1 Introduction

Oxygen is arguably one of life’s most vital molecules, participating in a multitude of indispensable biological processes. In the realm of oxygen driven enzymatic reactions, those involving short-lived intermediates proceeding through an early oxo- or peroxo-intermediate, such as in cytochrome P450 monooxygenase [1, 2], methane monooxygenase [3] and the heme-copper oxidase family [4, 5] have for long eluded detailed mechanistic study and structural elucidation. The fast time-scales at which these intermediates operate render traditional flow flash and mixing techniques inadequate for accurate detection [6,7,8]. Furthermore, traditional methods of studying such short-lived intermediates have relied upon methods for blocking the oxygen binding site with non-native ligands such as carbon monoxide. Applications of photolabile caged-oxygen carriers for time resolved studies of oxygen dependant enzymes have circumvented some conventional pitfalls associated with traditional flow flash techniques [6, 9]. Binuclear dioxygen bridged cobalt(III) complexes that have for long been known to release oxygen thermally, chemically and photochemically [10,11,12,13]. However, early bis(ethylenediamine)-based complexes were plagued by very poor photochemical quantum yields and exhibited low aqueous solubility, significantly hampering their utility in biochemical investigations [11]. More recently, a (µ-peroxo)(µ-hydroxo)bis[bis(bipyridyl)cobalt (HPBC) compound have been demonstrated to be a particularly versatile tool and have been successfully applied to time resolved spectroscopic studies of the quinol oxidase [14], bovine aa3-type [6, 15] and the ba3-type Cytochrome c oxidase [7, 15, 16]. The photoliberation reaction occurs at a rapid nanosecond time scale and produces molecular oxygen with high spatiotemporal control [17]. This type of controlled release addresses a significant challenge in studies which require a rapid reaction initiation prior to detection, such as pump-probe spectroscopic studies based on optical absorbance and similar experiments which use X-rays. The study of proteins and substrates in the field of structural biology has long sought out this possibility but a high degree of temporal control has been lacking for substrate dependent proteins and even innovative and sophisticated mixing approaches have struggled in achieving millisecond time resolution [18,19,20,21]. Photo-cages present a promising avenue for establishing temporal control in time resolved crystallographic experiments [22, 23]. Notably, three examples of X-ray crystallography-based studies involving a photolabile caged compound have been reported. These studies showcase the successful decaging of nitric oxide, ATP, and fluoroacetate in conjunction with their respective enzymes, leading to the acquisition of valuable structural information [23,24,25,26]. Here we set out to investigate if a caged oxygen compound could enable time resolved crystallographic studies of the ba3-type CcO, a member of the heme-copper oxidase family. The structure of the oxidized state of ba3-type CcO has been previously reported using serial femtosecond crystallography (SFX) [27, 28] and more recently, the time-resolved dissociation of carbon monoxide from the reduced carbon monoxide-bound state has been investigated [29]. However, to gain further structural insight, it is imperative to study the native substrate with the enzyme and current efforts are directed towards implementing a caged oxygen compound for time-resolved (TR) crystallographic studies of CcO. The success of this endeavour could provide a novel avenue for elucidating structural identity of the catalytic intermediates and could pave the way for similar studies of different oxygen-dependent systems. To enable a substrate-based TR-SFX study it is imperative that a sufficiently high concentration of the substrate is liberated on a shorter time scale than the reaction of interest, upon external stimuli such as by a single laser flash. This ensures that the reaction initiation is synchronous. In this work we provide appropriate and general conditions for the implementation of a photolabile caged oxygen carrier to trigger the native reaction cycle under conditions for crystallographic studies of cytochrome c oxidase, and with the required spatio-temporal control to facilitate further time resolved crystallographic studies.

2 Results and discussion

For a photocage to be suitable for biological applications, compatibility with the experimental conditions such as stability over time, temperature and other chemical components in the system is imperative. Currently, there are two well established methods for preparing the reduced state R of CcO: treatment with excess sodium dithionite (NaDt) [27] or a combination of sodium ascorbate and a mediator such as hexaammineruthenium(II) [6] or phenazine methosulfate (PMS) [30]. In contrast to prior studies [6] involving caged oxygen compound 1 (Fig. 1, b), where no significant degradation was observed, we observed a rapid degradation of compound 1 when exposed to both methods of reduction (Fig S4). Subsequent analysis determined that sodium dithionite and ascorbate lead to the degradation of 1 with an accelerated effect when a mediator was used in conjunction with the reductant. The rate of decomposition depends on the relative concentrations of compound 1 and the reductant, but nevertheless would restrict any experimental applications to a rapid mixing approach to avoid this detrimental side reaction. In pursuit of a mixing-free TR-SFX experiment, we initially embarked on an investigation to determine if compatibility with the reducing agent could be enhanced through structural modification of the caged oxygen compound.

Fig. 1
figure 1

a Overall reaction for the synthesis of compound 1–7. b Structure and isolated yields of the produced HPBC complexes 1–7. c The photoreaction of compound 1 illustrated with the ORTEP drawing of the crystal structures of compound 1 (left) and mononuclear cobalt(II) photoproduct (right). The nitrate counterions and hydrogens have been omitted for clarity. Compound 1: triclinic, space group P-1. Photoproduct 1: Monoclinic, space group C2/c. For CCDC depository numbers, see the Supporting Information.

Full size image

2.1 Synthesis of substituted caged oxygen compounds

Compounds 1–7 were synthesized to investigate steric and electronic effects on the oxygen liberation, according to Fig. 1a. Ligands (2 equivalents.) were dissolved and added to a solution of a cobalt(II) salt, which was then oxygenated with a slow addition of a hydroxide-base (0.5 equivalents.). The compounds were then isolated and purified by slow vapour diffusion. The previously described compound 1 was prepared with minor alterations to the original procedure [14] which in our hands generated a product with varying degrees of purity. Our adjusted protocol resulted in a lower yield but more consistently produced crystalline pure product.

Overall, the steric and electronic properties of the ligand were found to have a profound impact on both stability and the yield of isolation, which were generally low (Fig. 1b). Methylation at the para and meta-positions of the bipyridine ring (Fig. 1b, compound 2, 4) proved to be tolerated but not in the ortho-position (Fig S1, A). Sterically demanding ligands such as 1,1′-biisoquinoline (Fig S1, I) likely causes an unfavourable coordination angle to the metal in the final complex, while 2,2′-biquinoline (Fig S1, H) possibly suffers from unfavourable steric interactions due to substitution in the ortho-position. A para-methoxy substituted complex (Fig. 1b, compound 3) is tolerated but the corresponding hydroxyl-compound (Fig S1, O) was not, possibly due to increased solubility which likely prevents crystallization and isolation. Attempts to isolate by precipitating as either PF6 or BF4 salts were also unsuccessful. Phenyl and tert-butyl (Fig. 1b, compound 6, 7) substituents in the para-position are tolerated but this renders the complex insoluble in water, even as nitrate salts which typically have high aqueous solubility. The corresponding phenanthroline derivative could also be synthesized but suffers from poor aqueous stability and decomposes completely over several hours at room temperature (Fig. 1b, compound 5). All attempts to synthesize and isolate electron deficient complexes were unsuccessful (Fig S1, B–F, J, L–N), possibly explained by the fact that the oxygenation reactions of µ-peroxo-µ-hydroxo cobalt(III) complexes are assisted by ligand to metal sigma donation that stabilizes the dioxygen adduct [31, 32]. An unexpected cubane-type side product was identified by X-ray crystallography, from the reaction with 4,4′-bis(trifluoromethyl)-2,2′-bipyridine (supplementary crystal structure 2). This indicated that electron deficient ligands might allow for non-productive alternative reaction pathways, in addition to a likely faster decomposition in solution. Compound 1 could reliably be prepared in higher yields as a tetra fluoroborate salt (47%) with slightly lower water solubility compared to the nitrate salt, or as the hexafluorophosphate salt (> 90%) which exhibits poor aqueous solubility. Similar attempts to precipitate substituted compounds 2–7 resulted in the formation of brown precipitates with low yields and varying degrees of purity. Prolonged storage of compounds 1–7 at room temperature in solution typically results in thermal degradation and release of oxygen and the formation of red crystals, identified as photoproduct mononuclear cobalt(II) specie (Fig. 1c) and precipitation of a yellow side products were identified to be the corresponding tris(bipyridine)cobalt(II) mononuclear complex (Fig. 2, f) which is a side product from the reaction.

Fig. 2
figure 2

ORTEP drawings of obtained crystal structures. Hydrogens, counterions and solvents have been omitted for clarity. For CCDC depository numbers, see the Supporting Information. a Compound 2: monoclinic, space group P21/c. b Compound 4: monoclinic, space group P21/n. c Compound 6, triclinic, space group P-1. d Compound 7, monoclinic, space group P21/n. e Photoproduct of compound 3 obtained by crystallization with NH4BF4 additive after exposure to light. Monoclinic, space group I2/a. f Yellow side product obtained during crystallization of compound 4, triclinic, space group P-1

Full size image

2.2 Crystallography

Crystals, suitable for X-ray crystallography were obtained for the different compounds by slow vapour diffusion. Crystallization was carried out at temperatures between − 20 and  − 20 °C in the dark to obtain crystals for the HPBC complexes and some crystallization was performed after exposure to light to also obtain the photoproducts, which are easily identified as red plates in contrast to the black crystals for the HPBC complexes when mixtures was obtained. All HPBC compounds exhibits the characteristic (µ-peroxo)-(µ-hydroxo)-bridge. Furthermore, we observe µ(O)-µ(O) bond distances consistent to other closely related µ-peroxo complexes [13, 33] and slightly shorter than the earlier ethylenediamine based complexes [34]. Relevant bond lengths and angles are summarized in supplementary Table 1. We were unable to produce crystals of sufficiently good quality for compounds 3 and 5. However, the photoproduct for compound 3 was obtained (Fig. 2e).

Table 1 Determined half-lives of compound 1–5 (5 mM) mixed with sodium ascorbate (5 mM) + PMS (1.5 µM) and sodium dithionite (5 mM)
Full size table

2.3 Spectroscopic analysis

The optical absorbance spectrums for the water-soluble compounds 1–5 are shown in Fig. 3a and the non-water-soluble compound 6–7 in Fig. 3b. All compounds display a broad, moderately strong characteristic peak in the range of 390–450 nm arising from a π*\({\text{(O}}_{2}^{ - 2} ) \to\) dσ* ligand to metal charge transfer (LMCT) band (Fig. 3a, b) [33]. This is consistent with similar µ-peroxo-µ-hydroxo-cobalt compounds previously described in the literature [35]. The infrared stretch frequencies around 830–850 cm−1 of the O–O bond also support the character of the O2 as a peroxo-type ligand as opposed to the superoxide [36]. 5Me-substituted compound 4 exhibits a slight red shift of the 305 nm absorbance band compared to the unsubstituted compound 1. Phenanthroline based compound 5 does not share the 300 nm absorbance band and instead shows a strong absorbance peak around 270 nm. The para-methoxy substituted compound 3 exhibits a minor blue shift of the 300 nm peak, accompanied by a slightly lower extinction coefficient in this region and a higher coefficient in the 400 nm region. All the HPBC based compounds produced in this work are photolabile and upon light irradiation will dissociate into the corresponding cobalt(II) mononuclear complex and release molecular oxygen, as illustrated in Fig. 1c. To verify the previously demonstrated correlation between the disappearance of the 400 nm charge transfer band in the absorbance spectrum and liberation of O2, the water-soluble compounds 1–5 were illuminated with blue LED light (385 nm) and the oxygen concentration was measured over time. The dissolved oxygen concentration was measured using a fluorescence oxygen sensor (Ocean optics, NeoFox using the patch-based FOXY formulation) preparation of the sample. In Fig. 3c the absorbance spectrum for compound 4 is shown before (solid line) and after (dashed line) light illumination. The photolytic product for non-water-soluble compound 6 obtained by photolysis in MeCN was identified with X-ray crystallography to form the same type of mononuclear cobalt(II) complex as photolysis in water (see Supplementary crystal structure 1). This together with the similar absorbance spectrum changes (compare Fig. 3c, d) and visible bubble formation after exposure to light at higher concentrations shows that the same O2 liberating reaction occurs in non-aqueous solvents.

Fig. 3
figure 3

a Absorbance spectrum of compound 1–5 (0.025 mM) in water. b Absorbance spectrum of compound 6 (0.025 mM) and 7 (0.0125 mM) in MeCN. c Absorbance spectrum of compound 4 before (whole line) and after (dashed line) photolysis in water. d Absorbance spectrum of compound 6 before (whole line) and after (dashed line) photolysis in MeCN

Full size image

2.4 Reductant mixing and redox stability

Next, we probed the redox stability of the compounds 1–5 by exposing them to conditions typical for reduction of CcO. Employing a custom-built flow cell device [28] consisting of a 0.3 mm quartz capillary connected to a syringe pump, the time until 50% of the compound had degraded for compounds 1–5 was determined. Mixing was performed with 5 mM of compounds 1–5 and 5 mM of the reductant as this would be realistic reductant/oxygen cage concentrations for further studies. The stability of compound 1 could now be accurately determined and shows almost instantaneous decomposition when exposed to equimolar amounts NaDt (Table 1) and approximately 30 s half-life together with NaAsc + PMS. Although not completely detrimental for applications that can accommodate a very rapid mixing, this rapid degradation poses a significant challenge for the general use of caged oxygen compounds in crystallography-based studies. We rationalized that increasing the electron density of the cobalt centres by introducing an electron donating ligand would make the complex less susceptible to reduction and indeed compound 2–3 shows a significantly increased stability towards both reduction protocols, in addition to a significantly increased stability in solution at room temperature. Aqueous solutions of compounds 1,4 and 5 are not sufficiently stable at room temperature and will thermally degrade over several hours as opposed to compound 2 and 3, which show only minor degradation over several hours. The peak anodic current was measured with cyclic voltammetry and the voltammograms for compound 1–7 were obtained in MeCN. The obtained voltammograms are complex with multiple irreversible processes contingent on the scan range and direction but we found that the first reduction peak, which disappeared upon subsequent cycles, correlated well with our stability study, and was observed for compound 1–7.

The potential was also within a reasonable range (− 0.4 to − 0.2 V) compared to the reduction potential of relevant reducing agents. The peak anodic current was determined and a lower negative peak anodic current for the first reduction peak in the voltagram was found to be consistent with a compound more resilient towards chemical reduction (Table 1).Compound 5 shows an increased susceptibility towards chemical reduction which can be attributed to electronic factors, given that phenanthroline is a worse electron donor than bipyridine [37]. Compound 4 also falls under this category as a worse electron donor than compound 3 and 4 but is expected to be a slightly better donor than compound 1 which correlates well with the observed stability profile. Intrigued by the possibility of investigating further electron rich ligands, we first decided to evaluate the correlation between ligand electronics and the photolytic properties prior to such a commitment into further synthetic efforts as we suspected that the electronics would also influence the photo physical properties.

2.5 Photophysical properties and quantum yield

Preliminary photolysis experiments with the nitrate and the perchlorate salt of compound 1 raised concerns about the accuracy of some reported quantum yields. A prior study claimed a wavelength-dependent quantum yield, correlating to the extinction coefficient, with reported quantum yields ranging between 0.2 and 0.5 for wavelength between 300 and 320 nm, and a significantly higher overall quantum yield than our findings [14]. This is in contrast to the previously determined quantum yield of 0.04 at 355 nm for the perchlorate salt [6, 17]. Upon re-evaluation of the quantum yield for compound 1, it was discovered that the nitrate, perchlorate and tetrafluoroborate salts exhibited identical photolytic properties when illuminated across various wavelengths. The photon flux for several wavelength were determined by chemical actinometry using the well-established protocol for potassium ferrioxalate [38]. Our results align with first published quantum yield for the perchlorate salt [17] with quantum yields ranging 0.03–0.06 depending upon excitation wavelength (Table 2).

Table 2 Quantum yield of compound 1–7 in water and MeCN
Full size table

Surprisingly, we observed a decrease in quantum yield for compounds 2–7 compared to unsubstituted 1 across all investigated wavelengths. All compounds exhibit a higher quantum yield in MeCN compared to water, however, the stability of compounds 1 and 5 were significantly compromised in MeCN. These findings suggested to us that the relatively poor quantum efficiency of the photo-liberation reaction is an inherent property of HPBC based compounds and further structural modification is unlikely to improve the quantum yield significantly. We therefore turned to a different approach to find compatible conditions by identifying an alternative method of producing the reduced state for cytochrome c oxidase. By identifying less reductive conditions we reasoned that the slightly more redox resilient compounds 3 and 5 could overcome the previous chemical incompatibilities or simply the generic use of any HPBC based compound.

2.6 Photoliberation of oxygen

After an exhaustive search for less reducing conditions, we identified that a substantial excess of dithiothreitol (DTT) could produce CcO in the fully reduced state R. However, this process exhibited a significantly slower reduction kinetic compared to traditional methods, taking several minutes rather than sub-second duration that is typically observed. Notably, the addition of a fourfold molar excess of DTT to compound 1 did not lead to any discernible acceleration of decomposition, which was confirmed by NMR (Fig S5) and optical absorbance spectroscopy (Fig S6). Encouraged by the discovery of less reducing conditions for preparing the active state of the enzyme, we could turn our attention to establishing a protocol that would enable time resolved crystallographic study of CcO. In previous time resolved spectroscopic studies where molecular oxygen or caged oxygen has been utilized, a 30–45 fold excess of released oxygen was employed [7, 15]. In crystallographic studies, the concentration of protein prior to crystallization is typically in the domain of 50–80 μM. For example, once crystals of CcO form [27], the concentration of protein within microcrystals is 2.3 mM. The inhomogeneous distribution of protein within the total sample means that to achieve a homogenous reaction initiation, two primary criteria must be satisfied; (1) 2–3 mM photoproduced oxygen should be liberated upon a single laser excitation and, (2) the validation of retention of time resolution at these concentrations, meaning that one can distinguish between the different reaction intermediates during oxygen-initiated turnover of the enzyme. Efficient oxygen liberation is crucial. To assess whether a HPBC compound could deliver this high concentration upon a single laser pulse, different concentrations of compound 1, ranging from 1 to 30 mM, underwent exposure to a single 355 nm laser pulse (5 ns, approx. 2 mJ/pulse). The concentration of liberated oxygen was determined by the loss of absorbance at 400 nm, promptly recorded after each pulse. The experimental setup was designed to closely resemble the sample delivery and reaction initiation setup during a realistic X-ray based experiment (see SI) [28]. A thin quartz capillary (0.3 mm) connected to a syringe pump was used and we investigated if the required > 3 mM oxygen could be generated upon a single 355 nm laser flash. The selection of wavelength is motivated by general availability and that we envisioned that light penetration into the sample would be suitable at this wavelength. The importance of light penetration for homogeneous reaction initiation has been emphasized in previous studies [23, 39]. Our experiment shows that oxygen concentrations exceeding 3 mM could be achieved upon a single 355 nm laser pulse (Fig. 4). This corresponds to a concentration of O2 13 times higher than the normal air saturated water and 3 times higher than pure O2 saturated aqueous solution [40]. Concentrations above this level are considered to be supersaturated and will eventually lead to formation of bubbles. Supersaturation is reached because the diffusional escape of gas is slower than the timescale of photoliberation of O2 [17, 41]. While bubble formation was occasionally observed, it did not negatively impact the measurements by upstream O2 contamination or discontinuous flow. We, therefore, reason that oxygen concentrations locally will exceed the required 3 mM minimum during the timescale of the experiment since laser initiation would be followed quickly after by an X-ray pulse. While one might expect that a higher concentration of the compound would correlate with a greater oxygen release, our results revealed an optimal concentration threshold. Specifically, we observed an inner filter effect, characterized by a high optical density, at concentrations exceeding 7 mM, beyond which the photolytic oxygen yields plateaued and ceased to increase with higher concentrations of compound 1. Conversely, at concentrations below 3 mM, the compound achieved full photolytic yield, indicating that even higher yields might be possible with a thinner capillary or by optimizing the laser illumination to encompass multiple angles.

Fig. 4
figure 4

Photoreleased O2 upon a single 355 nm laser pulse (~ 1 J/cm2, 5 ns pulse) showing the percentage yield (dark blue) and the molar yield (blue) of released oxygen. Collected inside a 0.3 mm quartz capillary with approximately 0.01 mm thick walls

Full size image

Moreover, the use of thin jets, with diameters less than 100 µm, a common practice in viscous jet-based SFX experiments, could potentially enhance the oxygen concentration further when combined with multi-angle laser exposure. Encouraged by these results we moved on to simulate the conditions that would be realistic for a crystallography-based experiments.

2.7 Time-resolved spectroscopy

Typically, proteins can also absorb light in the 355 nm region and it is anticipated that this absorbance will decrease the photolytic yield of O2 due to competing absorption of the light. We, therefore, reasoned that there is a high probability that the liberated oxygen to protein ratio during crystallographic conditions will be lower than previous spectroscopic studies. A caged oxygen/protein ratio of 20 was selected as we reasoned that this would be realistically achievable and would showcase a worst-case scenario when the O2 could be a limiting factor. Time resolved optical difference spectra were recorded after a mixture of DTT reduced CcO and compound 1 in solution was exposed to a laser flash and the optical absorbance difference spectrum was recorded 2–2000 µs after the laser pulse (Fig. 5). Occasionally during the preparation of the sample, protein oxidation occurred prematurely. This was attributed to the slow thermal degradation of caged oxygen compound 1 when the sample was left for a prolonged time.

Fig. 5
figure 5

Time resolved optical absorbance difference spectrums of DTT reduced CcO (120 µM) and compound 1 (2.5 mM) 2, 10, 50, 100, 200, 500 µs, 1 and 2 ms after a 355 nm laser pulse. The spectral contribution of compound 1 has been removed and the reduced protein + compound 1 was used as the blank

Full size image

At higher concentrations this would outcompete the DTT mediated reduction, which is slow at lower pH. This occurs because DTT will be protonated to a higher extent, limiting its reducing capability, and effectively lowering the concentration of active reductant in the system. To mitigate this, a catalytic oxygen scavenger system [42] consisting of glucose oxidase, catalase and glucose was introduced to counteract the slow oxygen leakage. This prevented premature oxidation of the protein whilst not interfering with the kinetic binding of oxygen significantly. We obtained difference signals after laser excitation, consistent with the reaction of the reduced enzyme and molecular oxygen, which was achieved with a sufficient population to distinguish between various reaction intermediates (which is shown in Fig. 5) [16]. Notably, the first intermediate in the cycle, intermediate A, which occurs 5–10 µs after O2 binding [7, 9] remains unresolved in our experiment, possibly due to limitations in time resolution. The integration time of the spectrophotometer (30 µs) together with a pulse tail from the light source (≈ 10 µs) leads to less resolved populations for short lived intermediates. Following oxygen ligation, the O–O bond is broken and this results in oxidation of heme a3, which leads to a decrease in the difference spectrum at 445 nm and increase at 410 nm [9]. This is seen together with the oxidation of heme b, a decrease at 560 nm and increase at 610 nm, suggesting that a mixture of the intermediate A and the initial formation of PR can be seen during the 2–10 µs time interval (Fig. 5A, B). The subsequent 50 µs time point exhibits well resolved characteristic spectral features for the formation of ferryl intermediate seen as PR [15]. The highest population of PR is detected at this time point (Fig. 5A, B). Following the formation of PR, is the re-reduction of heme b by CuA forming the F-intermediate, whilst also protonating the binuclear centre [9]. This should result in complete reduction of heme b and the disappearance of the negative peak at 560 nm [9]. However, rapid reduction of the formed F-intermediate likely results in mixed populations of F and the oxidized state of the enzyme during our measurements. The 100 µs time point shows re-reduction of heme b, determined from the 560 nm peak that is re-oxidized upon the 200 µs time point (Fig. 5D). The following time points between 500 µs and 2 ms shows the final transition from the mixed F state to the oxidized state shown by complete disappearance of the transient 610 nm positive peak. (Fig. 5E, F). We estimate that the reaction is initiated in approximately 40–45% of the total protein content by the laser pulse, leading to a reaction initiation homogeneous enough to clearly distinguish the PR reaction intermediate. Increasing the concentration of caged oxygen and performing the laser illumination from more than one angle will likely lead to an even higher protein turnover yield/pulse.

Our protocol can successfully turn around the enzyme upon a single laser pulse, but the question remained if this would be possible to apply to a crystalline sample. With the added complication of a non-aqueous medium as the crystals are based on a liquid cubic phase (LCP) medium [43]. A LCP type phase is also commonly what is used for viscous sample injection during based SFX experiments [44]. We found that exposing an LCP sample containing DTT (15 mM) CcO microcrystals (3–6 µm, 50–60 µM) treated with the oxygen scrub protocol, and compound 1 (2.5 mM) to a single 355 nm pulsed laser produced a spectrum consistent with the transition from reduced state to oxidized CcO (Fig. 6). This was performed with our in-house flow cell using a 0.3 mm quartz capillary. Although a clean spectrum was not obtainable due to issues with light penetration, sample homogeneity and turbidity of the LCP phase, a difference signal consistent with oxidation could be obtained which consistently showed after multiple alternating dark/light cycles under flow conditions. A shift in the soret peak maximum from 425 nm towards 412 nm, consistent with oxidation was observed after 1 ms. A high flowrate was used, and no further significant change associated with protein oxidation was observed even after multiple laser pulses with a stationary sample was illuminated with multiple pulses.

Fig. 6
figure 6

Microcrystals of CcO (50–60 µM) in LCP mixed with caged oxygen compound 1 (2.5 mM) and DTT (15 mM) before (blue) and 1 ms after (yellow) a single laser pulse (5 ns, ~ 1 J/cm2, 355 nm). Peak at 610 nm is not a spectral feature but a result from a noisy background and the spectral contribution from the caged oxygen compound 1 has been removed

Full size image

A separate experiment was performed to investigate the ferryl-intermediate PM in more detail, which as opposed to the PM intermediate in other type oxidases [45], cannot be artificially isolated in a stable form by chemical means in the ba3-type oxidase. This was performed by first producing the carbon monoxide bound mixed-valence state of the enzyme [46] by incubating an aqueous solution of oxygen free of ba3-type CcO with carbon monoxide gas over several hours. The addition of caged oxygen compound 1 lead to familiar issues with premature oxidation due to the thermal oxygen leakage, which in this case was not mitigated by the addition of the oxygen scrub system. Consequently, we shifted our attention to the more redox stable compound 3. Despite its lower oxygen release potential, we rationalized that since the produced intermediate would be trapped due to a lack of further electrons to progress in the catalytic cycle, a lower concentration of O2 would be tolerated during the experimental conditions. Also, by studying a slightly slower millisecond time point after the initial formation, sufficient oxygen should have had time to diffuse into the sample in the case that the initial oxygen burst is not enough to achieve a high population of reacted enzyme. Upon addition of compound 3 to a final concentration of 2 mM, no premature oxidation of CcO (120 µM) in the MVCO state was observed within 1 h. Upon exposure to a single 355 nm laser pulse, difference signals consistent with the formation of the PM state were obtained. This is seen in Fig. 7 as a characteristic positive difference signal at 610 nm. Interestingly, this formed state was stable for about 30 s which was verified by exposing a stationary sample with a single laser pulse and recording spectrums over time. This demonstrates that a sufficiently high population of the PM intermediate can be generated upon a single laser pulse at crystallographic relevant concentrations in an aqueous solution.

Fig. 7
figure 7

Difference spectrum of carbon monoxide bound mixed valence Cytochrome c Oxidase (120 µM) and caged oxygen compound 3 (3 mM) as an aqueous solution, following a single laser pulse (5 ns, ~ 1 J/cm2, 355 nm). The negative feature at 560 nm is a result of a slight over-reduction of the enzyme during preparation which is oxidized upon oxygen liberation. The spectral contribution of the caged oxygen compound has been removed and the spectrum was cut at 440 nm due to light saturation

Full size image

The work presented herein paves the way for the broader implementation of caged oxygen compounds in time-resolved studies, particularly applications towards serial crystallography. Our newly synthesized compounds 2 and 3 demonstrate a marked improvement in stability profiles, addressing the pervasive challenge of compound premature degradation under our desired experimental conditions. However, the fortification of stability is accompanied by a compromise in the decreased efficiency of oxygen release. The ramifications of this are twofold. Enhanced stability extends the utility of these compounds, facilitating longer experimental durations with great improvements with regards to sample preparation, and minimizing the occurrence of premature degradation that could otherwise obscure accurate results. Conversely, the diminished efficiency requires an increased quantity of the compound, or the utilization of higher laser fluence, to accomplish the intended oxygen release levels. This interplay highlights the intricate equilibrium between stability and reactivity in the design of caged oxygen compounds. The protocol developed herein have been employed in a series of time-resolved X-ray investigations, the comprehensive findings of which will be delineated in subsequent publications.

3 Conclusion

In this study, we have successfully synthesized, characterized, and assessed several novel caged oxygen compounds, designed for advancing time-resolved studies of oxygen-depending systems. The presence of redox incompatibilities necessitated the development of an alternative reduction protocol, which has facilitated the general application of these compounds in experiments involving the reduced form of the CcO. Our experiments confirm that a single laser pulse is capable of liberating a significant amount of molecular oxygen from the caged compounds, sufficient to identify transient intermediates in the catalytic cycle of ba3-type CcO at concentrations viable for further crystallographic analysis. Notably, compounds 2 and 3 exhibit enhanced redox stability, mitigating the limitations previously encountered with compound 1, thus broadening the potential for application in systems that are necessitate reducing conditions. Furthermore, we present evidence for enzymatic turnover within microcrystals of reduced CcO, mediated by photoactive caged oxygen compounds.