Changes in active-site geometry on X-ray photoreduction of a lytic polysaccharide monooxygenase active-site copper and saccharide binding

The recently discovered lytic polysaccharide monooxygenases (LPMOs) are Cu-containing enzymes capable of degrading polysaccharide substrates oxidatively. The generally accepted first step in the LPMO reaction is the reduction of the active-site metal ion from Cu²⁺ to Cu⁺. Here we have used a systematic diffraction data collection method to monitor structural changes in two AA9 LPMOs, one from Lentinus similis (LsAA9_A) and one from Thermoascus auranti­acus (TaAA9_A), as the active-site Cu is photoreduced in the X-ray beam. For LsAA9_A, the protein produced in two different recombinant systems was crystallized to probe the effect of post-translational modifications and different crystallization conditions on the active site and metal photoreduction. We can recommend that crystallographic studies of AA9 LPMOs wishing to address the Cu²⁺ form use a total X-ray dose below 3 × 10⁴ Gy, while the Cu⁺ form can be attained using 1 × 10⁶ Gy. In all cases, we observe the transition from a hexa­coordinated Cu site with two solvent-facing ligands to a T-shaped geometry with no exogenous ligands, and a clear increase of the θ₂ parameter and a decrease of the θ₃ parameter by averages of 9.2° and 8.4°, respectively, but also a slight increase in θ_T. Thus, the θ₂ and θ₃ parameters are helpful diagnostics for the oxidation state of the metal in a His-brace protein. On binding of cello-oligosaccharides to LsAA9_A, regardless of the production source, the θ_T parameter increases, making the Cu site less planar, while the active-site Tyr—Cu distance decreases reproducibly for the Cu²⁺ form. Thus, the θ_T increase found on copper reduction may bring LsAA9_A closer to an oligosaccharide-bound state and contribute to the observed higher affinity of reduced LsAA9_A for cellulosic substrates.

1. Introduction

25–50% of all proteins contain metal cofactors (metallo­proteins) (Hoppert, 2011; Waldron et al., 2009) and govern a range of biological functions, such as non-enzymatic electron charge transfer, metal transport and storage (Hoppert, 2011), and catalysis (Chen et al., 2019). Redox activities, specifically, are often associated with enzymes containing iron and copper (Hoppert, 2011; Bowman et al., 2016). Protein Cu centres are involved in metal transport, regulation and oxidation of various com­pounds (Festa & Thiele, 2011), and are often defined by the number and character of ligands coordinating the metal (Solomon et al., 2014).

In X-ray crystallography experiments, crystals of metalloproteins absorb energies in the X-ray range more readily than a de-metallized form of the protein due to the presence of the transition metals (Handing et al., 2018). Metalloproteins are therefore more predisposed to radiation damage than proteins not containing metals. With the increase in flux and intensity of X-rays at the new synchrotron beamlines, X-ray studies on metalloproteins become more challenging. It is estimated that only 10% of the inter­acting X-ray photons are scattered by crystals during data collection (Bowman et al., 2016). The inter­acting but non-scattering photons may be absorbed and result in either heat increase or radiation damage (Helliwell, 1984). Photoreduction is therefore commonly observed in the structures of metalloproteins determined by X-rays. Metal centres may be found to change electronically and structurally (Bowman et al., 2016), as can be monitored by spectroscopy and/or shown by increasing bond distances and altered bond angles between the metal and its ligands (Bowman et al., 2016; Frankaer et al., 2014; Corbett et al., 2007; Antonyuk & Hough, 2011; Yano et al., 2005). Such distorted metal coordination geometry may lead to the incorrect identification of the metal ion and the catalytic mechanism, especially of a redox protein.

When performing a diffraction experiment, the incoming photons of the X-ray beam may reduce the metal of metalloproteins even at low doses. This makes the structure determination of the native/resting state difficult. One way of circumventing photoreduction is to use neutrons instead of X-rays for crystallographic studies of metalloproteins (Schröder & Meilleur, 2021), though this imposes severe limitations on the systems that can be studied, as much bigger crystals are required for neutron diffraction. Furthermore, reduction of transition metals by X-rays may be a convenient reaction trigger for studies of enzymatic reactions in crystals (Bourgeois & Weik, 2009; Bourgeois & Royant, 2005). Two strategies used to study reaction inter­mediates structurally are either a real-time approach (serial crystallography) or a trapping approach (conventional crystallography) (Bourgeois, 2017; Schotte et al., 2012).

Lytic polysaccharide monooxygenases [LPMOs, reviewed in Tandrup et al. (2018), Vu & Ngo (2018), Wang et al. (2020) and Eijsink et al. (2019)] were discovered just over 10 years ago and are metalloenzymes containing a type II Cu centre (Quinlan et al., 2011). A Cu²⁺ ion is coordinated by two histidines in a motif known as the histidine brace (His-brace), where the N-terminal His is often N^ɛ2-methyl­ated in LPMOs of fungal origin and coordinates Cu using both the N-terminal nitro­gen and the N^δ1 atom (see Fig. 1) (Quinlan et al., 2011). The His-brace motif is also found in other non-LPMO copper proteins (Labourel et al., 2020; Garcia-Santamarina et al., 2020; Udagedara et al., 2019; Cao et al., 2018). Known LPMOs degrade lignocellulose-, hemicellulose-, chitin-, starch- or pectin-containing biomass oxidatively, and have a boosting effect on saccharification together with glycoside hydro­lases (Harris et al., 2010; Lo Leggio et al., 2015; Quinlan et al., 2011; Vaaje-Kolstad et al., 2010; Zerva et al., 2020), thus attracting considerable industrial inter­est for the production of biofuel (Hemsworth et al., 2013; Müller et al., 2015; Beeson et al., 2015; Johansen, 2016) and other biotechnological applications (Ipsen et al., 2021). LPMOs are widespread in nature and have been identified in all domains of life, though so far not in mammals. In the CAZy (Carbohydrate-Active enZymes) database, they are classified as Auxiliary Activity families 9–17 (family 12 excluded) (Lombard et al., 2014), with AA17 recently discovered in oomycetes having activity against pectin (Sabbadin et al., 2021).

Figure 1 Illustrated is a simplified lytic polysaccharide monooxygenase (LPMO) reaction scheme. (a) LPMOs can engage in a variety of reactions depending on the specific enzyme. After Cu reduction, reactions with O2 and/or H2O2 may occur, involving saccharides or a variety of small mol­ecules, in some cases in several steps (Brander, Tokin et al., 2021; Brander, Lausten et al., 2021). Saccharide substrates produce smaller oxidized and/or non-oxidized saccharides as products. The first step in the reaction can be triggered either chemically by a reducing agent or using X-ray photons. The transition from Cu^2+ to Cu^+ results in expulsion of the equatorial (eq) and axial (ax) ligand water mol­ecules in the type II Cu site. These ligands may be different based on the crystallization liquor com­ponents. Upon binding of the substrate, the axial water mol­ecule is displaced. The LsAA9_A His-brace is shown at low (left; PDB entry 7pxi) and high (right; PDB entry 7pxn) X-ray dose (see results). (b) The overall structure of LsAA9_A binding cello­tetra­ose (PDB entry 6ydg). (c) The overall structure of TaAA9_A (PDB entry 3zud). (d) The bond angles referenced throughout the article are defined in Vu & Ngo (2018). Angles θ1, θ2 and θ3 are measured from N—Cu—N as indicated. θH–H is the angle between the two His imidazole ring planes. θT is the angle between the N^δ1—Cu—Nam plane and the N^ɛ2—Cu line. θH1 (also denoted θHN) is the angle between one His imidazole ring plane and the line from Cu to the Cu-inter­acting N atom of the same imidazole ring.

LPMOs cleave the glycosidic bond, producing both oxidized (C1 and/or C4) and non-oxidized chain ends. Their mechanism (see the simplified LPMO reaction scheme in Fig. 1) has been researched extensively since their discovery. An essential (though probably not rate limiting) step is the reduction of the metal cofactor Cu²⁺ to Cu⁺, which in nature can be carried by a number of electron donors, such as small mol­ecules or proteins found in their natural environment (Kracher et al., 2016; Várnai et al., 2018; Brenelli et al., 2018; Frommhagen et al., 2017; Wang et al., 2020). In vitro reductants such as ascorbic acid, gallic acid or cysteine are commonly used (Wang et al., 2020). As well as being essential for further reactivity, there are several reports indicating increased polysaccharide affinity on reduction of the active-site metal, suggesting that polysaccharide binding precedes further steps in the mechanism (Kracher et al., 2018; Hangasky & Marletta, 2018; Filandr et al., 2020; Brander, Tokin et al., 2021). Both O₂ and H₂O₂ have been suggested as natural co-substrates (Bissaro et al., 2017; Hangasky et al., 2018) and the LPMO mechanism is being assessed continuously for the involvement of either/both (Bissaro et al., 2020; Hedegård & Ryde, 2017; Courtade et al., 2020; McEvoy et al., 2021; Brander et al., 2020). The dependence on H₂O₂ for saccharide cleavage has been established recently specifically for one of the model LPMOs investigated here (Brander, Tokin et al., 2021), and is generally gaining increasing support. The H₂O₂-driven mechanism has the advantage that the LPMO active-site metal only needs to be reduced once, while in the O₂-driven mechanism, it needs to be reduced at every cycle (Bissaro et al., 2017). However, too high levels of H₂O₂ cause oxidative damage and destruction of the enzymes, com­plicating its use as co-substrate (Müller et al., 2018), and H₂O₂ will react with other chemical species, leading to unwanted deca­rboxylating reactions in com­plex substrates, such as lignocellulose, under industrially relevant conditions (Peciulyte et al., 2018).

For LPMOs, X-rays may possibly circumvent the need for additional reducing agents in structural studies of the reaction (Fig. 1). Previously, photoreduction of LPMOs has been studied in detail for members of AA10 and AA13 (Gudmundsson et al., 2014; Muderspach et al., 2019), and somewhat more qualitatively for AA9 (Frandsen et al., 2016).

Here we further investigated the gradual photoreduction of the LPMO active-site Cu²⁺ in a systematic way and monitored the structural changes in two AA9 LPMOs which have been studied extensively. We studied the primarily C4-oxidizing AA9 LPMO from the fungus Lentinus similis (LsAA9_A), which is active on soluble cello-oligosaccharides and polysaccharides. We exposed crystals of ligand-free and cello-oligosaccharide-binding LsAA9_A to different radiation doses. To investigate the effect of post-translational modification and/or crystallization conditions on the metal-centre X-ray-induced perturbation, we carried out a similar study with LsAA9_A produced in E. coli [LsAA9_A(Ec)], thus devoid of glycosyl­ation and His methyl­ation. We determined a total of six structures per substrate-free enzyme and six structures per enzyme solved with bound cello-oligosaccharide. Additionally, we have solved six structures of the C1/C4-oxidizing AA9 from Thermoascus auranti­acus (TaAA9_A, where oligosaccharides bound in the structure have so far not been obtained) to identify similarities between AA9s as the central Cu ion is reduced. We report how inter­atomic distances and other geometric parameters within the active site change as a function of the X-ray dose, and as Cu transitions from Cu²⁺ to Cu⁺. We have correlated our photoreduced structures to the chemically reduced structure of LsAA9_A and two structures solved from data collections at room tem­per­ature. Furthermore, we have monitored photoreduction in a Serial Synchrotron Crystallography (SSX) experiment of LsAA9_A. The results of the study are relevant both for the general study of metalloproteins by X-ray crystallography and the specific understanding of the LPMO mechanism.

2. Methods

2.1. Crystallization

Crystallization via sitting-drop vapour diffusion was performed in MRC two-drop plates set up by an ORYX-8 robot (Douglas Instruments), unless otherwise stated. Cello-oligosaccharides used for substrate-bound structures were purchased from Megazyme.

The fungal enzyme LsAA9_A was expressed in A. oryzae [LsAA9_A(f)], purified and deglycosyl­ated with endoH as described previously (Frandsen et al., 2016; Simmons et al., 2017; Tandrup et al., 2020). The protein sample was pre-incubated with Cu²⁺ acetate in equimolar amounts for 1 h at 4°C prior to setup. Crystals of LsAA9_A(f) were grown via sitting-drop vapour diffusion. Crystallization and optimization of the crystals has been described previously (Frandsen et al., 2016). The two crystals used for determining 12 of the structures presented here were grown with a reservoir consisting of 3.3 M NaCl and 0.1 M citric acid at pH 3.5. Drop and reservoir volumes were 0.5 and 100 µl, respectively, with a drop com­position of 0.3 µl protein (19 mg ml⁻¹ in 20 mM sodium acetate pH 5.5), 0.1 µl reservoir and 0.1 µl milli-Q water. Crystals appeared within 2 d. Prior to harvesting, the crystals were moved to a drop of 3.0 M NaCl and 0.1 M citric acid at pH 5.5 for 10 min to ensure an ordered active site (Frandsen et al., 2017). Subsequent soaking with oligosaccharide ligand was done for one of the crystals by moving it to a drop containing 3.0 M NaCl, 0.1 M citric acid pH 5.5, as well as 1.0 M of Cell₄ for 10 min. Crystals were flash frozen in liquid nitro­gen without any added cryoprotectant.

The LsAA9_A(f) crystal used for ascorbic acid soak was grown in a VDX plate (Hampton Research) by hanging-drop vapour diffusion. The crystal was grown in a 2 µl drop of 1.5 µl 3.5 M NaCl, 0.1 M citric acid pH 4.0 and 0.5 µl enzyme (23 mg ml⁻¹ in 20 mM Na acetate pH 5.5), with a reservoir of 1 ml. After growing a crystal of approximately 200 µm in each dimension, it was transferred to a drop consisting of 3.5 M NaCl, 0.1 M citric acid pH 5.5, 0.01 M ascorbic acid pH 5.5 for 20 min before flash freezing.

For room-tem­per­ature data, an LsAA9_A(f) crystal was grown using 15 µl 20.8 mg ml⁻¹ enzyme (Cu-loaded using Cu acetate in equimolar amounts) in 5 µl 3.0 M NaCl, 0.1 M citric acid pH 3.5 in a VDX plate (Hampton Research). The crystal grew to approximately 200 µm in each dimension within a few weeks and was transferred to a solution of 3.0 M NaCl, 0.1 M citric acid pH 5.5 for 10 min before being mounted on a cryo-loop (MiTeGen). A MicroRT capillary (MiTeGen) was pre­pared to contain 20 µl of 3.0 M NaCl, 0.1 M citric acid pH 5.5. The cryo-loop-mounted crystal was inserted into the MicroRT capillary to prevent the crystal from drying during data collection.

A room-tem­per­ature data set was collected at the BioMAX beamline of the MAX IV Laboratory from a crystal grown in 0.1 µl 3.0 M NaCl, 0.1 M citric acid pH 3.5 and 0.3 µl 19 mg ml⁻¹ LsAA9_A(f), also Cu-loaded. The crystal was equilibrated for pH, as above. The crystal was mounted in a 1.0 mm diameter glass capillary (WJM glass), plugged with 3.0 M NaCl, 0.1 M citric acid pH 5.5 in both ends and sealed with beeswax.

LsAA9_A(Ec) was produced in E. coli using the LyGo platform with the pLyGo-Ec-6 expression vector (Hernández-Rollán et al., 2021). The sample was purified using Q-Sepharose anion-exchange chromatography with a column equilibrated with 25 mM Bis-Tris pH 5.87. The sample was eluted using the same buffer over a 0–500 mM NaCl gradient. The protein, in 25 mM Bis-Tris pH 5.87 and 175 mM NaCl, was initially screened for crystallization conditions at a concentration of 1.83 mg ml⁻¹ in 0.3 µl drops (3:1 and 1:1 protein-to-reservoir ratio) with the commercially available Index (Hampton Research), PACT (Qiagen) and JCSG+ (Mol­ecular Dimensions) crystallization screens. At least 30 min before crystallization setup, equimolar quantities of acetate were added to the enzyme sample. Crystals were formed at room tem­per­ature in the A2 conditions of the Index screen [0.1 M sodium acetate trihydrate pH 4.5, 2.0 M (NH₄)₂SO₄], which was further optimized to 0.1 M sodium acetate pH 4.5 and 1.8 M (NH₄)₂SO₄. Crystals grown under this condition were flash frozen without added cryoprotectant.

An initial oligosaccharide-binding structure of LsAA9_A(Ec)-Cell₃ was obtained from a crystal soaked for 30 min in a 3 µl drop suspended over a reservoir with 100 µl 2 M (NH₄)₂SO₄ and 0.1 M sodium acetate in a VDX plate. The drop consisted of 16% glycerol, 0.3 M Cell₃ and LsAA9_A(Ec) protein stock to facilitate a final concentration of 0.3 mg ml⁻¹. The soaked crystals were mounted in fibre loops and flash frozen.

Crystals of LsAA9_A(Ec) soaked in cello­tetra­ose (Cell₄) were grown by the batch crystallization method in 1.5 ml micro­fuge tubes (Fisher Scientific) using a 1.8 mg ml⁻¹ protein concentration. The crystallization condition was 2.3 M (NH₄)₂SO₄, 0.1 M sodium acetate pH 4.2. A 200 µl batch was set up at 4°C with the protein and the crystallization condition in a 1:3 ratio. The crystals appeared within 4 d. For soaking, 0.5 µl of the crystal slurry was added to a drop containing 0.05 M Cell₄ in the crystallization condition with 18% glycerol and incubated from 5 to 30 min. The crystals were then flash frozen in liquid nitro­gen.

The crystals of LsAA9_A(Ec) used for photoreduction were grown in microbatch crystallization plates. Crystals not soaked in substrate were grown in 4 µl drops com­posed of 3 µl 2.0 M (NH₄)₂SO₄, 0.1 M sodium acetate pH 4.5 and 1 µl LsAA9_A(Ec) (1.8 mg ml⁻¹ in 25 mM Bis-Tris pH 5.8 175 mM NaCl). Crystals soaked in Cell₃ were grown in 3 µl drops com­posed of 2 µl 2.0 M (NH₄)₂SO₄, 0.1 M sodium acetate pH 4.5 and 1 µl LsAA9_A(Ec) (1.8 mg ml⁻¹ in 25 mM Bis-Tris pH 5.8, 175 mM NaCl). The drops were covered in a layer of paraffin and silicon oils in a paraffin–silicon ratio of 3:2. The drops were set up at 4°C and crystals appeared within 4 d. Crystals used in a soaking experiment were moved into a drop containing 0.5 M Cell₃, 2.0 M (NH₄)₂SO₄ and 0.1 M sodium acetate pH 4.5. Prior to flash freezing in liquid nitro­gen, the crystals were cryoprotected in 20%(w/v) glycerol, 2.0 M (NH₄)₂SO₄, 0.1 M sodium acetate pH 4.5.

TaAA9_A was produced as described in Harris et al. (2010) and purified using Q-Sepharose anion-exchange chromatography with a column equilibrated with 20 mM Tris pH 6.0. TaAA9_A was eluted in the same buffer over a 0.79–2.5 mM NaCl gradient. The sample was deglycosyl­ated in 20 mM MES pH 6.0, 125 mM NaCl by incubation with ∼0.05 units per mg TaAA9_A of endoH (Roche Diagnostics, 11643053001), and then buffer exchanged to 20 mM Na acetate pH 5.5. Prior to crystallization the sample was incubated for 1 h with equimolar amounts of Cu acetate. Crystals of TaAA9_A were grown in 0.4 µl drops com­posed of 0.3 µl TaAA9_A (15.9 mg ml⁻¹ in 20 mM sodium acetate pH 5.5) and 0.1 µl 0.02 M MgCl₂, 0.1 M HEPES pH 7.5, 22%(w/v) polyacrylic acid 5100 sodium salt.

2.2. Single-crystal X-ray data collection, structure determination and refinement

Data for the initial structure determination of LsAA9_A(Ec) were collected at 100 K at the BioMAX beamline at MAX IV (Lund, Sweden) (Ursby et al., 2020). The images were col­lected for 360° with an oscillation of 0.1° and a 100 ms exposure time for each image using a wavelength of 1.1 Å. The BioMAX beamline was also used to collect 360° data with an oscillation of 0.1° and an 11 ms exposure time for each image for LsAA9_A(Ec)-Cell₄ crystals. The autoprocessed (autoPROC pipeline) reflection file was used for structure determination. The highest resolution cut-off of diffraction data was generally chosen based on a CC_1/2 in the outer shell of above 50%.

For LsAA9_A(f), a data set was collected at room temper­ature using an in-house diffractometer at the Oak Ridge National Laboratory. The set-up consisted of a Rigaku HighFlux HomeLab instrument with a MicroMax-007 HF X-ray generator and an EIGER R 4M detector. Data were collected for 98° with 0.25° oscillation and a 10 s exposure per image. The data were processed using CrysAlis PRO (Rigaku) and scaled using AIMLESS/POINTLESS (Evans & Murshudov, 2013; Evans, 2006). Room-tem­per­ature data were also collected for LsAA9_A(f) at the BioMAX beamline of the MAX IV synchrotron. The glass capillary mounted sample was illuminated with X-rays for 360° with an oscillation of 0.1° and an exposure per frame of 0.01 s. The data were processed using XDS/XSCALE.

Data for the photoreduction study were collected at the P11 beamline of PETRA-III at DESY, Hamburg (Burkhardt et al., 2016). For each of the data sets, images were collected for 360° with an oscillation of 0.1° and a 100 ms exposure time for each image. Three data sets were collected on LsAA9_A(f) and LsAA9_A(f)-Cell₄ with the beam transmission set to 1, 10 and 100%. This corresponded to estimated photon fluxes of 10¹⁰, 10¹¹ and 10¹² photons/s, respectively. Each data collection was performed without changing the crystal orientation between collections. Subsets of the three data sets (two for each) were processed with the fewest number of images possible for a com­pleteness of ∼90%. The number of images for each data set was limited in an effort to trap the structures with a specific oxidation state of the Cu, thus being the basis for six structures per crystal, each at a different dose. The structures corresponding to the LsAA9_A(f) data sets were solved according to Table 1. The collection strategy was identical between the data sets collected from substrate-free crystals, and crystals soaked in Cell₄. For both the LsAA9_A(f) structures with and without substrate, anomalous signal could be used to determine the absence/presence of Cl⁻ at the active site.

Table 1
The final data sets derived from data collected at increasing photon flux on LsAA9_A(f) crystals

The 7.88 × 10³ Gy data set consists of the first 45° of the collection at 10¹⁰ photons/s. The 5.99 × 10⁴ Gy data set consists of the last 45° of the collection at 10¹⁰ photons/s. The 1.39 × 10⁵ Gy data set consists of the first 45° of the 10¹¹ photons/s collection. The 3.60 × 10⁵ Gy data sets consists of images in the range 100–180° of the 10¹¹ photons/s collection. The 1.45 × 10⁶ Gy data sets consists of the first 45° of the 10¹² photons/s collection. Finally, the 6.65 × 10⁶ Gy data set is from the last 45° of the 10¹² photons/s collection.

Photon flux	10^10 photons/s		10^11 photons/s		10^12 photons/s
Data range used (°)	0–45	315–360	0–45	100–180	0–45	315–360
X-ray dose (Gy)	7.88 × 10^3	5.99 × 10^4	1.39 × 10^5	3.60 × 10^5	1.45 × 10^6	6.65 × 10^6

Data-collection parameters for the photoreduction of crystals of LsAA9_A(Ec) and TaAA9_A were chosen to approximately match the relative doses received by the LsAA9_A(f) crystals described above. Using RADDOSE3D (Zeldin et al., 2013), the average diffraction weighted dose was calculated prior to X-ray exposure to find the optimal flux required to achieve similarly reduced active-site Cu in the resulting structures and calculated again after the experiment for a final dose.

Data for the photoreduction study were also processed and scaled using XDS/XSCALE (Kabsch, 2010). The lowest dose data sets were used as references during the processing of higher dose data sets to ensure consistent indexing. The highest resolution cut-off was generally chosen based on a CC_1/2 in the outer shell of above 40%. The consequence of following the set scheme in Table 1 was that in some cases a com­pleteness of less than 100% and a data redundancy <2 were obtained, but this was considered a reasonable trade-off for consistency. Phases for LsAA9_A(f) were obtained from the previously published high-resolution LsAA9_A(f) structure (PDB entry 5ach) (Frandsen et al., 2016). For LsAA9_A(Ec), the space group was different than for LsAA9_A(f), and the phases were obtained using MOLREP (Vagin & Teplyakov, 1997) and PDB entry 5ach as the search model. Phases for LsAA9_A(Ec)-Cell₄ and Cell₃ were obtained from the isomorphous oligosaccharide-free structures. For TaAA9_A, the phases were obtained from the previously published high-resolution TaAA9_A structure with a single mol­ecule in the asymmetric unit (PDB entry 3zud) (Quinlan et al., 2011).

For all structures, rigid-body and restrained refinement were performed in REFMAC5 (Murshudov et al., 1997) of the CCP4 suite (Winn et al., 2011), alternated with manual rebuilding and validation performed in Coot (Emsley & Cowtan, 2004). Further validation was performed using BAVERAGE and PROCHECK (Vaguine et al., 1999) of the CCP4 suite (Winn et al., 2011). For data collection and refinement statistics, see Tables S1–S6 in the supporting information. Figures of the resulting models were prepared in PyMOL (The PyMOL Mol­ecular Graphics System; Version 2.0.4; Schrödinger, LLC). The active-site distances and angles of the final models were measured in Coot and using the Python library Biopython (Cock et al., 2009). Animations of active-site changes during photoreduction (see Movies S1–S5 in the supporting information) were made in PyMOL using the morph feature with no trajectory refinement.

2.3. Serial synchrotron crystallography

The LsAA9_A(f)-SSX data set was collected at the ID29 beamline (de Sanctis et al., 2012) of the European Synchrotron Radiation Facility (ESRF). The data were collected using the Mesh & Collect routine (Zander et al., 2015), in a similar manner to that described recently for AoAA13 (Muderspach et al., 2019). For each crystal identified by the routine, data were collected over a total rotation of 10°, though only 2.5° were used for processing, with an oscillation of 0.1° per image at a wavelength of 0.98 Å. 13 data sets were processed individually with identical parameters using XDS (Kabsch, 2010). The hierarchical cluster analysis software ccCluster (Santoni et al., 2017) was used to determine which data sets could be merged and scaled based on their correlation coefficient distances. The resulting data set from this merge was used to solve the LsAA9_A(f)-SSX structure. Phasing and refinement were performed as described above.

3. Results and discussion

3.1. Overall structure overview for this study

This study includes five LPMO and substrate combinations with similar photoreduction protocols at cryogenic tem­per­atures: LsAA9_A produced in two recombinant systems giving different post-translational modifications and crystallized under different conditions in different space groups, in the presence/absence of oligosaccharide ligands, and TaAA9_A (without the oligosaccharide ligand). Table 2 summarizes the post-translational modifications, crystallization solution com­positions, crystal form characteristics and exogenous ligands for the five combinations investigated. LsAA9_A and TaAA9_A display the immunoglobin G-like β-sandwich fold observed for all LPMOs structurally characterized so far [see Figs. 1(b) and 1(c)]. The catalytic Cu ion is coordinated by the His-brace, fitting an elongated octa­hedral geometry, and lies in the middle of a relatively flat surface (Frandsen & Lo Leggio, 2016; Vu & Ngo, 2018; Tandrup et al., 2018), where the polysaccharide substrate has been demonstrated experimentally to bind in several AA9 LPMOs (Frandsen et al., 2016; Tandrup et al., 2020; Courtade et al., 2016). In both proteins, a tyrosine –OH group functions as one of the axial ligands to the Cu. In oligosaccharide-free LsAA9_A(f), the copper coordination sphere includes two water ligands in equatorial (eq) and axial (ax) positions (see Fig. 1). The chemical nature of the exogeneous equatorial and axial ligands changes depending on the crystallization conditions and the exogenous axial ligand is displaced upon oligosaccharide binding.

Table 2
A brief overview of the proteins and crystals used in this study

Protein	PDB entry	Post-translational modification	Oligosaccharide ligand	Occupied binding subsites	Mother liquor com­position	Protein buffer	Space group	No. of molecules in asymmetric unit	Maximum resolution in this study (Å)	Non-H2O exogenous ligands
LsAA9_A LsAA9_A(f)	7pxi	His1 N^ɛ2 methyl­ation; N-glycosyl­ation on Asn33			3.3 M NaCl, 0.1 M citric acid pH 3.5	50 mM sodium acetate pH 5.5	P4132	1	1.30
	7pxj
	7pxk
	7pxl
	7pxm
	7pxn
LsAA9_A LsAA9_A(f)	7pyd	His1 N^ɛ2 methyl­ation; N-glycosyl­ation on Asn33	1.0 M cellotetra­ose	−2 − 1 + 1 + 2	3.3 M NaCl, 0.1 M citric acid pH 3.5	50 mM sodium acetate pH 5.5	P4132	1	1.90	Cl^−†
	7pye
	7pyf
	7pyg
	7pyh
	7pyi
LsAA9_A LsAA9_A(Ec)	7pyl 7pym 7pyn 7pyo 7pyp 7pyq	‡			2.3 M (NH4)2SO4, 0.1 M sodium acetate pH 4.5	25 mM Bis-Tris pH 5.8, 175 mM NaCl	P41	1	1.30	SO4§
LsAA9_A LsAA9_A(Ec)	7pyu	‡	0.5 M cellotriose	−1 + 1 + 2¶	2.3 M (NH4)2SO4, 0.1 M sodium acetate pH 4.5	25 mM Bis-Tris pH 5.8, 175 mM NaCl	P41	1	1.20	Cl^−†
	7pyw
	7pyx
	7pyy
	7pyz
	7pz0
TaAA9_A	7pz3	His1 N^ɛ2 methyl­ation; N-glycosyl­ation on Asn138			20 mM MgCl2, 0.1 M HEPES pH 7.5, 22%(w/v) polyacrylic acid 5100 sodium salt	50 mM sodium acetate pH 5.5	P21	1	1.40	Acrylic acid§
	7pz4
	7pz5
	7pz6
	7pz7
	7pz8

†Cl present from crystallization conditions is often found in substrate-bound structures and is expected to inhibit the position binding either O₂ or H₂O₂ (Frandsen et al., 2016).
‡LsAA9_A(Ec) is lacking the N-terminal His methyl­ation, as expected from the bacterial expression (Fig. S1)
§Not coordinating directly to Cu. Present from crystallization conditions.
¶Cellotriose is also bound in a pocket at β-sheet 4 (Fig. S2). Binding in this pocket is not believed to have any biological relevance.

The bacterially expressed LsAA9_A(Ec) has recently been shown to be fully active (Brander, Tokin et al., 2021) and was also used in a high-resolution study to determine the proton­ation stages in the second coordination sphere of Cu (Banerjee et al., 2022). The structure of LsAA9_A(Ec) is very similar to the previously published fungally expressed LsAA9_A(f) structures (all-atom r.m.s. deviation for resting-state substrate-free structures: overall = 0.56 Å and His-brace = 0.10 Å), except for post-translational modifications and differences in the exogenous ligands to the Cu due to different crystallization conditions (Table 2 and Fig. S1).

The crystal packing around the substrate-binding site differs between the LsAA9_A(Ec) and LsAA9_A(f) crystal forms. LsAA9_A(f) has an unrestricted binding site, so that LsAA9_A(f) crystals soaked in Cell₄ in this study bind from subsite −2 to +2, as reported previously (Tandrup et al., 2020). In contrast, the −2 binding subsite of LsAA9_A(Ec) is blocked by crystal contacts. This prevents binding in the same mode observed previously for LsAA9_A(f) for both Cell₃ and Cell₄ (Frandsen et al., 2016; Tandrup et al., 2020). In the structure of LsAA9_A(Ec)-Cell₄, the substrate is modelled from subsite −1 to +3, with the glucose monomer in the +3 subsite being partially disordered (see Figs. S2 and S3). Thus, we have chosen LsAA9_A(Ec)-Cell₃ as a model for the photoreduction study. In this structure, Cell₃ is bound at subsites −1 to +2. In line with previous structural studies (Frandsen et al., 2016), the binding of oligosaccharides expels the axial water mol­ecule, thus only the equatorial exogenous ligand remains. A Cl⁻ ion occupies this position where O₂ or H₂O₂ are expected to bind during the reaction. For LsAA9_A(f), this is likely due to the high NaCl concentration in the crystallization conditions (upwards of 3.0 M), while for LsAA9_A(Ec), the NaCl concentration is much lower. Anomalous signal, routinely used to detect heavy atoms in crystals, could only determine the presence of Cl⁻ unambiguously in LsAA9_A(f)-Cell₄ structures (not shown), but Cl⁻ still fits the observed electron density more accurately for LsAA9_A(Ec)-Cell₃, com­pared to water and SO₄²⁻.

The structure of TaAA9_A presented here is similar to those reported previously (PDB entries 3zud and 2yet) (Quinlan et al., 2011), except that the active-site copper is well defined and free of the disorder observed previously. TaAA9_A has been used previously in describing the LPMO active site in theoretical studies (Hedegård & Ryde, 2018; Kjaergaard et al., 2014; Kim et al., 2014) and the structures here could therefore present an improvement for further calculations. Crystal contacts most likely prevent binding at negative subsites, assuming the oligosaccharide binds in a similar manner to previous oligosaccharide-binding structures with LsAA9_A and an AA9 from Collariella virescens (CvAA9_A) (Frandsen et al., 2016; Simmons et al., 2017; Tandrup et al., 2020), and it has not been possible so far to obtain an oligosaccharide bound in the structure of this protein, which also has no reported activity on oligosaccharides.

Crystals of fungal and bacterially expressed LsAA9_A (unsoaked and soaked in Cell₄ and Cell₃ oligosaccharide, respectively) and crystals of TaAA9_A were exposed to X-rays for three consecutive data collections, with an increasing flux of photons, and from these, images were selected to form data sets for structure determination at individual X-ray doses. The images in each data set were limited to the minimal amount necessary to solve the structures while still obtaining a com­plete and statistically reas­onable data set (see Methods section for details of the data reduction and Tables S1–S6 for statistics). We solved two structures from each of the three collections for all the crystals, giving snapshots at six X-ray doses for each. Two additional low-dose structures for the study come from the data for LsAA9_A(f) at room tem­per­ature (RT) from an in-house diffractometer and synchrotron to resolution limits of 1.8 and 1.9 Å, respectively (see Fig. S4). As a reference for the Cu⁺ state, we used a chemically reduced structure obtained from LsAA9_A(f) crystals soaked in 10 mM ascorbic acid.

These reference structures were used to establish which parameters should be monitored to follow the transition from Cu²⁺ to Cu⁺. Previously, a tendency towards a slight increase in the Cu—His1-N distances and slight decreases in the Cu—His1 N^δ1 and Cu—His N^ɛ2 distances have been reported on photoreduction (Gudmundsson et al., 2014). However, as can be seen in Table 3, the differences are very small between our Cu²⁺ and Cu⁺ reference structures. The transition from Cu²⁺ to Cu⁺ can instead be followed more clearly by measuring the inter­atomic distances between Cu and its ligands (see Table 3), including the equatorial and axial water mol­ecules (see Fig. 1). In terms of distances, we here define the Cu state based on distances <2.2, 2.2–2.9 and >2.9 Å between Cu and the equatorial ligand as Cu²⁺, a mix and Cu⁺, respectively. For the axial Cu ligand, the distances are defined here as <2.7, 2.7–3.2 and >3.2 Å, respectively. While due to Jahn–Teller distortion (Jahn & Teller, 1937) even longer distances have been observed, at distances longer than these we can probably be confident that there is no longer a coordination bond. Compared to earlier studies on LPMOs, the lowest-dose Cu–water ligand distances found here are slightly longer than for other type II Cu sites investigated using X-ray doses in a similar range (Frankaer et al., 2014). For reference, in aqueous six-coordinated com­plexes of Cu²⁺, equatorial water mol­ecules are at a distance of around 2.0 Å from the ion and axial water mol­ecules are up to 0.3 Å longer (de Almeida et al., 2009). These distances have been used previously to examine the active-site geometry of LPMOs (Vu & Ngo, 2018), and the definitions of distance magnitudes are based on previous LPMO studies and small-mol­ecule Cu com­plexes (Gudmundsson et al., 2014; Vu & Ngo, 2018; Persson, 2010). Based on these cut-offs, our reference low-dose LsAA9_A(f) structure is very close to a full Cu²⁺ state. For the chemically reduced crystals, structures at two X-ray doses were obtained (a low dose of 2.08 × 10³ Gy and a high dose of 1.70 × 10⁷ Gy) (see Fig. S5). Consistent with a fully Cu⁺ state, the distances to the equatorial and axial water mol­ecules are large even at low dose (4.0 and 3.7 Å) and do not increase with dose (4.0 and 3.5 Å, respectively) (see Table 3 for all measured active-site distances).

Table 3
Structural parameters within the LPMO Cu site

The lowest and highest X-ray dose structures are listed, together with additional reference structures. Distances (between the indicated atom and Cu) and angles were measured in Coot (Emsley & Cowtan, 2004) and using the Biopython module (Cock et al., 2009). In cases where a ligand atom is modelled in a double conformation, the distance is an average weighted by the occupancy of the atom. Further structural parameters for all structures are presented in Tables S7 and S8. Res = high-resolution limit. Distances are measured from Cu to the indicated atom. θ₁, θ₂, θ₃ and θ_T (°) are defined in Vu & Ngo (2018). The lowest/highest/average distance and θ₁, θ₂, θ₃ and θ_T (°) for Cu²⁺/Cu⁺ structures are from Vu & Ngo (2018). ECR = estimated coordinate error based on the R value; value extracted from the REFMAC5 output PDB file.

	PDB entry	Dose (Gy)	Res. (Å)	Rwork/Rfree (%)	N^δ1 (Å)	NAm (Å)	N^ɛ2 (Å)	OTyr (Å)	Eq (Å)	Ax (Å)	θ1, θ2, θ3 (°)	θT (°)	ECR (Å)
LsAA9_A(f) (RT)	7pxr		1.80	14.67/16.62	1.9	2.2	2.0	2.8	2.2	2.8	88.6, 92.9, 176.6	3.3	0.081
LsAA9_A(f) (RT-sync)	7pxs	1.91 × 10^3	1.90	16.94/19.56	1.9	2.2	2.0	2.8	2.2	2.6	92.7, 94.7, 170.5	5.9	0.106
LsAA9_A(f) (Ascorbic acid)	7pxu	2.08 × 10^3	1.80	19.52/22.65	1.9	2.3	2.0	2.9	4.0	3.7	94.8, 96.3, 168.4	3.4	0.119
	7pxv	1.70 × 10^7	1.50	18.74/20.70	1.8	2.3	2.0	2.8	4.0	3.5	93.2, 97.4, 168.6	4.4	0.063
LsAA9_A(f)	7pxi	7.88 × 10^3	1.57	18.30/20.99	1.9	2.2	2.0	2.7	2.2	2.7	92.9, 89.6, 177.1	1.5	0.080
	7pxn	6.65 × 10^6	1.65	20.35/22.18	1.9	2.3	2.0	2.7	3.3	3.3	91.7, 99.7, 167.5	5.1	0.057
LsAA9_A(Ec)	7p­yl	1.49 × 10^4	1.70	14.83/18.83	2.0	2.1	2.0	2.7	1.9	2.6	91.8, 94.9, 172.6	3.3	0.094
	7pyq	6.35 × 10^6	1.60	14.42/16.78	2.0	2.2	2.0	2.8		3.8	93.3, 102.9, 162.0	7.8	0.069
LsAA9_A(f) (SSX)	7pxt	7.02 × 10^4	2.40	18.67/24.76	1.9	2.3	1.9	2.7	3.9	2.9	84.9, 99.6, 171.5	7.1	0.319
TaAA9_A	7pz3	5.37 × 10^3	1.90	20.81/25.63	1.9	2.2	2.1	3.0	2.2	2.0	94.0, 91.4, 174.6	−1	0.207
	7pz8	3.12 × 10^6	1.40	15.12/16.89	1.9	2.2	2.0	2.9	2.5	3.0	94.6, 96.1, 169.3	1	0.055
LsAA9_A(f)-Cell4	7pyd	7.88 × 10^3	2.21	22.47/28.71	2.0	2.3	2.0	2.5	2.3		96.4, 94.1, 161.1	15.6	0.246
	7pyi	6.65 × 10^6	2.05	21.80/26.00	1.9	2.5	2.0	2.7	3.8		98.1, 110.1, 147.9	14.5	0.184
LsAA9_A (Ec)-Cell3	7pyu	1.49 × 10^4	1.40	14.19/16.22	2.0	2.1	2.0	2.6	2.3		91.3, 91.7, 169.2	10.4	0.052
	7pz0	9.81 × 10^6	1.20	13.40/14.75	1.9	2.3	1.9	2.7	3.8		93.0, 98.8, 165.7	8.2	0.030
LsAA9_A(Ec)-Cell4	7pxw	2.14 × 10^6	1.40	11.71/15.98	2.0	2.3	2.0	2.7	3.9		92.4, 99.5, 165.3	8.5	0.047
Cu^2+
Lowest					2.1	1.9	1.9				88, 85, 168	−2.4
Highest					2.4	2.3	2.4				103, 96, 178	4.2
Average					2.2	2.0	2.1				93, 92, 174	0.75
Cu^+
Lowest					2.0	1.9	1.9				93, 93, 155	−14
Highest					2.3	2.1	2.3				99, 103, 169	16
Average					2.2	2.0	2.0				96, 99, 163	3.9

Upon reduction of the metal from Cu²⁺ to Cu⁺, the geometry is expected to change from a trigonal bipyramid, distorted square pyramid, seesaw or elongated octa­hedral geometry (the latter applicable to the AA9 LPMOs studied here) towards a T-shaped geometry, with increased deviations from 90° in the defined angles θ₁ and θ₂, a decrease in θ₃ and an increase in θ_T [see Fig. 1(d)] (Vu & Ngo, 2018). The θ₁, θ₂, θ₃ and θ_T values in the chosen reference structures are consistent with the previous findings for the geometrical differences between Cu²⁺ and Cu⁺ geometries, and a steady increase in θ₂ and a decrease in θ₃ are the clearest trends which are common to the three types of crystals investigated. We have further observed a decrease in θ_H–H and θ_H1, with an increase in θ_HN in the fully chemically reduced structures com­pared to the reference Cu²⁺ structure (see Fig. S6). Most angles do not change upon continued exposure of the chemically reduced LsAA9_A(f), except θ_H1 and θ_H–H, which decrease further (see Table S7). Thus, all these angles have also been monitored through the photoreduction studies as useful indicators of the Cu ion state (Tables S7 and S8).

3.2. Photoreduction of substrate-free AA9s

The lowest dose LsAA9_A(f) cryotem­per­ature structure was solved from a data set totaling a dose of 7.88 × 10³ Gy [Figs. 1 and 2(a)]. Detailed transitions (all X-ray doses) can be seen in Figs. S7 and S8, and Movies S1 and S2, while different geometrical parameters under photoreduction are monitored in Figs. S6 and S9. Efforts were made to expose other crystals to com­parable doses. For LsAA9_A(Ec), the lowest dose was 1.49 × 10⁴ Gy [Fig. 2(c)]. In both cases, the active-site distances and angles agree well with a Cu²⁺ oxidation state of type II Cu (Vu & Ngo, 2018; Solomon et al., 2014; Frandsen et al., 2016) [see Figs. 2(a) and 2(c)], although θ₃ and θ_T may indicate that photoreduction has started for LsAA9_A(Ec), which is collected at a slightly higher dose. In all the low-dose structures, Cu has been modelled in full occupancy. The distance between Cu and the equatorial water is 0.2 Å shorter in LsAA9_A(Ec) and is probably affected by a nearby sulfate ion at a hydrogen-bond distance of 2.5 Å. The key measured geometrical parameters are presented in Table 3, with an additional analysis in Tables S7 and S8. In the lowest-dose substrate-free structures of fungal/bacterial LsAA9_A, the distance to the equatorial water ligand is in the range 2.2–2.3 Å, while that to the axial ligand is in the range 2.6–2.7 Å. For com­parison, QM/MM models have shown distances of 2.16/2.25, 2.08/2.53 and 2.21/2.34 Å for the equatorial/axial water mol­ecule, depending on the parameters of the calculation (Theibich et al., 2021), which could indicate that even at the lowest X-ray dose possible, a small amount of photoreduction has occurred.

Figure 2 LPMO Cu sites in LsAA9_A(f), LsAA9_A(Ec) and TaAA9_A at the lowest and highest X-ray doses. LsAA9_A(f) at (a) low and (b) high dose. LsAA9_A (Ec) at (c) low and (d) high dose. TaAA9_A at (e) low and (f) high dose. The distances between Cu (orange spheres) and the equatorial/axial water mol­ecules (red spheres) increase with increasing X-ray dose in the transition from Cu^2+ to Cu^+. In some cases, the water mol­ecules have been modelled in a double conformation (b) or left unmodelled due to a lack of electron density (d). Distances below 2.2/2.7 Å are shown as full lines (coordination distance), below 2.9/3.2 Å as dashed lines (close to coordination distance) and above 2.9/3.2 Å with no line (not coordinating) for the equatorial/axial ligands, respectively. 2Fo–Fc electron density is shown at a 1.0 contour level as blue mesh.

As the X-ray dose increases, so do the distances to the exogenous water ligands, which gradually become disordered and/or disappear, as expected (see Fig. 3). For all inter­mediate doses of LsAA9_A(f), the equatorial water mol­ecule has been modelled in a double conformation, each with 50% occupancy. For LsAA9_A(Ec), the equatorial water mol­ecule becomes disordered at an X-ray dose above 3.33 × 10⁵ Gy, where the mol­ecule can no longer be modelled confidently. At inter­mediate doses of LsAA9_A(Ec), the axial water is modelled in a double conformation, each with 50% occupancy, or in an alternative conformation with sulfate (see Fig. S8), likely hindering the migration of the water.

Figure 3 Measured distances as a function of the average diffraction weighted radiation dose. Distances are measured from the Cu atom to either the equatorial ligand or the axial water mol­ecule. In substrate-binding structures, no measurement is taken for the axial water, as it has been displaced by substrate. For LsAA9_A(Ec), the density for the equatorial water disappears com­pletely after the third dose structure. All distances are listed in Table S7.

Both LsAA9_A(f) and LsAA9_A(Ec) structures with the highest X-ray doses (6.65 × 10⁶ and 6.35 × 10⁶ Gy, respectively) exhibit exogenous ligand distances and other parameters agreeing with a predominant Cu⁺ state, as reported previously (Vu & Ngo, 2018; Gudmundsson et al., 2014; Frandsen et al., 2016) and in agreement with our chemically reduced LsAA9_A(f) structures. In LsAA9_A(f), both the axial and the equatorial water mol­ecules have increased distances from Cu com­pared to the lowest dose (0.6 Å more for the axial and 1.2 Å more for the equatorial) and >3.3 Å from Cu. Thus, the Cu atom has effectively lost its two ligands when reduced from Cu²⁺ to Cu⁺ [see Table 3 and Figs. 2(b) and S7]. In the higher-dose structures, the equatorial water mol­ecule has migrated away from the Cu ion and is coordinated by Gln162 and/or His147 of the secondary coordination sphere instead (Fig. S10). In LsAA9_A(Ec), there is no residual density for the equatorial water at the highest dose, while the axial water is 0.5 Å further away than in the lowest dose. Compared to the chemically reduced crystals, the measured Cu-site distances are shorter even in the highest dose LsAA9_A(f) structure. This might imply that the LsAA9_A(f) structure is not fully reduced and that it may indeed be difficult to fully photoreduce an LPMO. Alternatively, photoreduction at 100 K may impede full migration of bound water.

Looking at the θ angles in the two substrate-free LsAA9_A variants (Figs. 4 and S6), Cu reduction is associated with an increase in θ₂, θ_T and θ_HN (>8, >3 and >3°, respectively). Additionally, a decrease in θ₃, θ_H–H and θ_H1 (>9, >3 and >9°, respectively) can be observed over the transition, although the absolute values between the two enzymes differ somewhat.

Figure 4 Measured active-site angles θ1, θ2 and θ3 as a function of the average diffraction weighted radiation dose. A consistent observation for all proteins in the photoreduction study is an increase in θ2 and a decrease in θ3, while θ1 is less consistent. All angles are listed in Table S7.

In a previous EXAFS study, it was indicated that the Cu²⁺ structure of TaAA9_A could be well represented by four N/O ligands at an average distance from Cu of 1.98 Å (Kjaergaard et al., 2014), whereby one of the close ligands was lost on photoreduction, leaving the Cu atom in a T-shaped coordination. The equatorial ligand distances in the lowest-dose TaAA9_A structure presented have an average of 2.1 Å, which is within the agreement level that can be expected given an estimated coordinate error of 0.207 Å for this structure (see Table 3); however, in the crystal structure, an axial water is additionally very close at a distance from Cu of 2.0 Å and is significantly closer than in the LsAA9_A substrate-free structures above. The distance increases steadily with dose to 3.0 Å [see Figs. 2(e) and 2(f), with a more detailed transition in Fig. S11 and Movie S3]. The equatorial water mol­ecule sits at 2.2 Å from the Cu atom in the lowest-dose structure and reaches the longest distance of 2.6 Å at an inter­mediate dose (1.13 × 10⁶ Gy). Similar to LsAA9_A(Ec), the structure of TaAA9_A has a small mol­ecule (modelled as acrylic acid) near the equatorial position [see Figs. 2(e) and 2(f)], and additionally, a HEPES mol­ecule near the axial position, present from crystallization conditions. It is therefore possible that the presence of these ligands affects the observed distances. However, the axial distance in the EXAFS study may also be affected by some photoreduction even when precautions were taken. TaAA9_A may also be slightly less sensitive to photoreduction than LsAA9_A, as the changes in θ values are slightly smaller for the TaAA9_A transition (Figs. 4 and S6).

For all three substrate-free enzymes used in the photoreduction study, we have obtained a reasonably consistent picture of transition from Cu²⁺ to Cu⁺ in diverse AA9 LPMOs under different crystallization conditions. We observe a loss in coordination between Cu and water ligands at higher doses, changing the coordination from an elongated hexa­coordinated geometry to a T-shaped geometry with Cu coordinated solely by protein ligands, as discussed previously (Gudmundsson et al., 2014; Vu & Ngo, 2018; Kjaergaard et al., 2014). Previous studies of EfaCBM33/EfAA10, using doses in a com­parable range to what has been used here (Gudmundsson et al., 2014), showed that at doses higher than 5.94 × 10⁵ Gy the exogenous water ligands were expelled from the initial trigonal bipyramidal geometry. This points towards the T-shaped geometry being essential when accepting the oxygen co-substrate, despite the initial difference in the water-ligand coordination com­pared to the investigated AA9s. A structure of AoAA13, which has similar Cu geometry to AA9, solved with an X-ray dose of 3.19 × 10⁴ Gy, exhibits a fully reduced Cu site with no visible solvent-facing ligands, and already at the lowest dose (2.57 × 10³ Gy) has lost the axial water ligand (Muderspach et al., 2019), suggesting it is more prone to photoreduction than the analyzed AA9s and AA10.

3.3. Serial synchrotron crystallography and room-tem­per­ature structures of LsAA9_A

Serial synchrotron crystallography (SSX) experiments are an effective method for reducing radiation damage com­pared to a conventional MX experiment (de la Mora et al., 2020; Ebrahim et al., 2019; Mehrabi et al., 2021) and could allow better control of photoreduction in metalloproteins. Data collected on 13 crystals were scaled together, resulting in the 2.4 Å LsAA9_A(f)-SSX data set. The dose received by the resulting structure (7.02 × 10⁴ Gy) is com­parable to the oligosaccharide-free LsAA9_A(f) structure with a dose of 5.99 × 10⁴ Gy, and indeed shows a similar geometry, indicating close to full photoreduction, though with a somewhat sharper θ₁ and wider θ₃ angle. The θ_T angle in the SSX structure is greater than in any of the six LsAA9_A(f) substrate-free structures, or the chemically reduced structures. The axial water mol­ecule was found at 2.9 Å in the LsAA9_A(f)-SSX structure, which agrees with the substrate-free LsAA9_A(f) structure. However, the equatorial water mol­ecule found at 3.9 Å is very indicative of a fully reduced Cu site, despite efforts to limit the dose experienced by each crystal. In part, this could be due to the water in the LsAA9_A(f)-SSX structure having only a single conformation at this position (see Figs. S7 and S12). Perhaps unsurprisingly, SSX does not protect from photoreduction at a similar total dose and it would have been difficult to obtain a lower dose with the applied protocol. The distances observed within the active site are listed in Table 3, and in Tables S7 and S8.

We collected room-tem­per­ature data for the LsAA9_A(f) crystals, both from an in-house diffractometer and from a syn­chro­tron, at the lowest possible dose (Fig. S4). Here LsAA9_A(f) (RT) from the in-house data has distances of 2.2 and 2.8 Å for the equatorial and axial ligands, respectively, which correlate well with a Cu²⁺ state (Vu & Ngo, 2018; Gudmundsson et al., 2014; Frandsen et al., 2016). Perhaps surprisingly, data collected from a synchrotron source at room tem­per­ature was similarly able to produce a structure containing a Cu²⁺ site [LsAA9_A(f), RT-sync], based on the Cu–water distances and θ₁–θ₃ angles (some of the other parameters resemble more a Cu⁺ site). It was in this case possible to achieve a much lower dose than for the corresponding cryogenic tem­per­ature structures (1.91 × 10³ Gy).

Due to the rapid decay of the diffraction intensity without cryocooling, a fully photoreduced structure could not be determined using this strategy. Curiously, the LsAA9_A(f) (RT) structure contains a Cl⁻ ion in the axial position, which was confirmed by an anomalous signal (see Fig. S4), which in LsAA9_A(f) (RT-sync) appears to be the more regularly observed water ligand.

3.4. Photoreduction of LsAA9_A crystals soaked with oligosaccharides

Data on LsAA9_A crystals soaked in cello-oligosaccharides were collected using the same protocol as for the oligosaccharide-free crystals. As for the unbound structures, methyl­ation of His1 does not seem to affect photoreduction to any significant extent, nor does it seem to have a structural effect on substrate binding (Frandsen et al., 2016). For both the LsAA9_A(f) and LsAA9_A(Ec) oligosaccharide-bound/unbound structures, the so-called pocket water increases its distance to His1-N with increasing X-ray dose (Figs. 2, 5 S7, S9, S10, S13 and S14). Here also the distance from the equatorial ligand to Cu increases with increasing X-ray dose experienced by the crystal (Tables 3 and S7, and Fig. 5).

Figure 5 LPMO Cu sites in the oligosaccharide-bound structures of LsAA9_A(f) [(a) 7pyd and (b) 7pyi] and LsAA9_A(Ec) [(c) 7pyu and (d) 7pz0] at the lowest and highest X-ray doses. For both LsAA9_A(f) and LsAA9_A(Ec) at higher X-ray doses, the equatorial Cl^− ion (green spheres, present from crystallization conditions) increases its distance relative to the Cu atom. The Tyr164-O to Cu distance is, in all cases, shorter than for the substrate-free structures presented in Fig. 2. Active-site distances are listed in Table 3. 2Fo–Fc electron density is shown at a 1.0 contour level as blue mesh.

At the lowest doses [Figs. 5(a) and 5(c)], the equatorial solvent-facing ligand, modelled as a Cl⁻ ion, is 2.3 Å from the Cu atom, while in the highest-dose structures, the distance has increased to 3.9 and 3.8 Å for LsAA9_A(f)-Cell₄ and LsAA9_A(Ec)-Cell₃, respectively. For com­parison, the average distance to the equatorial ligand is 2.1 Å between different QM/MM models in a recent theoretical study (Theibich et al., 2021). In several inter­mediate X-ray doses of LsAA9_A(f)-Cell₄, the Cl⁻ ion was modelled in a double conformation, with 90% occupancy for the conformation furthest from the Cu atom and 10% occupancy for the conformation closest to the Cu atom, indicating a mixed state between Cu²⁺ and Cu⁺. The distance from the Cl⁻ ion to Cu increases much more rapidly with dose in the LsAA9_A(f)-Cell₄ structures (3.8 Å at 3.60 × 10⁵ Gy) com­pared to LsAA9_A(Ec)-Cell₃ (3.8 Å at 9.81 × 10⁶ Gy) (Figs. S13 and S14, Table S7, and Movies S4 and S5). It is unclear if these differences are due to the differences between the fungally and bacterially expressed enzymes, or if it is a result of the conditions under which the crystals were grown. At any rate, while in the case of the substrate-free LsAA9_A(Ec) structure the equatorial position was partially occupied by sulfate (and partially by a water mol­ecule), this does not seem to be the case for the equatorial ligand at LsAA9_A(Ec)-Cell₃, which is most consistent with fully occupied chloride, although this could not be confirmed with an anomalous signal.

As for the transition from Cu²⁺ to Cu⁺ in the unbound structures, for the oligosaccharide-bound structures there is a clear increase in the θ₂ angle and a clear decrease in the θ₃ angle (see Fig. 4). Furthermore, all the oligosaccharide-bound structures at any dose have values considerably greater than 0 and greater than those of the corresponding unbound structures for the θ_T angle, i.e. 18.7 ± 3.3° for LsAA9_A(f) and 8.7 ± 0.9° for LsAA9_A(Ec) (see Fig. S6). Thus, a significantly greater θ_T angle seems to be a characteristic of oligosaccharide-bound LsAA9_A. To further test this hypothesis, we calculated the average value for all the available structures of LsAA9_A with bound saccharide ligands, excluding xylooligosaccharides, as they are questionable as true LPMO substrates (Simmons et al., 2017). The average was 11.49 ± 2.03° (PDB entries 5aci, 5acf, 5acj, 5n05, 5nkw, 5nlr, 5nls and 6ydg). As θ_T also increases with dose in the unbound structures, a better com­parison may be between low-dose structures for all saccharide-free and saccharide-bound LsAA9_A structures to date (Table S9), which also shows a greater θ_T value for bound (12.0 ± 3.1°) com­pared to free (2.4 ± 0.9°).

Several reports have suggested a higher affinity of at least some AA9 LPMOs for cellulosic substrates in their Cu⁺ as opposed to their Cu²⁺ form (Kracher et al., 2018; Bertini et al., 2018; Courtade et al., 2016; McEvoy et al., 2021). This has also been demonstrated recently for LsAA9_A (Brander, Tokin et al., 2021). For both LsAA9_A(f) and LsAA9_A(Ec) oligosaccharide-binding structures, there are no considerable structural changes in inter­actions (e.g. protein-accessible surface or hydrogen bonding) in the Cu²⁺ to Cu⁺ transition that can explain the increased affinity. However, since the axial ligand is already lost during reduction, no penalty must be paid for the loss of this inter­action in the Cu⁺ form. Furthermore, we have shown that both photoreduction and, to an even larger extent, oligosaccharide binding increases the θ_T angle. Thus, the oligosaccharide-free Cu⁺ state is closer in active-site geometry to the oligosaccharide-bound state, which might contribute favourably to binding.

3.5. The active-site tyrosine in oligosaccharide-free and oligosaccharide-bound structures

The role of the Cu-site tyrosine has been proposed to be for protection against auto-oxidation (Paradisi et al., 2019; Singh et al., 2019). Recently, deprotonation of tyrosine as part of the mechanism was investigated through theoretical calculations and spectroscopic measurements to elucidate the formation of the LPMO reaction inter­mediates cis/trans-[Tyr-O^.–Cu²⁺–OH]⁺ (McEvoy et al., 2021). In the unbound state, the formation of Tyr inter­mediates is associated with a reduction in the Tyr-O to Cu distance of 0.27–0.47 Å (McEvoy et al., 2021). Thus, though a shortening has been reported from a com­parison of experimental structures (Frandsen et al., 2016), it was important to confirm it in this more systematic study.

Here, for both LsAA9_A variants, the distance between Tyr164 and Cu increases by 0.1 Å with increasing radiation dose, while for TaAA9_A, the distance decreases by the same value. These differences may not be significant. However, confirming the previous crystallographic studies (Frandsen et al., 2016; Simmons et al., 2017), the distance from Tyr164-O to Cu is shorter (0.1–0.2 Å) for all the LsAA9_A(f)-Cell₄ and LsAA9_A(Ec)-Cell₃ structures com­pared to the corresponding unbound structures (see Fig. 6). This difference may again not seem significant when considering only DPI-based coordinate errors (Kumar et al., 2015; Murshudov et al., 1997). However, there are a number of considerations suggesting that it is significant, at least for the low-dose LsAA9_A Cu²⁺ structures from this and previous studies (Frandsen et al., 2016), which are listed in Table S9, with the relevant distances and errors presented in Movie S6. First of all, taking each structure as an independent estimate of bond length, the average for the three saccharide-free structures is 2.71 ± 0.031 Å, while for the three saccharide-binding structures, the average is 2.50 ± 0.044 Å, indicating that the real coordinate error may be lower than estimated. Secondly, to investigate the structural changes without influence of the refinement protocol and the presence of the saccharide in the model, we calculated difference maps where only rigid-body refinement was applied to a model of a sac­charide-free structure (protein and Cu only, as a single rigid body to maintain a fixed distance) against data for a sac­charide-binding structure. Maps calculated for three such pairs show clearly that the data demand the copper–Tyr distance to be shortened (Fig. S15). The reduction in distance between Tyr-O and Cu is also reproduced qualitatively in the QM/MM-optimized structures of LsAA9_A, where the distance was found to decrease by 0.2–0.3 Å upon LsAA9_A binding the substrate (Theibich et al., 2021). While the role of tyrosine remains unclear, the shortening of the Tyr-O to Cu bond has been shown here to be reproducible for LsAA9_A.

Figure 6 The measured Cu—Tyr-O distance over the Cu^2+ to Cu^+ transition for LsAA9_A(f) and LsAA9_A(Ec). A slight increase in the distance is found over the transition and a more significant reduction in the distances for all doses is found once substrate is bound. This is further presented for the low-dose structures in Movie S6.

4. Conclusion

For metalloproteins, radiation damage in the form of photoreduction is common, and balancing X-ray dose and diffraction intensity is crucial. For LPMOs, photoreduction of the central Cu atom is difficult to avoid in a diffraction experiment, but may reveal details for the first step of the LPMO reaction.

We recommend that researchers wishing to investigate the LPMO Cu²⁺ state plan their experiments to limit the total dose to a few kGy or less if possible and, if necessary, to use helical data collections to reduce the dose. A dose in the MGy range should be used to catch Cu⁺ states, or possibly trigger the LPMO reaction. Similarly, we were successful in collecting data on chemically reduced crystals which support a Cu⁺ state.

Through our photoreduction study we have monitored various geometrical parameters, which are indicative of the two LPMO Cu states. As the observed changes in distance to protein ligands are small, the best structural diagnostics for the Cu²⁺ to Cu⁺ transition in these (and probably other octa­hedrally coordinated Cu sites) are the θ₃ and θ_T values, where values lower than 170° and higher than 3°, respectively, are strong indicators of a reduced site. Additional indicators may be found in the angles θ_H–H and θ_H1, which appear lower for a reduced site, and θ_HN, which is slightly higher for the Cu⁺ state, although the absolute values are inconsistent between the investigated proteins. Non-protein ligands may also be helpful in identifying the Cu state, as we and others have demonstrated, and the final T-shaped geometry of the His-brace seems to be a defining feature of LPMO Cu⁺ sites. The exogenous ligands can be affected by the closeness of other mol­ecules present in the active site, which must be considered when inter­preting structural data. Thus, when a single-crystal X-ray experiment is performed, these values alone cannot firmly assign a structure as Cu²⁺ or Cu⁺.

Upon binding of saccharide substrate, the Cu–Tyr distance in AA9s is found consistently to be shortened, regardless of the Cu state. Based on several independently determined structures and an analysis of difference electron-density maps, we find this small shortening of ∼0.2 Å significant for the Cu²⁺ structures of LsAA9_A. With previous indications that the active-site Tyr is important for the mechanism, the experimental structures now available at both Cu stages could allow further com­putational exploration. Correlated to the Cu–Tyr distance changes, while in Cu²⁺ saccharide-free structures the equatorial ligands and Cu are close to being coplanar, saccharide binding induces clear deviation from planarity, as seen in changes of the θ_T angle. As the reduction of Cu also induces geometric changes to the Cu site, including an increase in the θ_T angle, the observed greater affinity of Cu⁺ LPMOs for poly- and oligosaccharides may be partly explained by being closer structurally to the oligosaccharide-bound form.

5. Abbreviations

AA: auxiliary activity; ASC: ascorbic acid; ax: Axial; CAZy: Carbohydrate-Active enZymes database; Cell₃: cellotriose; Cell₄: cello­tetra­ose; EndoH: endoglycosidase-H; Eq: equatorial; LPMO: lytic polysaccharide monooxygenase; LsAA9_A(Ec): LsAA9_A produced in E. coli; LsAA9_A(f): LsAA9_A produced in A. oryzae; LsAA9_A: family AA9 LPMOs from Lentinus similis; MX: macromolecular crystallography; QM/MM: quantum mechanics/mol­ecular mechanics; RT: room tem­per­ature; SSX: serial synchrotron crystallography; TaAA9_A: family AA9 LPMOs from Thermoascus auranti­acus.