 Å, exposing the cytoplasmic cleft.

Metarhodopsin II is the active species: it binds transducin, the rod G protein, and catalyses GDP→GTP exchange at hundreds of transducins per second, while rhodopsin kinase and arrestin quench it within tens of milliseconds, so one Rh* turns over on the order of tens of transducins before shutoff. Each activated transducin uncages a PDE6 subunit that hydrolyses roughly 10³ cGMP per second, cyclic-GMP-gated channels close, and one absorbed photon produces a current change near 1 pA. The gain is entirely chemical: GTP and cGMP pay for the amplification, not the photon.

How It Is Measured — and How Vision Is Specified

The mechanism was won with successively faster instruments. Low-temperature UV-vis trapped the intermediates. Femtosecond pump-probe transient absorption with sub-20 fs pulses resolved the 80 fs passage through the conical intersection (Polli, 2010). Femtosecond stimulated Raman spectroscopy caught the HOOP modes appearing in real time (Kukura et al., Science, 2005) — vibrational structure, not just kinetics. Palczewski and colleagues solved bovine rhodopsin to 2.8 Å in 2000 (PDB 1F88), the first GPCR structure ever determined, followed by bathorhodopsin (2G87) and metarhodopsin II (3PXO), and in 2023 a SwissFEL free-electron-laser experiment (Gruhl et al., Nature) caught rhodopsin by room-temperature time-resolved serial crystallography at 1 ps.

Function is measured electrically: suction-electrode recording from a single rod (Baylor, Lamb and Yau, 1979) resolves the single-photon response directly, and the clinical standard is the ISCEV dark-adapted 0.01 electroretinogram, whose rod-driven b-wave needs 20–30 minutes of dark adaptation. The chemistry is written into photometric law: the CIE scotopic luminous efficiency function V′(λ) peaks at 507 nm — rhodopsin's 500 nm absorption filtered by the eye's own optics.

Worked example. Rhodopsin's molar extinction coefficient at 500 nm is 40,600 M⁻¹cm⁻¹ (Wald and Brown, 1953). A human rod outer segment is ~25 µm long with rhodopsin at ~3 mM averaged over its volume, so the axial absorbance is A = εcl = 40,600 × 0.003 × 0.0025 cm ≈ 0.30. Transmission is 10⁻⁰·³⁰ ≈ 0.50, so about half of the 500 nm photons entering a rod end-on are absorbed. Multiply by the quantum yield 0.65 and roughly one incident photon in three produces an isomerization — which is why Hecht, Shlaer and Pirenne (1942) could place absolute visual threshold at only 5–14 absorbed quanta.

Dark Noise and the 420-Year Barrier

The same bond can flip without light, and that sets the floor on how well any eye can see. Thermal isomerization in the dark occurs about once per 420 years per rhodopsin molecule. A rod holds ~10⁸ rhodopsins, so roughly one spontaneous event every two minutes leaks through in each cell (10⁸ × 7.5×10⁻¹¹ s⁻¹ ≈ 0.008 s⁻¹, matching the 0.006–0.01 s⁻¹ dark events recorded from primate rods) — electrically indistinguishable from real photon responses.

Worked example. A 420-year lifetime is 420 × 3.156×10⁷ s = 1.33×10¹⁰ s, so the rate constant is k ≈ 7.5×10⁻¹¹ s⁻¹. Put that into the Arrhenius form with a conventional 10¹³ s⁻¹ attempt frequency: Ea = RT·ln(A/k) = (1.987×10⁻³ kcal mol⁻¹ K⁻¹ × 310 K) × ln(1.3×10²³) ≈ 0.62 × 53 ≈ 33 kcal/mol. That is a formidable barrier — more than half the photon's 57 kcal/mol. (Measured activation energies from rod dark-noise experiments come out lower, near 20–25 kcal/mol, so the accepted picture has the thermal route going through a vibrationally hot ground state with a much smaller effective prefactor; Luo and colleagues argued in 2011 that light and heat trigger the very same molecular event.)

The trade-off shows across pigments: long-wavelength cone pigments sit behind a lower thermal barrier, so they isomerize spontaneously orders of magnitude more often — one reason cones are far less sensitive in dim light. Mutations that weaken the Glu113 salt bridge, such as G90D and A292E, do it chemically: the pigment turns constitutively active in the dark, the rods behave as though under permanent dim illumination, and the result is congenital stationary night blindness. The P23H mutation fails one step earlier — it never folds around a chromophore at all, and is the commonest dominant cause of retinitis pigmentosa in the US.

Failure Modes, Look-alikes and Misconceptions

It is not the enzymatic re-isomerization. The usual confusion is with RPE65, the isomerohydrolase in the retinal pigment epithelium that drives all-trans-retinyl ester back to 11-cis-retinol on the energy of ester hydrolysis — thermal, enzymatic, minutes long, and the real limit on dark adaptation: the 30–40 minutes it takes to see properly after a camera flash is RPE65 turnover, not anything femtosecond. Biallelic RPE65 mutations cause Leber congenital amaurosis type 2, the target of voretigene neparvovec (2017), the first FDA-approved gene therapy for an inherited disease.

It is not microbial rhodopsin. Bacteriorhodopsin isomerizes all-trans → 13-cis in ~0.5 ps at similar quantum yield, but it pumps protons instead of activating a G protein and re-isomerizes thermally in milliseconds — no visual cycle.

Bleaching does not destroy retinal. The pigment loses visible colour because the Schiff base deprotonates at metarhodopsin II (500 nm → 380 nm), not because anything is consumed. The retinal is eventually hydrolysed off intact and recycled.

The photon does not power the signal. Hubbard and Wald established that light's only job is the cis–trans flip — Wald shared the 1967 Nobel Prize in Physiology or Medicine for that line of work.

Free all-trans-retinal is toxic. If it is not cleared fast enough it condenses with phosphatidylethanolamine into the bis-retinoid A2E, which accumulates as lipofuscin in the pigment epithelium — the chemistry behind Stargardt disease when the ABCA4 flippase fails. Upstream, vitamin A deficiency (serum retinol below 0.70 µmol/L) starves the cycle of chromophore and causes night blindness, still a leading preventable cause of childhood blindness.

Retinal isomerization compared with the reactions it is most often confused with

Process	Isomerization	Driving energy	Timescale and yield
Rhodopsin (vision)	11-cis → all-trans	One 500 nm photon	~200 fs, quantum yield 0.65
Free retinal Schiff base in methanol	11-cis → all-trans (mixed products)	One photon	~3 ps, quantum yield ~0.2
Bacteriorhodopsin (proton pump)	all-trans → 13-cis	One 568 nm photon	~0.5 ps, quantum yield ~0.6
RPE65 in the pigment epithelium	all-trans-retinyl ester → 11-cis-retinol	Ester hydrolysis (enzymatic, thermal)	Minutes; sets dark-adaptation speed
Thermal activation in the dark	11-cis → all-trans	Ambient thermal energy	Once per 420 years per rhodopsin