Transient intermediates in LPMOs
Stopped flow UV-vis spectroscopy studies were performed for the reaction of the shunt reagent, meta-chloro-perbenzoic acid (m-CPBA), with three different CuI-LPMOs, which had been prepared by the reduction of a CuII-LPMO with either dithionite or ascorbate. One of the LPMOs (TaAA9) has a tryptophan residue close to the active site as described above, whereas two others (LsAA9 and CvAA9) have structures with a more distant tryptophan22,23 (Fig. 1c). In the former, in accord with previous studies20, we observed the appearance of two optical intermediates in stopped-flow measurements that lie on divergent pathways, viz. a tryptophanyl radical with characteristic bands at 520 and 548 nm (Fig. 2a), and a longer-lived tyrosyl radical with characteristic bands at ca. 420 nm (Extended data Fig. 1). These species are formed with 65% and 26% conversions, respectively, based on known extinction coefficients of neutral tryptophanyl and tyrosyl radicals.
In contrast, in the cases of CvAA9 (Extended data Fig. 2) and LsAA9 (Extended data Fig. 3), where the tryptophan is located ca. 10 Å away from the histidine brace active site, no significant oxidation of tryptophan was observed. Instead, oxidation of a tyrosine, as evidenced by clear absorption with lmax = 414 nm (hereafter Int2) was preceded by the appearance of a short-lived species with lmax = ~360 nm (hereafter Int1), and a further weak absorption in the 600 – 800 nm region of the spectrum. These kinetic data are in accord with the expected lower degree of electronic coupling of the tryptophan residue with the active site in LsAA9 and CvAA9 as compared to TaAA9, and would appear to confirm the hypothesis that, in the absence of substrate, the lifetime of a reactive intermediate, Int1, is controlled by coupling to adjacent redox-active amino acids. It should be noted that although similar intermediates are observed in both the LsAA9 and CvAA9 enzymes, there are some differences in the intensity and lifetime of each of these species.
Global modelling of the observed rates of formation and decay of the absorptions at 361 nm and 414 nm could be achieved with a sequential A®Int1®Int2®B model, where A is CuI-LPMO and B is the CuII resting state of the enzyme (assigned from the lack of any distinct spectral features in its visible spectrum at the enzyme concentrations used in these experiments and the CW-EPR spectra of samples after the reaction had complete, see below). The sequential nature of this reaction pathway, which likely involves, or is triggered by, O–O bond cleavage of the oxidant (see Discussion), is evident when comparing the kinetic transients at 361 nm and 414 nm (Fig. 2b inset). Using a molar absorption coefficient (Y• ε420 nm) of 2,600 M−1·cm−1, we estimate that ~35-40% of reduced LsAA9 and ~55% of reduced CvAA9 are converted to Int2 (via Int1), through this pathway. Using this model, we then determined the rates of formation and decay of Int1 and Int2 under a variety of conditions in order to establish the dependency, if any, of their lifetimes on: 1) the oxidant, 2) the reductant which was used to generate CuI-LPMO precursor, 3) the pH and 4) the buffer. To this end, we developed an expression system for LsAA9 in Escherichia coli, which gave access to the significant amounts of protein needed to survey all of the different conditions, and which further facilitated the site-directed mutagenesis studies described below (Supplementary information Fig. S1).
Within error, the rates of formation and reaction of Int1 and Int2 and their spectral features were found to be independent of pH (6 to 8), buffer (phosphate, MES) and initial reductant (Fig. S2-S4 and Extended data Fig. 4). Furthermore, no optical intermediates were observed when the CuII form of the enzyme was used, demonstrating that the observed stopped-flow spectra arise from oxidation of the CuI-state of the LPMO (Fig. S29). We then examined the role of H2O2, peracetic acid (PAA) and m-CPBA as oxidants, which all gave similar patterns of behaviour and, importantly for our later arguments, identical spectra for Int1 and Int2. The rate of conversion from Int1 to Int2 was found to be independent of H2O2 concentration (12.8 ± 0.1 s-1 and 11.5 ± 0.02 s-1 with 10 equiv. and 50 equiv. of H2O2, respectively) and oxidant (PAA and m-CPBA, 9.1 ± 0.02 s-1 and 9.0 ± 0.01 s-1, Fig. S5-S9). The rate of Int2 decay was also independent of oxidant concentration/identity (Fig. S5-S9). The only significant differences between oxidants were seen in the rates of formation of Int1 (Extended data Fig. 5). Noticeably, Int1 did not accumulate when a stoichiometric amount of H2O2 was used (Fig. S7) but was observed at higher equivalents. In contrast, Int1 formation was observed with stoichiometric quantities of the more reactive peroxy-acid oxidants, m-CPBA or PAA (Fig. S8, S9). For instance, rates of formation of 38.1 ± 0.1 s-1 and 38.9 ± 0.1 s-1 at 277 K were observed with one equivalent of m-CPBA and PAA respectively, and rates of 14.8 ± 0.2 s-1 and 99.1 ± 0.3 s-1 at 277 K were observed with 10 equiv. and 50 equiv. of H2O2, respectively (Fig. S5-S9). The rates of Int1 and Int2 formation were also measured in D2O buffer to determine whether any solvent kinetic isotope effects (SIEs) were associated with either step. An SIE of 1.6 ± 0.2 was observed on the rate of formation of Int1, which is consistent with the value reported in pathways involving heterolytic cleavage of the oxidant O–O bond20 (Extended data table 1).
Intermediates 1 and 2 do not oxidize oligosaccharide substrates
To explore the reactivity of Int1 with oligosaccharide substrates, we next carried out stopped-flow measurements in the presence of cellopentoase (G5), a known substrate for LsAA912. No optical intermediates were observed when oxidants were added to a mixture of G5 and CuI-LsLPMO (Fig. S31), showing that Int1 and Int2 are not formed under these conditions, or that their rate of decay is much faster than their rate of formation in the presence of substrate. We therefore turned to double mixing stopped-flow experiments, where CuI-LsLPMO is mixed first with stoichiometric amounts of oxidant, allowed to age for 50 ms to generate a maximal concentration of Int1, and then mixed with G5. Under these conditions, Int1 decayed ca. 40-fold faster in the presence of G5 than in the absence of substrate (i.e. with 500 μM G5, rate = 225.1 ± 3.2 s-1 vs 6.0 ± 0.1 s-1 with buffer only) (Fig. 2c and S16). Int2 decay (bi-exponential) was also accelerated (i.e. with 500 μM G5 = 38.4 ± 0.8 s-1 vs 0.20 ± 0.001 s-1 with buffer only). These rates of decay were dependent on substrate concentration (Fig. S11–S19). However, the presence of the G2 product in the double mixing experiments had no effect on the rate of decay of either intermediate (Fig. S20, S21).
Despite the fact that the rates of decay of Int1 and Int2 are accelerated by substrate binding, it cannot be inferred from these data that either intermediate is part of the oxygenase’s productive catalytic cycle. As such, we undertook two experiments to determine the role, if any, of Int1 and Int2 in the productive catalytic cycle of LsAA9. In both experiments, the double mix experiment described above was repeated using both H2O2 and m-CPBA as oxidants but mixed at the second stage with either: i) a known cellotetraose substrate for LsAA9 which fluoresces upon cleavage, or ii) analysing the double-quench samples described above (with variable double-mixing times of 50, 300 and 1000 ms after the initial mix) for any cleavage products and their amounts as a function of Int1 or Int2 concentration (Extended data Fig. 6 and 7). In both cases, no evidence of intermediate-dependent substrate cleavage was found, showing that neither Int1 nor Int2 are likely to be parts of the productive catalytic cycle of LsAA9. In contrast, control experiments where reactions are initiated by the addition of H2O2 to a mixture of substrate and enzyme lead to product formation as anticipated.
Intermediate 2 is a ferromagnetically-coupled (S = 1) CuII-tyrosyl complex
The relatively long lifetime (1–5 s) of Int2 allowed us to trap this species using standard freeze-quench methods. Thus, using an Int2 sample quenched after ~1 s, we investigated its temperature and power-dependent CW-EPR spectra and compared these to two control samples (CuII-LsLPMO and a sample quenched after 5 s). Spectra of Int2 collected over the temperature range of 8 – 20 K showed the presence of three different EPR active species for the ~1 s sample; i) signals from a CuII ion (S = ½), with spin Hamiltonian parameters that match that of the LsAA9 resting state, ii) a sharp signal from an organic-like radical (g ~ 2.00) without noticeable magnetic-splitting (S = ½) and with low intensity (ca. < 1% of the integrated CuII intensity) and iii) half-field, forbidden transitions, ‘DmS = ±2’ at 1500 G (Fig. 2d) and allowed ‘DmS = ±1’ transitions. The observation of the two S = ½ signals (resting-state CuII signal and sharp signal at g ~ 2.00) show that the trapped sample contains products that likely arise from the direct homolytic fission of the peroxide bond to give CuII-OH and an organic-based radical. From modelling of the low temperature EPR data, this constitutes ca 62% of the whole sample, in good agreement with stopped-flow data, and shows that this is the dominant pathway, similar to a previous report20. In addition to the products from the homolytic pathway, the half-field transition and its temperature- and power dependent behaviour reveal the formation of a separate triplet state (S = 1) species (calculated dipolar coupling of T = [+3417 +917 -4334], and J = -100 MHz (modelling are not sensitive to the magnitude of the exchange coupling), spin-spin distance is given for two extreme cases with dipolar interaction tensor is assumed to be approximately axial, T ~ [-T, -T, +2T] with T = 2167 MHz (Cu…O = 2.88 Å) and a rhombic dipolar tensor, T ~ [-T, 0, +T] with T = 4344 MHz (Cu…O = 2.29 Å)). In combination with the fact that Int2 has clear absorption features that match those of a tyrosyl radical, such a coupling can only realistically arise from unpaired electrons on CuII ion and a tyrosyl radical that lies adjacent to the Cu in the active site of LPMOs (Cu…O = 2.6 Å, Fig. 1). The triplet EPR signals are not observed in CuII-LsLPMO nor in the sample freeze-quenched after ~0.5 and ~5 s (Fig. 2d, S32-S40), indicating direct correlation between the population of tyrosyl radical observed in stopped-flow measurement and formation of the triplet state (S = 1).
Mutagenesis of active site residues perturbs intermediate formation and decay
To confirm the identity of the Tyr radical in Int2, we prepared Y164F and W64F variants of LsAA9. The active site Tyr164 and the neighbouring Trp64 are thought to form part of a charge transfer pathway from the active site to the protein surface14. Stopped-flow analysis of Y164F showed the formation of a single intermediate species (Fig. S22), with very similar absorbance features to Int1. This intermediate is not converted into Int2, but instead decays to a non-descript spectrum at a rate similar to that of Int2 decay in the wild-type enzyme. In contrast to the wild type enzyme, temperature- and power-dependent CW-EPR analysis (5 – 20 K) on a freeze-quench trapped sample (0.5 – 1 s after reaction, 60-80% Int1 relative concentration, Fig. 3) revealed no half-field signals associated with a triplet species, showing that Int2 is not formed in the reaction and that Int1 is not EPR active (Fig. S40, see below for further discussion).
Oxidation of the reduced W64F variant leads to the formation of Int1 and Int2 (Fig. S23) at similar rates to wild-type LsLPMO. However, the rate of decay of Int2 in W64F is >10-fold slower than in the wild-type enzyme, suggesting the tyrosyl radical of Int2 is reduced by rapid ‘hole-hopping’ from Trp64. This behaviour is consistent with the site of tyrosyl radical identified in EPR studies of Int2 being Y164, affirming the identity of Int2 as a ferromagnetically coupled CuII…Y164• pair.
The roles of two additional conserved residues, His147 and Gln162, in intermediate formation and decay were also explored. His147 lies adjacent to the histidine brace and participates in π-stacking interactions with His78. With His147 replaced by Phe, Int1 and Int2 were formed as in the wild-type enzyme, albeit at slightly different rates (k1 = 61.9 ± 0.2 s-1 and k2 = 3.6 ± 0.01 s-1), showing that His147 does not contribute significantly to the electronic features of Int1 and Int2 (Fig. S24). Interestingly no intermediates were observed upon oxidation of H147A and H147Q variants, suggesting that p-stacking interactions with His78 may be important for Int1 formation and/or stability, plausibly by restricting accessible conformations of His78 (Fig. S25, S26). We next explored the role played by Gln162. This residue is conserved across all AA9 LPMOs and is well-positioned to interact with H2O2 (or O2 derived species) bound at the vacant equatorial coordination site. Previous studies have shown that mutation of this conserved glutamine abolishes catalytic activity with both H2O2 and O2 as oxidants24, and QM/MM calculations invoke an important role for the glutamine in orientating a peroxide within the active site25. Stopped-flow measurements with a Q162A variant revealed that no intermediates are accumulated upon oxidation with H2O2 (Fig. S28), consistent with Gln162 playing a role in the formation of a CuI…HO–OH species with the peroxide bound at the active site. In contrast, using m-CPBA as the oxidant, where the O–O bond strength is expected to be weaker than in H2O226, led to rapid accumulation of Int2 (Fig. S27) with no observable accumulation of Int1. Although not observed directly, we presume Int2 formation still proceeds via Int1, but that the Q162A mutation decreases Int1 stability, preventing its accumulation on the timescale of the experiment before formation of Int2.
Intermediate 1 is an open-shell singlet (S = 0) CuII-(histidyl radical)
The characterization of Int2 as a CuII-tyrosyl radical pair is informative as to the nature of Int1, not least in the fact that Int1 must exist at an oxidation state level which is one higher than a formal CuII state. Given the potential of Int1 to be a high valent Cu-based species, therefore, we sought to trap this species and determine its Cu oxidation state and electronic nature. To this end, we took advantage of the extended lifetime of Int1 in the Y164F variant, using rapid freeze quench methods (at 100 ms after reaction, Supplementary information), to characterize it using a combination of CW-EPR spectroscopy and HERFD-XAS.
Consistent with the EPR-silent nature of Int1 described above, CW-EPR spectra at 150 K of Y164F Int1 trapped at 100 ms also showed no signals associated with a triplet species. Subsequent annealing of the sample from 150 K in 10 K steps to room temperature afforded a single CuII spectrum with spin Hamiltonian parameters (gz = 2.265, ǀAzǀ = 445 MHz) consistent with a CuII state of LsAA9, where this species starts to appear in the spectrum at ~220 K (Fig. S41). Such magnetic behaviour reveals that the sample trapped at 100 ms post-mixing is mostly Int1, and—in combination with the EPR data described above—is likely an S = 0 spin singlet, although whether this is in an open or closed-shell form cannot be determined from the EPR data alone. As a control, a freeze-quench sample trapped at 380 ms gave a single CuII spectrum of the resting state, which only slightly gained intensity upon annealing to 273 K, showing that the sample trapped at this timescale is likely a mixture of Int1 and the final product of the reaction, but dominated by the latter.
Copper K-edge HERFD-XAS spectra (collected at 10 K to avoid significant photoreduction of the sample, Supplementary information) of samples trapped at 100 and 380 ms, and subsequent annealing through a temperature range of 150 to 250 K, revealed in all cases an edge position at 8985.8(3) eV, consistent with a sample that is principally in the CuII oxidation state19. An intense rising-edge feature at 8981.9(3) eV is also observed, likely to be either a CuII shakedown transition often observed in the K-edge XANES spectra of CuII species27, or a 1s-4p transition that is observed in the equivalent CuI form of the enzyme at the same position, which could arise from partial photoreduction of the sample19. In the samples trapped at 100 ms, which had been annealed to 150 K and 250 K, two weak pre-edge features at 8977.7(3) eV and 8979.2(3) eV were also observed (Fig. 4 and S44), which are not observed in the sample trapped at 380 ms, at which stage it is expected that the concentration of Int1 will have been substantially reduced.
The weak pre-edge feature at 8977.7(3) eV is assigned to a dipole-disallowed, quadrupole-allowed 1s-3d(x2-y2) transition, commonplace in CuII enzymes and complexes. The transition is also commensurate with previous Cu K-edge XAS studies on wild-type CuII-LsAA9, which exhibit the same transition at 8977.4 eV19. The pre-edge feature at 8979.2(3) eV, on the other hand, is not a common absorption feature in CuII complexes, and—in the CuII oxidation state—can only realistically arise from a dipole-forbidden Cu(1s) to ligand LUMO/SOMO transition. Moreover, for such a transition to have any intensity, the ligand orbital must have some overlap with the orbitals of the metal. Such requirements are fulfilled by a radical based on a ligand that is directly coordinated to the CuII. Indeed, such a weak pre-edge transition has been observed before for oxidized forms of LsAA9 at 8982.8 eV, which in this earlier case was assigned to a MLCT transition in an inactive CuII-tyrosyl species (which, at pH 7, is only formed upon prolonged exposure to hydrogen peroxide)19. However, we can rule out the source of the absorption at 8979.2(3) eV in the Int1 sample, as arising from the previously reported CuII-tyrosyl species in LsAA9, since the latter occurs at 8982.8 eV, some 3.6 eV higher in energy.
In making an assignment of the pre-edge transitions in Int1, we turn to previous work on the UV-vis spectra of transient imidazolyl radicals, CuII-(imidazolyl radicals) and CuII-(histidyl radicals), all of which exhibit semi-intense absorptions at ca. 360 nm (e > 2,000 M-1 cm-1)28,29, and also that of an isolated imidazolyl radical with an intense band at 365 nm (e ~ 2,000 M-1 cm-1)30. The analogy between these absorptions and that in Int1 directs us towards the possibility that Int1 is, in fact, a CuII-(histidyl radical) complex. Indeed, given the need from the HERFD-XAS data for the radical ligand to have significant overlap with the copper orbitals, a histidyl radical offers itself as one of only two viable species, where the other is one in which the radical ‘hole’ exists on the exogenous ligand coordinated to the CuII. While it is not possible to separate the two possibilities on the basis of the combined UV-vis and HERFD-XAS data, TD-DFT calculations for the pre-edge region of the XAS of Int1 were performed; these calculations give reliable predictions into the number and relative intensities of any transitions of the possible species31. As such, we performed TD-DFT calculations on an optimized structure of the active site of LsAA9 for a CuII-(histidyl radical) where the hydrogen atom of the C2 carbon of His1 had been removed to give the His1-radical. The C2 site for radical formation was chosen in the knowledge that C2-H atom abstraction to give histidyl radicals has been observed for CuI-superoxide dismutases treated with peroxide32, and that 2-oxo-histidine is the principal product from the reaction of AA10 LPMOs with H2O2 in the absence of substrate14. We performed the analogous calculations on an optimized structure of the active site of LsAA9 for a CuII-oxyl.
As expected, TD-DFT calculations on both CuII-oxyl and CuII-(histidyl radical) species afford two weak pre-edge transitions in the XAS spectrum: 1s-3d(x2-y2) and 1s-ligand radical transitions (Fig. 4). For the CuII-(histidyl radical) the calculated energy separation of the two transitions is 2.3 eV with near equal intensity, which is a close match to the experimental separation of 1.8(4) eV and the relative intensities of the two pre-edge peaks following correction for the rising background signal of the pre-edge transition which, is not modelled in the TD-DFT calculations. (Fig. 4). The equivalent TD-DFT data for the CuII-oxyl species give a separation of 3.6 eV of two peaks with unequal intensity. Thus, the combined data from XAS, UV-vis spectroscopy and TD-DFT are most consistent with Int1 being a transient CuII-(histidyl radical), which is formed following treatment of CuI-LsAA9 with peroxides.
Int1 and Int2 are part of an enzyme repair pathway
To explore the functional role of the Int1/Int2 charge transfer pathway, we next carried out steady-state assays with wild-type LsAA9 using a known cellotetraose substrate which fluoresces upon cleavage. For comparison, assays were also performed with the Y164F variant where the charge transfer pathway has been disabled. Assays were performed in the presence of H2O2 and were initiated by the addition of ascorbate as a reductant. The wild-type and Y164F variants show similar catalytic behaviour, with rapid initial formation of the anticipated fluorescent product followed by apparent deactivation (Fig. 5). However, the origin of the observed deactivation is different in the two variants. In the wild-type enzyme, activity can be restored by the addition of reductant, ultimately resulting in fluorescence intensity anticipated following complete conversion of substrate to product. In contrast, the activity of the Y164F variant cannot be restored with reductant and increases in reaction conversion can only be achieved by the addition of fresh enzyme. Taken together, from these experiments we can infer that the Int1/Int2 charge transfer pathway forms part of an in-built enzyme repair mechanism that protects the active site from oxidative damage during uncoupled turnover by restoring the copper-histidine brace to its resting state, which can then re-enter the catalytic cycle through reduction (Fig. 5).