The role of local and remote amino acid substitutions for optimizing fluorescence in bacteriophytochromes: A case study on iRFP

Bacteriophytochromes are promising tools for tissue microscopy and imaging due to their fluorescence in the near-infrared region. These applications require optimization of the originally low fluorescence quantum yields via genetic engineering. Factors that favour fluorescence over other non-radiative excited state decay channels are yet poorly understood. In this work we employed resonance Raman and fluorescence spectroscopy to analyse the consequences of multiple amino acid substitutions on fluorescence of the iRFP713 benchmark protein. Two groups of mutations distinguishing iRFP from its precursor, the PAS-GAF domain of the bacteriophytochrome P2 from Rhodopseudomonas palustris, have qualitatively different effects on the biliverdin cofactor, which exists in a fluorescent (state II) and a non-fluorescent conformer (state I). Substitution of three critical amino acids in the chromophore binding pocket increases the intrinsic fluorescence quantum yield of state II from 1.7 to 5.0% due to slight structural changes of the tetrapyrrole chromophore. Whereas these changes are accompanied by an enrichment of state II from ~40 to ~50%, a major shift to ~88% is achieved by remote amino acid substitutions. Additionally, an increase of the intrinsic fluorescence quantum yield of this conformer by ~34% is achieved. The present results have important implications for future design strategies of biofluorophores.


Nomenclature of chromophore geometry
The geometry of the biliverdin chromophore is denoted according to the configuration (Z/E; Z -"zusammen", E -"entgegen") and conformation (s/a; s -syn, a -anti) of the methine bridges between the pyrrole rings A, B, C, and D. In the Pr ground state, the configuration is ZZZssa, in the Pfr state ZZEssa. In the protein, the cofactor is covalently attached to a cysteine via the ethylidene substituent at ring A. The C atoms in rings A-D are numbered for reference. Photon absorption leads to a photoinduced Z/E isomerization of the chromophore around the methine bridge between rings C and D.

Fluorescence data analysis
The fluorescence decays were analyzed employing a Levenberg-Marquardt algorithm for the minimization of the reduced χ r 2 after iterative reconvolution with the instrumental response function The quality of the fit was judged by the value of χ r 2 and by the degree of randomness of residuals (difference between the experimental data points and the fit at each time point ν t ) to check for the absence of any correlation of the deviations in a certain time interval. For this judgment, the autocorrelation function of the residuals was calculated, which was around 1.0 for the whole time interval. For these calculations, the software of Globals Unlimited® (University of Illinois, Urbana, USA) was used. It was found that a sufficient quality of the fit could be achieved for fitting with n=3 decay components (χ r 2 = 1.05). No significant improvement was obtained with n>3 decay components. Figure S1. Normalized absorption spectra of P2PG, iRFP, and the mutants obtained via route A (right) and route B (left) showing the typical Soret (around 400 nm) and Q bands (around 700 nm) due to absorption of the biliverdin chromophore. Spectra measured before and after red-light irradiation (660 nm LED) are shown in black and right, respectively. Note that the photoactive variants only undergo a phototransformation to the Meta-R state which, due to the low extent of photoconversion, is typically reflected only by a decrease of the Q-band absorption. All spectra were normalized to the protein absorption peak at 280 nm. Subsequently, iRFP was used as a reference for determining the fluorescence quantum yields of the other phytochrome variants.

RR spectra in the low frequency region
Below 700 cm −1 there are three medium intense Raman bands which may be assigned in analogy to previous vibrational analysis of the BV chromophore in the Pfr state of Agp2 (see manuscript). In that case, theoretical calculations predicted modes of considerable Raman intensity in this region that involve torsional and out-of-plane deformation coordinates of rings D, C, and ring B, as well as the N-H out-of-plane coordinate of ring B. Guided by the predicted Raman intensities and frequencies, we assign the peak at 673 cm −1 (674 cm −1 ) of iRFP (P2-PG) to two closely spaced modes involving mainly torsional and out-of-plane deformation coordinates of rings C and D ( Figure S3). The 651 cm −1 band in the spectrum of iRFP (656 cm −1 for P2-PG) displays a small upshift upon H/D exchange which can be rationalized in terms of the involvement of the N-H out-of-plane coordinates in H 2 O but not in D 2 O. An even larger contribution of this coordinate is likely to be the origin for the disappearance of the 659 cm −1 band of iRFP (663 cm −1 in P2-PG) in D 2 O, where a new band at distinctly higher frequencies (690 cm −1 ) is detected instead. We therefore tentatively assign the 659cm −1 band (iRFP) to a mode of significant N-H out-of-plane character, presumably localized at ring B, and the 651-cm −1 band to a torsional and out-of-plane deformation mode involving ring D (vide supra).
Inspection of the spectral changes brought about by the substitutions according to route A and B shows that the upshift of the HOOP mode from 810 cm −1 (P2-PG) to 813 cm −1 (iRFP) is produced by the single mutations at either the position 202 (D202T) or 258 (Y258F) ( Figure S4). The latter replacement is also responsible for the intensity increase of the 663 cm −1 band whereas the frequency shift of this band largely depend on the remote amino acid substitutions. Figure S3. RR spectra of P2PG and iRFP in the HOOP region, measured from samples in H 2 O (black) and D 2 O (red). Figure S4. RR spectra of the P2-PG variants obtained via route A (left) and route B (right), compared with the spectra of the wild-type P2-PG and iRFP. The spectra, measured from the proteins in H 2 O, display the region of the HOOP modes.

IR difference spectra
Infrared difference spectroscopy IR difference spectroscopy can be employed only to those variants which undergo a photoisomerisation ( Figure S5). Due to the lack of the PHY domain, the spectra of these variants do not display any major difference signals in the amide I and amide II region. However, the C=O stretching mode of ring D, which in the Pr state of P2PG is observed at 1711 cm −1 , can clearly be identified in all photoactive variants, albeit at slightly higher frequencies. The corresponding ring A C=O stretching at 1737 cm −1 (P2PG) can barely be detected suggesting that the position of this mode remains largely unchanged upon photoconversion. Figure S5. IR difference spectra obtained by subtracting the spectrum of the parent Pr state from that of the photoproduct obtained by 660 nm LED irradiation. Thus, negative and positive bands refer to the parent state and the photoproduct (usually a Meta-R state), respectively. The signals of the ring D and A carbonyl functions are observed between 1710 and 1725 cm −1 and above 1730 cm −1 , respectively. As expected, no major signals are observed in the amide I region.
6. RR spectra of further mutants Figure S6. RR spectra of various mutants derived from P2PG.