Structure of a Zinc Porphyrin-Substituted Bacterioferritin and Photophysical Properties of Iron Reduction

Abstract

The iron storage protein bacterioferritin (Bfr) binds up to 12 hemes b at specific sites in its protein shell. The heme b can be substituted with the photosensitizer Zn(II)-protoporphyrin IX (ZnPP), and photosensitized reductive iron release from the ferric oxyhydroxide {[FeO(OH)]_n} core inside the ZnPP-Bfr protein shell was demonstrated [Cioloboc, D., et al. (2018) Biomacromolecules 19, 178–187]. This report describes the X-ray crystal structure of ZnPP-Bfr and the effects of loaded iron on the photophysical properties of the ZnPP. The crystal structure of ZnPP-Bfr shows a unique six-coordinate zinc in the ZnPP with two axial methionine sulfur ligands. Steady state and transient ultraviolet–visible absorption and luminescence spectroscopies show that irradiation with light overlapping the Soret absorption causes oxidation of ZnPP to the cation radical ZnPP^•+ only when the ZnPP-Bfr is loaded with [FeO(OH)]_n. Femtosecond transient absorption spectroscopy shows that this photooxidation occurs from the singlet excited state (¹ZnPP*) on the picosecond time scale and is consistent with two oxidizing populations of Fe³⁺, which do not appear to involve the ferroxidase center iron. We propose that [FeO(OH)]_n clusters at or near the inner surface of the protein shell are responsible for ZnPP photooxidation. Hopping of the photoinjected electrons through the [FeO(OH)]_n would effectively cause migration of Fe²⁺ through the inner cavity to pores where it exits the protein. Reductive iron mobilization is presumed to be a physiological function of Bfrs. The phototriggered Fe³⁺ reduction could be used to identify the sites of iron mobilization within the Bfr protein shell.

Graphical Abstract

Zinc(II) porphyrins (ZnPs) have a long history as photosensitizers of redox reactions.^1,2 These reactions typically occur via either oxidation or reduction of the photogenerated ³ZnP* or ¹ZnP* states. Naturally occurring ZnPs tend to aggregate in aqueous solution, which greatly lowers the photosensitization efficiency. Embedding these ZnPs within heme binding sites of proteins inhibits aggregation, and several ZnP-substituted proteins have been shown to function as both reductive and oxidative photo-sensitizers.^3-7

Clark and Kurtz reported that Zn(II)-protoporphyrin IX (ZnPP) can be quantitatively inserted into the heme binding sites of the iron storage protein bacterioferritin (Bfr).⁸ The native Bfr structure consists of an approximately 12 nm diameter protein shell containing 24 identical subunits (Figure 1A) arranged as 12 head-to-tail subunit pairs, each sandwiching an iron-protoporphyrin IX (heme b) (Figure 1B,C).⁹ The iron of heme b is axially coordinated by two methionine sulfurs (Figure 1C), one from each subunit at the dimer interface. The heme b propionate substituents protrude into the ~8 nm diameter interior cavity, which can store up to ~3000 irons as a ferric oxyhydroxide polymer {core iron or [FeO(OH)]_n}.¹⁰ This core iron forms via initial autoxidation of Fe²⁺ at a ferroxidase center (FC) within in each subunit, and a nonfunctional FC severely limits core formation. The FC can be occupied by up to two irons (shown as orange spheres in Figure 1C) located symmetrically on either side of a heme b. The heme b in Bfr facilitates reduction and release of core iron and may function as an ET mediator from exogenous reductants.^11-13 The structure of ZnPP is identical to that of heme b shown in Figure 1B, except that Zn(II) is substituted in place of iron. Clark and Kurtz¹⁴ also prepared an iron-free dimeric form of ZnPP-Bfr (ZnPP-Bfr dimer) and showed that it can serve as a photosensitizer for platinum nanoparticle-catalyzed H₂ generation in aqueous solution.¹⁴ Our photophysical studies of the ZnPP-Bfr dimer¹⁵ indicated that this photosensitization occurred as diagrammed in Scheme 1 via either an oxidative quenching pathway in which ³ZnPP* is oxidized to the ZnPP cation radical, ZnPP^•+, or a reductive quenching pathway in which ³ZnPP* is reduced to the ZnPP anion radical, ZnPP^•−, by a sacrificial electron donor (SED). The singlet excited state, ¹ZnPP*, may also participate. Completing the redox cycle requires the presence of a SED.

Figure 1.

Relevant structural features of Escherichia coli Bfr. (A) Cartoon rendering of the 24-mer protein shell with two head-to-tail dimer subunits highlighted as a blue/green pair and the 12 hemes b highlighted as yellow atom spheres. (B) Heme b structure. (C) One of the 12 head-to-tail homodimers viewed from the interior cavity and approximately perpendicular to the plane of the embedded heme b (stick rendered). Two iron atoms in the FC of each subunit are shown as orange spheres. Protein drawings were generated in PyMOL (Schrödinger, LLC) using coordinates from Protein Data Bank entry 3e1n.⁹

Scheme 1.

Oxidative and Reductive Quenching Pathways for Photosensitized H₂ Production by the ZnPP-Bfr Dimer and Pt NPs

Porphyrins are also traditional photodynamic therapy drugs, their efficacy being attributed primarily to toxic singlet oxygen generation.^16-18 However, singlet oxygen generation is inefficient in hypoxic tumor environments.¹⁹ Cioloboc et al.²⁰ showed that the ZnPP-Bfr 24-mer could be loaded with ~2500 irons within the protein shell as [FeO(OH)]_n analogously to the native Bfr. They further showed that irradiation with purple light overlapping the ZnPP Soret absorption in the presence of the biological reducing agent, NADH, under a N₂ atmosphere, caused release of up to 100% of the iron as Fe²⁺, which reacted with hydrogen peroxide to produce hydroxyl radical. Under an aerobic atmosphere, singlet O₂ was also produced. Purple light irradiation of melanoma cells that had taken up the iron-loaded ZnPP-Bfr resulted in intracellular iron release and decreased cell survival compared to unirradiated cells or of irradiated cells without the ZnPP-Bfr. The system is diagrammed in Scheme 2. Other “photo-Fenton” therapy approaches have recently been reported.²¹

Scheme 2.

Photosensitized Fe²⁺ Release and Production of Hydroxyl Radical and Singlet O₂ from Iron-Loaded ZnPP-Bfr

Although the heme in Bfr is well established to have bis(methionine) axial ligated to the iron (Figure 1C), the axial ligation state of the ZnPP in ZnPP-Bfr 24-mer has not been established. In the presence of coordinating ligands, Zn in ZnPs is typically five-coordinate with one axial nitrogen or oxygen ligand.^4,22,23 Six-coordinate Zn in ZnPs is relatively uncommon,^24-28 and ligation of thioether to ZnPs is rare.^3,29-31 X-ray absorption studies of the ZnPP-Bfr dimer at 18 K showed a mixture of five- and six-coordinate Zn in ZnPP with both short (2.4 Å) and long (2.8 Å) Zn–S distances.¹⁵ The work presented here assesses the structure of ZnPP in the ZnPP-Bfr 24-mer (hereafter simply termed ZnPP-Bfr) as well as the photophysical properties of the iron-loaded ZnPP-Bfr to better understand the process of photosensitized reductive iron release.

MATERIALS AND METHODS

Reagents and General Methods.

Protoporphyrin IX was from Frontier Scientific. All other chemicals were purchased from either Fisher Scientific or Sigma-Aldrich at the highest available purities. All aqueous solutions were prepared in water that had been passed through a Milli-Q ultrapurification system (Merck Millipore, Inc.) to achieve a resistivity of 18 MΩ. ZnPP was prepared by metalation of protoporphyrin IX with zinc acetate dihydrate.⁸ Expression, isolation, purification, and quantification methods for Escherichia coli ZnPP-Bfr have been reported previously.⁸ Samples for metal quantification were prepared by mixing 100 μL of the protein solution (containing ~10 μM ZnPP) with 500 μL of freshly prepared aqua regia and then further diluting the mixture to a 7 mL total volume with water. Metal concentrations were quantified by inductively coupled plasma optical emission spectrometry (ICP-OES) at the University of Texas at San Antonio Department of Civil and Environmental Engineering. The ZnPP-containing proteins were stored in aluminum foil-covered containers and manipulated in low-light environments.

Iron Loading and Sample Preparation.

Iron-loaded ZnPP-Bfr (ZnPP-Bfr-Fe) was prepared from the as-isolated ZnPP-Bfr (ZnPP-Bfr-ai) as described previously²⁰ by addition of multiple aliquots of an aqueous ferrous ammonium sulfate stock solution to air-saturated solutions containing ~1 μM Bfr 24-mer in 50 mM 3-morpholinopropane-1-sulfonic acid (MOPS) (pH 6.5). A sample of ZnPP-Bfr-ai was similarly treated with a single aliquot of a ferrous ammonium sulfate stock solution to achieve a molar ratio of added iron to Bfr 24-mer of 200. This mixture was incubated for 30 min at room temperature and then centrifuged at 5500g for 10 min. Some experiments used ZnPP-Bfr-ai that had been further treated with sodium dithionite and EDTA to remove residual iron, as described previously.³² This iron-depleted ZnPP-Bfr is hereafter termed Fe_depl-ZnPP-Bfr. For photophysical and photochemical experiments, the proteins were exchanged into 150 mM NaCl by being passed over desalting columns. The solutions were then transferred to a Vacuum Atmospheres glovebox and allowed to equilibrate with the anaerobic N₂ atmosphere.

X-ray Crystallography.

Automated screening for crystallization of an analytical gel filtration-purified preparation of ZnPP-Bfr-ai was carried out using the sitting drop vapor diffusion method with an Art Robbins Instruments Phoenix system in the X-ray Crystallography Core Laboratory at the University of Texas Health Science Center at San Antonio (UT Health San Antonio). Crystals were obtained at 22 °C from the Qiagen pH Clear II Suite condition 88 [2.4 M sodium malonate (pH 6.0)] by mixing crystallization reagents with a protein solution containing 50 mM MOPS (pH 7) in a 1:1 (v/v) ratio and a total drop volume of 400 nL. Crystals were flash-cooled in liquid nitrogen prior to data collection at beamline 24-ID-E of the Advanced Photon Source (Argonne, IL). Diffraction data were processed using XDS³³ to a resolution limit with a CC_1/2 of 0.48.³⁴ Additional data were acquired at beamline 24-ID-C of the Advanced Photon Source for a second crystal from the same crystal drop near the zinc absorption edge (1.2822 Å, 9670 eV) and at a low energy remote (1.3890 Å, 9400 eV) to verify the identity of zinc ions in the structure. These data were processed using xia2.^33,35-37 The structure was determined by the molecular replacement method implemented in PHASER³⁸ using coordinates from Protein Data Bank (PDB) entry 2Y3Q³⁹ as the search model. Model coordinates were refined using PHENIX,⁴⁰ including simulated annealing, and alternated with manual rebuilding using Coot.⁴¹ Models was verified using composite omit map analysis.⁴² The zinc protoporphyrin IX molecules were modeled in two overlapping orientations rotated 180° about the zinc ion as they are able to adopt either orientation between Met ligands.⁴³ Data collection and refinement statistics are listed in Table 1. The crystallographic data have been deposited in the Protein Data Bank (PDB entries 6P8K and 6P8L).

Table 1.

Data Collection and Refinement Statistics

Data Collectiona
PDB entry	6P8K	6P8L
X-ray source	APS 24-ID-E	APS 24-ID-C	APS 24-ID-C
space group	P42212	P42212	P42212
cell dimensions
a, b, c (Å)	207.9, 207.9, 142.7	207.8, 207.8, 142.7	207.9, 207.9, 143.1
α, β, γ (deg)	90, 90, 90	90, 90, 90	90, 90, 90
wavelength (Å)	0.97918	1.2822	1.3190
resolution (Å)	19.96–1.70 (1.79–1.70)	92.65–2.10 (2.21–2.10)	92.98–2.50 (2.64–2.50)
Rpim	0.048 (1.039)	0.067 (0.689)	0.071 (0.712)
CC1/2	(0.48)	(0.69)	(0.74)
mean I/σI	11.3 (0.7)	8.2 (1.1)	8.7 (1.2)
completeness (%)	99.9 (100)	99.8 (99.9)	99.8 (99.8)
redundancy	13.5 (12.6)	7.3 (7.5)	7.2 (7.2)
Wilson value (Å2)	19.4	30.2	40.4
Refinement
resolution (Å)	19.96–1.70	92.98–2.11
no. of reflections	337073	178718
Rwork, Rfree	0.192, 0.218	0.213, 0.240
no. of monomers per asymmetric unit	12	12
no. of atoms
protein	15832	15788
ligand	735	734
solvent	2530	1598
B-factor (Å2)
protein	21.3	34.3
ligand	36.2	43.3
solvent	36.8	44.8
root-mean-square deviation
bond lengths (Å)	0.010	0.007
bond angles (deg)	2.007	1.972
Ramachandran plot (%)
favored	99.8	99.6
allowed	0.2	0.4
outliers	0.0	0.0

^aValues in parentheses are for the highest-resolution shell.

Ultraviolet–Visible (UV–vis) Absorption Spectral Time Courses.

ZnPP-Bfr solutions in 150 mM NaCl (pH 7) were made anaerobic by being placed in a glovebox (Vacuum Atmospheres, Inc.) exposed to the N₂ atmosphere for 8–12 h. The solutions were adjusted to 1.5–5 μM ZnPP and placed in 1 cm path length quartz cuvettes. For irradiation, the cuvettes were sealed with a screw cap and rubber septum and removed from the glovebox. Irradiation was carried out as described previously²⁰ at room temperature using a 300 W halogen lamp focused through a slide projector lens with the samples in 1 cm path length quartz cuvettes placed 10 cm from the lens. The light was passed through a HOYA 62 mm UV-IR multicoated filter with cutoffs below 390 nm and above 700 nm. A 426 ± 25 nm band-pass filter (90 mm by 40 mm, Optical Filter Shop) was placed between the projector lens and sample cuvette. UV–vis absorption spectra were recorded on an Ocean Optics Flame fiber optic spectrophotometer.

Transient Photophysical Methods.

ZnPP-Bfr solutions (~3 μM in ZnPP) were used for all experiments. All experiments were conducted at room temperature (~23 °C). Steady state photoluminescence spectra were recorded on an Edinburgh FLS1000 photoluminescence spectrometer. Fluorescence dynamics were measured by time-correlated singlephoton counting (TCSPC) in a PicoQuant FluoTime 300 fluorescence lifetime spectrophotometer. The excitation source was a Coherent Chameleon Ti:sapphire ultrafast laser. The Chameleon laser provides 120 fs pulses at an 80 MHz repetition rate that can be tuned from 680 to 1080 nm. The 80 MHz pulse train was sampled using a Coherent model 9200 pulse picker to provide excitation pulses at a 4 MHz repetition rate and was frequency doubled using a Coherent second-harmonic generator. All samples were placed in 1 cm quartz cuvettes. An excitation wavelength of 430 nm was chosen for all samples, and the instrument response function was measured using a Ludox scattering solution.

Nanosecond transient absorption spectroscopy measurements were performed on an Edinburgh L900 transient absorption spectrometer. An Opotek Radiant 335 tunable laser was used as the excitation source. The Opotek pump laser was combined with an optical parametric oscillator in a single closed compartment and has tunable range from 410 to 710 nm, pulse lengths of 4–6 nm, and a repetition rate of ≤10 Hz. Experiments were carried out typically at pulse energies of 4–5 mJ. Spectra were recorded using a 150 W, ozone-free xenon flash lamp as the probe source and the PMT and iCCD camera as the detectors. All data were collected using the xenon arc lamp as a probe in pulsed mode up to 10 Hz. The pulse is capable of supplying an additional current of 100 A for up to 6 ms. The iCCD camera (Andor DH320T) was used as the detector for the spectral mode to record transient absorption (TA) spectra, and a PMT (Hamamatsu R928) was used for the kinetic mode to measure lifetimes. The TA spectra were recorded from 350 to 850 nm with a 4 or 50 ns initial camera delay and with different subsequent delay times depending on the triplet lifetimes. The samples for nanosecond TA spectroscopy were prepared in anaerobic aqueous solutions with an optical density of ~0.5 at the excitation wavelength (434 nm) in a 10 mm path length quartz cuvette.

Ultrafast pump–probe experiments were performed using femtosecond transient absorption spectroscopy on an Ultrafast TA system with broadband capabilities. This system is equipped with a Coherent Astrella laser (Ti:sapphire amplifier, 5 W CW output), an ultrafast optical parametric amplifier (Coherent OPerA Solo OPA), and a Helios Fire transient absorption spectrometer. The Astrella laser emits a fundamental beam of 800 nm with a pulse duration of 100 fs and and a pulse repetition rate of 1 kH direct into the OPA, which was tuned to generate a bean light of 434 nm. The pump beam was then directed into a Helios Fire (Ultrafast Systems) automated femtosecond transient absorption spectrometer where the beam passed through a mechanical chopper, a depolarizer, and a neutral density filter to tune the beam to 200 nJ/pulse of power before hitting the stirred sample (~0.5 OD). Residual 800 nm light was directed into an automated 8 ns delay stage on the Helios Fire instrument. Afterward, the beam was focused into a calcium fluoride crystal to generate a visible probe ranging from 420 to 700 nm. The pump and broadband probe beams were overlapped both spatially and temporally on the sample solution, and the transmitted probe light from the samples was collected on the broadband UV–vis detectors to record the time-resolved excitation difference spectra. Kinetic traces at specific wavelengths were constructed by extracting data from the multiwavelength spectral data taken at many different pump–probe delay times. All spectral data were processed and analyzed using Surface Xplorer.

RESULTS

X-ray Crystal Structure of ZnPP-Bfr.

The crystal structure shows a 24-mer assembly containing 12 ZnPPs (Figure S1A) with overall structural features virtually identical to those described above and shown in Figure 1 for the heme-containing Bfr.^39,43 The 1.7 Å resolution model (PDB entry 6P8K) is used here to describe the ZnPP-Bfr structure. Figure 2 shows the ZnPP structure superimposed on a composite omit map.⁴² This map minimizes model bias going into the map calculation as the electron density region in view was calculated without using local model atom coordinates. Difference anomalous Fourier analysis of X-ray diffraction data collected at the Zn absorption edge and at a low-energy remote wavelength identified the metal atoms at the center of the porphyrin as zinc (Figure S2A, PDB entry 6P8L). We observed significant difference anomalous Fourier peaks (>3.0σ above background) at the Zn absorption edge but not at the discriminating low-energy remote wavelength, which indicates the metal identities are Zn while ruling out the presence of Fe (which would generate a significant peak at the low-energy remote wavelength). The ZnPPs are sandwiched at the interfaces of the 12 subunit dimers. Each ZnPP is axially coordinated by two Met52 Sγ atoms, one from each sandwiching subunit. The Zn–Met52 Sγ distances average 2.4 Å for the six crystallographically independent ZnPPs. The two Zn-S axes on opposite sides of the ZnPP plane are tilted approximately 10° toward opposite porphyrin meso carbons. The Sγ–Zn–Sγ axis leans ~166° toward the propionate side of the porphyrin ring. Although the fitted positions of the ZnPP zinc atoms lie approximately in the porphyrin plane, the oblong shape of the zinc anomalous difference electron density shown in Figure 2A indicates some out-of-plane disorder of the Zn. Two slightly differing positions of ZnPP at the subunit dimer interfaces were fitted to the electron density. These positions are related by a pseudo-2-fold rotation axis parallel to the porphyrin plane and lying between the propionate substituents. The analogous 2-fold positions have been previously reported for the heme in E. coli Bfr.⁴³ The propionate substituents of the ZnPPs protrude into the internal cavity of the 24-mer (Figure S1B). A water-filled cavity (not shown) is observed between the buried vinyl/methyl substituent side of the ZnPP and the 24-mer outer surface. The shortest Zn–Zn distance between any two ZnPPs in the 24-mer is ~42 Å, and the shortest interporphyrin ring carbon–carbon distance is ~35 Å Here again, these distances are very similar to those reported for the heme b-occupied Bfr.

Figure 2.

Two stereoviews of the ZnPP environment in the ZnPP-Bfr crystal structure approximately (A) parallel and (B) perpendicular to the porphyrin plane. The ZnPP carbon backbone is colored yellow with pyrrole nitrogens and propionate oxygens colored blue and red, respectively, and the gray colored crossing indicates the positions of the zinc atoms. Met52 side chains are shown with Sγ colored orange. A 2F₀ – F_c composite omit map, contoured at 1σ, is shown as blue mesh for the ZnPP and Met52 residues. Anomalous difference Fourier electron density (calculated for a wavelength of 0.97918 Å) is shown as red mesh. Portions of surrounding protein subunit helical backbones are shown as green and blue ribbons.

In the ZnPP-Bfr structure, a single metal atom identified as zinc by difference anomalous Fourier analysis at the zinc X-ray absorption edge was found in each FC (Figure S3). This zinc is pseudotetrahedrally coordinated by three Glu residues and one His residue that are known to provide ligands to one of the irons in the FC.⁹ Two FC zinc atoms are located 12–13 Å from the zinc atom in each ZnPP (Figure S3). Non-zinc/iron metal atoms were found to occupy each of six pores centered along the three 4-fold rotational axes of the protein shell. Each of these metal atoms is surrounded by four Nδ2 atoms of N148 residues lining the pores (Figure S1A). On the basis of other Bfr crystal structures⁴⁴ and the high concentration of Na⁺ in the ZnPP-Bfr crystallization buffer, these atoms were modeled as sodiums.

Iron and Zinc Contents.

Iron and zinc were the only metals found by ICP-OES in significant amounts in our ZnPP-Bfr preparations. The iron and zinc contents in the uncrystallized proteins are listed in Table 2. The source of the apparently “excess” Zn in ZnPP-Bfr-ai and ZnPP-Bfr-Fe [i.e., above the 10–12 expected from the ZnPP plus the 24 Zn observed to occupy the FC in our X-ray crystal structure (Figure S3)] is unclear. It could conceivably be due to small amounts of uncomplexed Zn²⁺ in the ZnPP preparations added to the bacterial overexpression cultures.⁸ Crystals of E. coli heme-containing Bfr soaked in a zinc salt were shown to incorporate two zincs into the FC.⁹ However, two-zinc occupancy of the FC has also been verified by anomalous scattering in at least one other crystal structure of E. coli heme-containing Bfr,⁴³ and that report does not mention addition of exogenous zinc to the bacterial overexpression cultures, the as-isolated Bfr or the Bfr crystals.

Table 2.

Iron and Zinc Contents of E. coli ZnPP-Bfrs Used in This Work

sample	Fea	Zna
Bfr-ZnPP-ai	11(3)	50(10)
Bfr-ZnPP-Fe	2600(100)	50(10)
Fedepl-ZnPP-Bfr	3(1)	19(3)

^aMoles of metal per mol 24-mer; average of three replicate determinations with standard deviations in parentheses.

Photophysical and Photochemical Properties.

We investigated the effects of loaded iron on the photophysical properties of ZnPP-Bfr-Fe compared to those of the as-isolated protein ZnPP-Bfr-ai. To avoid interference by O₂, all photochemical experiments were conducted with samples under a N₂ atmosphere.

The UV–vis absorption spectrum of ZnPP-Bfr-ai has been reported previously and is characterized by a Soret band absorption maximum at 433 nm.²⁰ Irradiation of ZnPP-Bfr-ai with filtered light overlapping the Soret band (426 ± 25 nm) did not result in significant UV–vis spectral changes after 60 min (Figure 3). Irradiation of ZnPP-Bfr-Fe also did not result in significant changes in the UV–vis absorption spectrum, except for a slight decrease in the sloping background of [FeO(OH)]_n absorption near 330 nm (Figure S4).

Figure 3.

UV–vis absorption spectral time courses for a solution of ZnPP-Bfr-ai (~5 μM in ZnPP) in 150 mM NaCl (pH 7) under a N₂ atmosphere upon irradiation with 426 ± 25 nm light for the indicated times in minutes. Zero time indicates the spectrum obtained just prior to irradiation. R indicates the spectrum after return to “dark” and incubated for 15 min.

To more clearly observe possible UV–vis spectral changes of the ZnPP upon irradiation, we added a smaller amount of iron, 200 Fe²⁺ atoms per ZnPP-Bfr-ai 24-mer in an air-saturated solution. As one can see in Figure 4, irradiation with the filtered Soret-targeted light resulted in a decrease in Soret- and Q-band absorptions and the appearance of a new absorption feature at 685 nm within 5 min that maximized at ~15 min and persisted for a 60 min irradiation. These spectral changes are characteristic of oxidation of ZnPP to its cation radical, ZnPP^•+.¹⁵

Figure 4.

UV–vis absorption spectral time courses for solutions of ZnPP-Bfr-ai (~3 μM ZnPP in panel A or ~1.5 μM ZnPP in panel B) treated aerobically with 200 equiv of Fe²⁺ as ferrous ammonium sulfate per Bfr 24-mer in 150 mM NaCl (pH 7) and then placed under a N₂ atmosphere and irradiated with 426 ± 25 nm light for the indicated times in minutes. Zero time is just prior to irradiation. (A) Spectra recorded at the indicated times in min during a 60 min total irradiation time. (B) Spectra recorded just prior to irradiation (0 min), after irradiation for 5 min, or after irradiation for 5 min and a return to the dark for 15 min.

The steady state luminescence spectra of ZnPP-Bfr-ai and ZnPP-Bfr-Fe (Figure S5) show fluorescence emission maxima at 590 and 650 nm and phosphorescence emission at 745 and 830 nm. These spectra closely resemble those reported for the ZnPP-Bfr dimer¹⁵ and for other ZnPP-substituted heme binding proteins.^4,45-50 The transient fluorescence decay time courses at 650 nm upon flash excitation at 434 nm are shown in Figure 5. A single decay component was observed for ZnPP-Bfr-ai with a fitted τ₃ of 1.7 ns (Table 3). This fluorescence decay is due to a combination of radiative relaxation of ¹ZnPP* to ZnPP and intersystem crossing (ISC) to ³ZnPP*. The decay time constant is similar to those reported for ¹ZnPP* in other ZnPP-substituted proteins with differing axial Zn ligand environments.^4,45,47,49 The fluorescence decay time courses for ZnPP-Bfr-Fe and ZnPP-Bfr-Fe with 10 mM NADH were more complex and best fit with three time constants. Time constant τ₃ was identical to that for the single decay component observed for ZnPP-Bfr-ai (Table 3). The three-component fit resolved two more rapid decay phases with time constants τ₁ and τ₂ of 70–90 ps and 0.6–0.7 ns, respectively (Table 3). Because the corresponding relaxation times in the presence and absence of NADH are similar to each other (Table 3), we infer that the τ₁ and τ₂ components are due to Fe²⁺ or Fe³⁺ proximal to the ZnPP in ZnPP-Bfr-Fe. The τ₂ component possibly reflects accelerated ISC due to the proximity of paramagnetic Fe²⁺ or Fe³⁺. The 70–90 ps fluorescence decay component, τ₁, is shorter than those reported for paramagnetic metal ion quenching of ZnPP.⁵¹ Fits using other kinetic models, including a stretched exponential (with two adjustable parameters),⁵² were unsatisfactory. As noted below, the fluorescence decay components listed in Table 3 using the multiexponential expression correlated with those from independent fits to transient absorption decays.

Figure 5.

Fluorescence decay time courses at 650 nm (λ_ex = 434 nm) for ZnPP-Bfr-ai, ZnPP-Bfr-Fe, and ZnPP-Bfr-Fe with 10 mM NADH. Solutions were under a N₂ atmosphere in 150 mM NaCl (pH 7) with protein concentrations adjusted to ~3 μM ZnPP.

Table 3.

Fluorescence Decays for ZnPP-Bfr^a

sample	τ1 (ns)	τ2 (ns)	τ3 (ns)
ZnPP-Bfr-ai	–	–	1.7
ZnPP-Bfr-Fe	0.09 (22%)	0.73 (26%)	1.7 (52%)
ZnPP-Bfr-Fe with NADH	0.07 (44%)	0.64 (27%)	1.6 (29%)

^aAt 650 nm with a λ_ex of 434 nm. Solutions were under a N₂ atmosphere and contained 150 mM NaCl (pH 7). NADH when present was at a concentration of 10 mM.

UV–vis transient difference absorption spectroscopy upon nanosecond excitation at 434 nm (ns TA) was used to examine the dynamics of ³ZnPP* in ZnPP-Bfr. Ns TA spectra recorded after a 4 ns delay are shown in Figure 6. The positive absorption feature at ~475 nm is characteristic of ³ZnPP* in other proteins.¹⁴ The corresponding ns TA time courses are shown in Figure S6 for a 50 ns time delay. From these time courses, a ³ZnPP* relaxation time τ_T ~ 8 ms) was fitted at 475 nm. This relaxation time is similar to those assigned to ³ZnPP* in other ZnPP-substituted proteins (~10 ms)⁴ and is not significantly affected by loaded iron in the presence or absence of NADH. However, at early pulse delay times (<10 ns), an absorption feature reaching a maximum at ~685 nm is evident in difference spectra of ZnPP-Bfr-Fe and ZnPP-Bfr-Fe with NADH, but not in that of ZnPP-Bfr-ai (Figure 6, inset). We associate this feature with that assigned to ZnPP^•+ in the steady state absorption spectra (Figure 4).¹⁵ The 685 nm absorption thus appears in the presence but not in the absence of an [FeO(OH)]_n core. This 685 nm feature decays much more rapidly [τ > 200 μs (Figure S7)] than does the 475 nm feature assigned to ³ZnPP*.

Figure 6.

Transient UV–vis absorption difference spectra recorded with a 4 ns delay after nanosecond pulsed laser excitation at 434 nm of ZnPP-Bfr, ZnPP-Bfr-Fe, or ZnPP-Bfr-Fe with 10 mM NADH. The inset shows the 550–800 nm region on an expanded absorbance scale. The solution was under a N₂ atmosphere in 150 mM NaCl (pH 7) with the protein concentration adjusted to ~3 μM ZnPP.

Transient difference absorption spectral time courses upon 434 nm femtosecond pulsed laser excitation (fs TA) of the ZnPP-Bfrs are shown in Figure 7, and Table 4 lists the corresponding relaxation time constants. The absorption feature at ~510 nm is characteristic of ¹ZnPP*,¹⁵ and its ISC gives rise to the characteristic ³ZnPP* feature at ~475 nm. The 510 nm fs TA for ZnPP-Bfr-ai was fit well to a single exponential with a τ of 1.6 ns (τ₃ in Table 4). This value is essentially identical to that of the single fluorescence decay component observed for ZnPP-Bfr-ai (τ₃ in Table 3). This parallel absorbance/fluorescence behavior thus largely reflects ¹ZnPP* to ³ZnPP* ISC. The corresponding fs TA for ZnPP-Bfr-Fe differed in two significant ways from that of ZnPP-Bfr-ai. First, the time course at 510 nm for ZnPP-Bfr-Fe was best fit to three relaxation components, the longest of which (τ₃ = 1.2 ns) is similar to that associated with ISC in ZnPP-Bfr-ai. The other two components have much shorter time constants: τ₁ = 4 ps and τ₂ = 90 ps (Table 4). The longer and two shorter components show only minor changes in the presence of NADH. We can, therefore, presume that the shorter relaxations are due to the effects of loaded Fe²⁺ or Fe³⁺. A second differing behavior in the fs TA of ZnPP-Bfr-Fe in the presence and absence of NADH is the rapid appearance and persistence of the 685 nm absorption assigned to ZnPP^•+, which is not apparent in the corresponding fs TA for ZnPP-Bfr-ai. On the basis of these observations, a straightforward explanation for the shorter fs TA relaxation transients, τ₁ and τ₂, is electron transfer (ET) from ¹ZnPP* to loaded Fe³⁺. The 90 ps relaxation component in the fs TA (τ₂) has a counterpart in the fluorescence decay component, τ₁ (Table 3). Our fluorescence transient instrument would not be capable of resolving a counterpart to the 3.4–3.8 ps fs TA component (τ₁ in Table 4).

Figure 7.

Transient absorption difference spectral time courses upon femtosecond 434 nm laser excitation of ZnPP-Bfr, ZnPP-Bfr-Fe, and ZnPP-Bfr-Fe with 10 mM NADH. The left column panels show full spectral time courses, and the right column panels show corresponding single-wavelength time courses with fits superimposed on the 510 nm time courses. Solutions were under a N₂ atmosphere in 150 mM NaCl (pH 7) with the protein concentrations adjusted to ~3 μM ZnPP. The negative absorption feature at ~650 nm in the full spectrum panels is due to stimulated emission.

Table 4.

fs TA upon 434 nm Excitation of ZnPP-Bfr.^a

sample	τ1	τ2	τ3
ZnPP-Bfr-ai	–	–	1600 ± 150 (57%)
ZnPP-Bfr-Fe	3.8 ± 0.3 (43%)	91 ± 16 (19%)	1200 ± 130 (28%)
ZnPP-Bfr-Fe/NADH	3.4 ± 0.4 (40%)	87 ± 16 (18%)	1300 ± 140 (29%)

^aτ values in picoseconds were obtained from fits to absorbance changes at 505–510 nm. Laser power of 240 μW.

To probe possible ET from ¹ZnPP* to iron in the FC, we also obtained fs TA time courses for Fe_depl-ZnPP-Bfr and Fe_depl-ZnPP-Bfr to which 48 irons per 24-mer had been added from a ferrous ammonium sulfate stock solution to the air-saturated protein solution (Figure 8). The 24 FCs can potentially bind up to 48 iron (or zinc) ions. On the basis of the iron and zinc stoichiometries of the Fe_depl-ZnPP-Bfr (Table 2) and previous reports,^9,10,53 we expected at least some of the 48 added irons would occupy the FC. However, fits to the 510 nm ΔA time course of the 48 iron-treated Fe_depl-ZnPP-Bfr (Figure 8, bottom panel) showed only the nanosecond decay component closely resembling those of the untreated Fe_depl-ZnPP-Bfr (Figure 8, top panel) and ZnPP-Bfr-ai (Figure 7, top panel, and Table 4).

Figure 8.

Transient absorption difference spectral time courses upon femtosecond 434 nm laser excitation of Fe_depl-ZnPP-Bfr and Fe_depl-ZnPP-Bfr to which 48 Fe atoms per Bfr 24-mer added as ferrous ammonium sulfate to an air-saturated protein solution. Solutions for fs TA were under a N₂ atmosphere in 150 mM NaCl (pH 7) with protein concentrations adjusted to 3–5 μM ZnPP. The negative absorption feature at ~650 nm is due to stimulated emission.

DISCUSSION

The mechanisms and pathways for iron mobilization from Bfr are less well characterized than those for oxidative iron uptake and [FeO(OH)]_n core formation.^44,54 In vitro reductive iron mobilization from iron-loaded E. coli Bfr can be achieved using various reducing agents that either passively diffuse through pores in the protein shell or transfer electrons across the protein shell. Core iron is reduced to soluble Fe²⁺, which exits through pores in the protein shell and can be chelated by various exogenous ligands.¹² An obvious role for the heme in Bfr is mediation of ET from exogenous reductants to core Fe³⁺, although in our view this role has not been conclusively established for E. coli Bfr.

ZnPP substituted into heme binding sites in other proteins is either known or thought to have five-coordinated Zn(II) typically with an axial His ligand.^4,22,55 Axial coordination of Tyr to Zn in other ZnPP-substituted proteins may also occur.^4,49 Evidence for six coordination of the Zn porphyrin in ZnPP substituted into cytochrome c peroxidase²⁸ and the covalently attached Zn porphyrin in cytochrome c has been obtained.^3,30 The bis(thioether) axial coordination of zinc in ZnPP-Bfr (Figure 2) is to the best of our knowledge unprecedented for zinc porphyrins in either synthetic compounds or proteins. Our results show that the photophysical properties of ZnPP in ZnPP-Bfr are generally similar to those of ZnPP substituted into myoglobin and other proteins.^4,45,47,49 Co et al.⁷ have demonstrated and quantified the energetics of electron transfer processes for ³ZnP* and ¹ZnP* in ZnP-substituted myoglobin. Although the energetics typically favor oxidative over reductive quenching for both singlet and triplet excited states, both quenching pathways have been observed in ZnPP-substituted proteins.^3,4,7,56-58

The redox orbital of ¹ZnPP* or ³ZnPP* is more delocalized over the porphyrin ring than that of heme iron. This delocalization and the highly reducing nature of ¹ZnPP* and ³ZnPP* (ZnPP^•+/³ZnPP* E°′ ≤ −0.8 V vs NHE)^7,59,60 would likely make these excited states more efficient than ferrous heme in Bfr (estimated ferric/ferrous heme E°′ ~ −0.48 V) at reducing non-heme Fe³⁺ within the protein shell or the [FeO(OH)]_n core (estimated Fe³⁺/Fe²⁺ E°′ ~ −0.42 V).^{61 3}ZnPP* does not appear to participate in ET to core iron. Our evidence rather supports one-electron oxidation of ¹ZnPP* to ZnPP^•+ by Fe³⁺ in iron-loaded ZnPP-Bfr. Two rapid (picosecond time scale) ¹ZnPP* relaxations were observed for ZnPP-Bfr-Fe (τ₁ and τ₂ in Table 4) concomitant with ZnPP^•+ formation in the fs TA of ZnPP-Bfr-Fe. These picosecond phases were not observed in ZnPP-Bfr-ai but were observed when as few as 200 Fe²⁺ ions were added to ZnPP-Bfr-ai and allowed to oxidize in an aerobic atmosphere.

Given the proximity of two FCs to each ZnPP (Figure S3), one of the rapid ¹ZnPP* decay phases in the fs TA could be due to ET from ¹ZnPP* FC to Fe³⁺. However, exogenously added Zn²⁺ is known to have a higher affinity for the Bfr FC than does added Fe²⁺,^10,32,53,62 and the high zinc content of ZnPP-Bfr-ai and ZnPP-Bfr-Fe could prevent binding of added iron to the FC. Previous reports indicate that the first 48 Fe²⁺ ions per 24-mer added to Bfr bind to the FC and are autoxidized to Fe³⁺ on a time scale of seconds.^10,53 We, therefore, examined Fe_depl-ZnPP-Bfr, which, on the basis of its 3 iron and 19 zinc/24-mer stoichiometry (Table 2) and assuming 10–12 of these zincs belong to ZnPP, should leave a significant portion of its FCs unoccupied by either zinc or iron. However, while a sample prepared by aerobic addition of 48 Fe²⁺ ions to Fe_depl-ZnPP-Bfr gave rise to the ¹ZnPP* signal in fs TA, it was not accompanied by the spectral signature of ZnPP^•+, and the time course did not show the rapid phases attributed to oxidation of ¹ZnPP*. Our results, thus, do not support ET from ¹ZnPP* to FC Fe³⁺ under these conditions.

We interpret the collective results according to Scheme 3, which implies oxidative quenching of the ¹ZnPP* by ensembles of proximal and distal Fe³⁺ in [FeO(OH)]_n clusters lining the inner surface of the protein shell. Scheme 3 does not include “back electron transfer” from Fe²⁺ to ZnPP^•+, which we propose is responsible for the “dark” return of the ZnPP^•+ absorption to that of ZnPP in steady state irradiation experiments (Figure 4B). The inner surface of the protein shell contains numerous glutamate and carboxylate side chains that provide potential attachment sites for Fe²⁺ and Fe³⁺ ions.^9,32,44 These iron binding sites would in turn provide nucleation sites for [FeO(OH)]_n cluster formation. The available X-ray crystal structures of Bfrs do not show core [FeO(OH)]_n or distinct [FeO(OH)]_n clusters.^44,63 However, TEM of ZnPP-Bfr-Fe shows iron cores of non-uniform density, possibly composed of multiple [FeO(OH)]_n clusters.²⁰ An electron photoinjected into Fe³⁺ on the surface of an [FeO(OH)]_n cluster would be delocalized by hopping to other Fe³⁺,⁶⁴ thereby slowing “back electron transfer” to the strongly oxidizing ZnPP^•+. The Fe²⁺ could, thus, migrate through [FeO(OH)]_n, dissociate from the cluster, and exit through pores in the protein shell.

Scheme 3.

Proposed Photoinitiated ET from ¹ZnPP* to [FeO(OH)]_n Clusters at the Inner Surface of the Protein Shell in ZnPP-Bfr-Fe

NADH in large excess did not affect the kinetics of ¹ZnPP* or ZnPP^•+ decay in ZnPP-Bfr-Fe, consistent with much slower ET from NADH to ¹ZnPP* or ZnPP^•+. Nevertheless, photosensitized iron release did not occur in the absence of NADH.²⁰ We also noted that visible light irradiation of the iron-loaded heme-containing Bfr with excess NADH did not result in significant iron release.²⁰ We, therefore, presume that the oxidative quenching process is followed by slow re-reduction of ZnPP^•+ to ZnPP by NADH, possibly involving the NAD radical (NAD^• in Scheme 2).^65,66 This notion is consistent with the observed quantitative release of Fe²⁺ from ZnPP-Bfr-Fe over the course of 3 h upon irradiation centered on the ZnPP Soret band in the presence of excess NADH.²⁰ Reductive iron mobilization is presumably an essential function of Bfrs.^12,67 Our photosensitized reductive approach on ZnPP-Bfr could potentially identify the sites of reduction and release of iron.

CONCLUSIONS

The X-ray crystal structure of ZnPP-Bfr shows a unique six-coordinate zinc in the ZnPP with two axial methionine sulfur ligands with some disorder of the zinc out of the porphyrin plane. Both steady state and transient UV–vis absorption spectroscopies show that irradiation with light overlapping the Soret absorption causes oxidation of ZnPP to the cation radical, ZnPP^•+, only when the ZnPP-Bfr is loaded with iron. Fs TA shows that this photooxidation occurs from the singlet excited state (¹ZnPP*). The results are consistent with two distinct populations of Fe³⁺ in [FeO(OH)]_n clusters that oxidize the ¹ZnPP* on the picosecond time scale. Much slower “back electron transfer” from Fe²⁺ to ZnPP^•+ was also observed.