Twisted Carotenoids Do Not Support Efficient Intramolecular Singlet Fission in the Orange Carotenoid Protein

Singlet exciton fission is the spin-allowed generation of two triplet electronic excited states from a singlet state. Intramolecular singlet fission has been suggested to occur on individual carotenoid molecules within protein complexes provided that the conjugated backbone is twisted out of plane. However, this hypothesis has been forwarded only in protein complexes containing multiple carotenoids and bacteriochlorophylls in close contact. To test the hypothesis on twisted carotenoids in a “minimal” one-carotenoid system, we study the orange carotenoid protein (OCP). OCP exists in two forms: in its orange form (OCPo), the single bound carotenoid is twisted, whereas in its red form (OCPr), the carotenoid is planar. To enable room-temperature spectroscopy on canthaxanthin-binding OCPo and OCPr without laser-induced photoconversion, we trap them in a trehalose glass. Using transient absorption spectroscopy, we show that there is no evidence of long-lived triplet generation through intramolecular singlet fission despite the canthaxanthin twist in OCPo.

strate (S151, Ossila; 15×20×1.1 mm) and 40 µL of the protein-trehalose mixture drop-cast in the center of the imaging spacer. The substrate was then placed in vacuum as above; pressure was released under a continuous flow of ultra-pure nitrogen gas and a glass microscope cover slip (ThermoScientific; 22×22 mm, No.1 thickness) was attached to the upper imaging spacer.

S1.2 Steady-state absorbance spectroscopy
Steady-state absorbance of samples were measured in a commercial Cary double-beam spectrometer (Cary 60 UV-Vis Spectrophotometer, Agilent Technologies). Both zero and baseline corrections were applied with a blank sample; in the case of trehalose-encapsulated samples, a trehalose blank is used.

S1.3 Transient absorption spectroscopy
Picosecond transient absorption spectroscopy was undertaken with a commercial spectrometer (Helios, Ultrafast Systems) outfitted with a Ti:Sapphire seed laser (MaiTai, Spectra Physics) providing 800 nm pulses (84 MHz, 25 fs nominal FWHM) and a Ti:Sapphire chirped-S4 pulse amplifier (Spitfire Ace PA-40) amplifying 800 nm pulses (10 kHz, 12 W average power, 40 fs nominal FWHM). Tunable pump pulses for excitation were generated by seeding a part of the 800 nm beam in an optical parametric amplifier (TOPAS Prime, Light Conversion). An optical chopper was used to modulate the pump frequency to 5 kHz. An intensity spectrum for the pump used in visible-probe measurements is shown in Figure S4. Pump beam spot sizes were measured at the sample position with a CCD beam profiler (BC106N-VIS/M, Thorlabs), and used in subsequent calculations to tune a 200 µJ cm −2 pump fluence.
Supercontinuum probes were generated with a part of the 800 nm pulse focused on either a sapphire crystal for visible probes (450-800 nm) or a YAG crystal for NIR probes (800-1600 nm). Pump-probe delay was controlled with a motorized delay stage with a random stepping order. The signal was dispersed with a grating and detected with CMOS or InGaAs sensors for visible or NIR probes, respectively. The pump and probe polarizations were set to the magic angle. Surface Xplorer 4.3.0 (Ultrafast Systems) was used in processing the transient absorption datasets. Noisy edges of the spectra were trimmed, and the program's bad spectra replacement procedure was applied. A background correction ('subtract scattered light') was then applied using the spectra before any apparent response from the sample. For visible-probe data, chirp correction was applied, choosing points at the first apparent signal for a given kinetic. For NIR-probe data, no chirp was discernible, hence no chirp correction was applied.
Time zero is adjusted to the time of maximum initial signal. Further processing and some analysis was performed with original Python code.

S6
S2 Additional steady-state absorbance data  Figure S3: Resonance Raman spectra of blank trehalose glass and OCPo in trehalose. The blank trehalose spectrum (black) is almost entirely noise, confirming that the trehalose-sucrose glass is not contributing significant Raman signal in any of the samples incorporating trehalose glass. OCPo in trehalose (green) is shown as a comparison, normalized to the peak ν 1 intensity. The blank spectrum has been scaled to give a similar noise magnitude at the high-wavenumber end. Both spectra are single scans (not averaged). Measurements were undertaken at room temperature. A 532 nm pump was used.  Figure S4: Intensity spectrum of the pump used in the transient absorption experiments while using the visible probe. This 532 nm-centered (set in WinTopas4) excitation profile was used in taking the visible-probe data shown in main text Figure 3 and Figure 4. The 532 nm-centered profile used in taking the NIR-probe data was not taken, and may have had a slightly different spectrum owing to day-to-day variation in the tunable pump generation. The spectrum has been normalized to the maximum intensity. S10

S4.2 Global lifetime analysis
Global lifetime analysis on the visible-probe transient absorption data of OCPo and OCPr was performed. This was done using the Glotaran 1.5.1 software package (http://glotaran.org), 5 a GUI for the R package TIMP. 6 Data used had already been processed with the steps outlined in Section S1.3; in particular, a chirp correction had already been applied, so that a term to account for chirp did not need to be included in the fitting.
Noisy regions in the data due to pump scatter were excluded for all times to ensure a good fit of the rest of the data. Noisy red and blue ends in the data associated with tails of the probe were also excluded, so that the fitted wavelengths were 430 nm to 780 nm. The fitting was weighted favourably at later delay times for good fits of any long-lived features; Table S1 shows the weighting applied. Due to the strong coherent artifact feature in the first 0.5 ps, only data beyond that time was fitted. Thus, in the model, terms to account for the coherent artifact and S 2 states were not included. This left a relatively simple fitted model of a number of wavelength-dependent decay-associated difference spectra (DADS) decaying exponentially in parallel. Table S1: Weightings applied to time-ranges of the visible ps transient absorption data for the global lifetime analysis. Note that only data >0.5 ps was fitted, and that the maximum time delay in these experiments was ∼40 ps. OCPr data are shown in Figure S5 and S6, respectively. Fitting a 2-component parallel decay model in an artifact-free region of the visible-probe data beyond the initial coherent S11 artifact and S 2 -associated response gives two decay-associated difference spectra (DADS)  Only the wavelength range 430-780 nm and times >0.5 ps were fitted, and noisy data from 520.5-549.5 nm due to significant pump scatter was excluded from the fit. DADS lifetimes are specified in the legend. Residuals = Data − Fit. See text for further details.
We note that sample degradation (caused the ∼200 µJ cm −2 pump fluence) likely affects S12 the fitted time constants and DADS profiles, and that the maximum time delay used was about 40 ps. S13 Only the wavelength range 430-780 nm and times >0.5 ps were fitted, and noisy data from 522-548 nm due to significant pump scatter was excluded from the fit. DADS lifetimes are specified in the legend; multiplications refer to scalings applied to the DADS. Residuals = Data − Fit. See text for further details.