The gene sequence encoding dye-decolorizing peroxidase from the slime mold D. discoideum AX4 (GenBank: EAL70759.1) was codon optimized for Escherichia coli, synthesized and subcloned into the BamHI/XhoI cloning site in a pGEX-6P1 vector harboring a Human Rhinovirus 3C excision site and a glutathione transferase (GST) tag at the N-terminal region (BioCat GmbH, Germany). The pGEX-6P1-DdDyP construct was transformed into an OverExpress E. coli C43(DE3), a chemically competent strain (Sigma-Aldrich, Germany). For purification, a single colony from a freshly streaked plate was inoculated into a Luria-Bertani (LB) or ×2 yeast extract-tryptone (×2 YT) enriched media containing 100 μg mL⁻¹ final concentration of ampicillin and incubated at 37°C ± 1.0°C for overnight (15–16 h). A starter culture was used to inoculate 6 × 1 L of LB or ×2 YT containing 100 μg mL⁻¹ ampicillin and incubated at 37°C ± 1.0°C or lower temperatures until the optical density (OD₆₀₀) reaches 1.0–1.25 before inducing the expression of DdDyP with 1.0 mM of isopropyl β-D-1-thiogalactopyranoside. The culture was then incubated for additional 7 or 20 h for expression at 37 ± 1.0 and 20°C ± 1.0°C, respectively. The cells were harvested by centrifugation with ×13,881 g for 30 min on an F9-6-×1000 LEX rotor (Thermo Fischer Scientific, Germany) at 4°C and pellets were stored at −80°C until used.

For protein purification, frozen cells were thawed using warm tab water (∼40°C) and diluted with ×3–5 of lysis buffer containing 0.05 M Tris-HCl, pH 8.0 and 0.15 M NaCl supplemented with 1.0 mM final concentration of phenylmethylsulfonyl fluoride protease inhibitor or a tablet of EDTA-free protease inhibitor cocktail (Sigma-Aldrich, Germany). The cells were lysed with 35 cycles of sonication at 4°C on ice using 50% amplitude and 25 s sonication pulse with 1.5 min interval. Lysate was clarified with centrifugation at ×52,400 g at 4°C for 45 min and the supernatant was filtered with a 0.45 µm syringe filter and mixed with 5–10 mL glutathione sepharose high-performance resin pre-equilibrated with lysis buffer, followed by incubation at 4°C with gentle rotation for 3 h. The mixture was loaded into an empty gravity column and the GST-tagged DdDyP was eluted with ×5 column volume of an elution buffer containing 0.05 M Tris-HCl, pH 8.0, 0.15 M NaCl and 15–20 mM L-Glutathione (reduced form). The GST tag was then removed using HRV 3C protease with 1:20 enzyme to protein ratio at 4°C for overnight followed by passing the mixture through a glutathione sepharose column pre-equilibrated with lysis buffer. Purified protein was characterized with SDS-PAGE and UV-visible spectrophotometry. For crystallization the protein was further purified with gel-filtration using Superdex 75 10/300 Increase column (Cytiva, Sweden). Purified protein was concentrated to 20–30 mg mL⁻¹ in lysis buffer and stored at −80°C until further use. Overexpression, purification and crystallization were carried out at the XBI BioLab of the European XFEL facility (Han et al., 2021).

A control heme reconstitution experiment was conducted as described previously (Chen et al., 2015). In brief, purified apo-DdDyP from LB expression was mixed with hemin chloride with ∼1:2 M ratio in a buffer containing 50 mM Tris-HCl, pH 7.0 and 150 mM NaCl, followed by incubation on ice for 30 min. The heme reconstituted holo-DdDyP protein was then passed through a PD-10 desalting column (Cytiva, Sweden) to remove the excess hemin chloride.

All spectra were recorded on a Shimadzu UV-2700 PC spectrophotometer (Shimadzu Co., Japan) using a cuvette with 1.0 cm pathlength in a range of 200–700 nm at room temperature (20°C ± 2.0°C). For measurements purified DdDyP was diluted with lysis buffer to a concentration of 0.4 mg mL⁻¹ and the lysis buffer was used as a reference. All spectra were processed using the Origin software 2022b (OriginLab Co., United States).

Crystallization screening was performed using a NT8 Formulatrix robot (Formulatrix, United States). Hit was obtained from the C12 condition (20% PEG 6000, 0.1 M HEPES, pH 7.0, 0.01 M ZnCl₂) of the PACT++ crystallization screen (Jena Bioscience, Germany) with 10 mg mL⁻¹ of purified DdDyP. This condition was further optimized to 15% PEG 6000, 0.1 M HEPES-NaOH, pH 7.0, and 0.01 M ZnCl₂ crystallized with 15 mg mL⁻¹ final concentration of purified DdDyP, which gave rise to a maximum crystal size of 200 μm × 100 μm × 25 μm at 20°C ± 1.0°C in 4–6 weeks. Crystals were harvested directly from the drops using nylon loops and flash-cooled in liquid nitrogen.

X-ray diffraction datasets were collected at the P11/PETRA III beamline at DESY (Hamburg, Germany) using a flat focus with 20 × 20 μm² (v × h) beam area at the sample position, 12.0 keV photon energy, and a photon flux of ∼2 × 10¹⁰ photon sec⁻¹ and an exposure time of 100 ms for a total wedge of 360° with 0.1° oscillation recording step on EIGER 16M detector (Burkhardt et al., 2016). Data collection was performed at cryogenic temperature, 100 K. Diffraction datasets were processed using the program XDS, and scaled with XSCALE in the XDS graphic user interface (Kabsch, 2010). The initial phase was obtained by molecular replacement using the DdDyP peroxidase active structure (PDB ID: 7ODZ) as a reference model with the program Phaser in the phenix software (Afonine et al., 2012). The model was then refined in phenix and manually corrected in coot (Emsley et al., 2010). Radiation dose was estimated using the program RADDOSE-3D as described previously using the abovementioned parameters (Bury et al., 2018). For channel and cavity calculations MOLEontile tool (https://mole.uplo.cz/method) was used with the coordinate obtained from the final cycle of refinement (PDB ID: 8OHY) as a template (Sehnal et al., 2013). The interfaces of heme b and the oligomeric states analysis of DdDyP were calculated using PISA (Protein Interfaces, Surfaces, and Assemblies) server (Krissinel and Henrick, 2007).

For heterologous expression, a DdDyP peroxidase gene was cloned into a pGEX-6P1 vector which has an HVR 3C excision site and a GST-tag in its N-terminal region as reported previously (Rai et al., 2021), however, we used the OverExpress E. coli C43(DE3) strain instead of BL21-derived Rosetta (DE3) strain for overexpression and purification of DdDyP. Note that both strains were derived from BL21(DE3). DdDyP was expressed at high and low temperature (37 ± 1.0 and 20°C ± 1.0°C) in LB and ×2 YT medium with different yields, ranging from 2.1 to 14.1 mg of protein in average per 25 g of cells, respectively (Table 1). Figures 2A, B shows the SDS-PAGE analyses of the typical expression and purification of DdDyP in E. coli C43(DE3) strain. Protein expressed at 37°C ± 1.0°C, however, has transparent to pale brownish colour, whereas those of 20°C ± 1.0°C exhibited a darker brownish colour (Figure 2C). The proteins purified from high and low temperature have a reasonable purity with less aggregation (Figure 2) with ∼35 kDa molecular weight as confirmed by the SDS-PAGE analysis. However, no crystallization hit was obtained from these conditions despite several attempts with various crystallization screens (Table 1). Since that our purified DdDyP has sufficient purity for crystallization, yet we did not successfully crystallize it, we concluded that the instability of the protein during expression at 37°C ± 1.0°C may be the cause for the unsuccessful crystallization. This is likely due to the improper protein folding, and thus lowering the temperature of the expression may be one method for achieving a stable and correctly folded protein (Francis and Page, 2010; Huang et al., 2021). Indeed, when we expressed DdDyP at 20°C ± 1.0°C, it gave rise to a darker brownish protein (Figure 2C), a typical heme peroxidase colour of a native protein, especially when expressed in a ×2 YT enriched medium, which is richer than LB. Intriguingly, DdDyP expressed in ×2 YT at lower temperature was the only condition that resulted in a successful crystallization yielding dark brownish crystals, an indication that heme b is preserved during purification and crystallization (Figure 2D). The ×2 YT also yielded several times higher amount of protein than that obtained in LB (Table 1).

To examine the quality of the electronic absorbance of purified DdDyP we used a UV-visible spectrophotometer. Figures 3A–D shows the spectral analysis of the purified DdDyP protein from different conditions. DdDyP shows weak absorbance Soret peak at λ = 405 nm and a Reinheitszahl (R _Z) (A _soret/A ₂₈₀) value of ∼0.27 when expressed in LB at 37°C ± 1.0°C (Table 1). This R _Z value is about 2 times higher than that obtained previously on peroxidases that were expressed using BL21(DE3) strain (Rai et al., 2014; Fiege et al., 2018). As shown in Table 1, we observed that the R _Z value increases ×2–3 times to reach ∼0.93 with λ = 402 nm of the Soret peak when expressing DdDyP in enriched ×2 YT medium at low temperature (20°C ± 1.0°C), indicating a heme occupancy of 90%–95% which is comparable to that of the secreted heme peroxidase (DdPoxA) (Nicolussi et al., 2018). This value is comparable to our heme reconstitution reference (Figure 3D) and significantly higher (about 7 times) than those previously reported, when DyP-peroxidases were expressed without adding heme supplements or δ-ALA during expression (Krainer et al., 2015; Fiege et al., 2018). Heme biosynthesis in E. coli relies on an L-glutamate, glycine, succinyl-CoA and other nitrogenous biochemicals which are abundant in both tryptone and yeast extracts—major components in ×2 YT and LB medium (Layer et al., 2010; Krainer et al., 2015). The ×2 YT medium has a double amount of tryptone and yeast extracts comparing to LB medium. The L-glutamate, which is a key substrate in the heme biosynthesis substrate δ-ALA in E. coli, is ∼3.5 times higher than in LB (Lessard, 2013). Importantly, the ×2 YT medium has higher concentrations of accessible Fe²⁺ and Mg²⁺ which are required for the ferrochelatase and δ-aminolevulinic dehydratase (the porphobilinogen (hemB) synthase), respectively (Sudhamsu et al., 2010; Zhang et al., 2015). Both enzymes are key to the biosynthesis of heme from L-glutamate in E. coli (Woodard and Dailey, 1995; Pranawidjaja et al., 2015; Zhang et al., 2015; Feige et al., 2018; Ge et al., 2023). This might indicate a higher level of heme biosynthesis, and thus its native incorporation into DdDyP, in ×2 YT than that in LB (Table 1). The heme reconstituted DdDyP from LB expression shows a Soret absorbance at 406 nm and an electron transfer (ET) (Q-band) at λ = 497 nm plus two additional ET bands at 536 nm and 576 nm as well as a charge transfer (CT) component at 636 nm, preserving some bacterial peroxidase features (Chen and Li, 2016). This region differs significantly from that previously reported in DdDyP, which showed an ET and CT band at 506 nm and 636 nm, respectively (Rai et al., 2021). Interestingly, our purified DdDyP displays similar features in the Q-band and ET regions to those of the secreted DdPoxA heme peroxidase (Nicolussi et al., 2018). We also observed that the DdDyP with a natively incorporated heme has a broad ET peak with λ _max = 508 nm, which is slightly red shifted with Δλ = 9 nm and Δλ = 2 nm, comparing to that of the ET bands of the reference (Figures 3C, D) and a previous work, respectively (Rai et al., 2021). The Q-band region also reveals a unique shoulder at the ET band with 567 nm absorbance (Figure 3C).

Several DdDyP structures have been resolved so far including a single native structure that is in complex with cyanide (PDB: 7O9L) (Rai et al., 2021), however there is no available structure that describes the native resting state. Here, to get insight into the heme binding pocket in its native form, we crystallized the native DdDyP peroxidase bound to a natively incorporated heme and compared it with that resolved in complex with cyanide. A single crystal with a size of 150 × 80 × 30 μm³ size was used for diffraction data collection (Figure 2D). The crystal data collection and refinement statistics are shown in Table 2. Native DdDyP peroxidase is crystallized in a tetragonal space group P4₁ 2₁ 2 similar to previously reported structures (Rai et al., 2014), with exception that the X-ray data of the current crystal condition can be equally processed and resolved in an additional space group (P4₃ 2₁ 2) (Table 2). Moreover, the crystal unit cell exhibited significantly shorter axes, giving rise to about 35% smaller cell volume with 52.8% solvent content and 2.62 ų Da⁻¹ of Matthew’s coefficient (V _m), indicating the presence of a single molecule per asymmetric unit. The solvent content is decreased by ∼20% comparing to that previously reported (Rai et al., 2021). This is more likely due to a relatively high concentration of the dehydrating precipitant (∼30% PEG 6000), as previously reported in other systems (Umena et al., 2011; Koua et al., 2013). Such high PEG concentration causes a shrinking in the protein crystals by mechanism of dehydration which shortens the axes and leads to a tightly packed unit cell (Supplementary Figure S1). The average radiation dose on a single crystal was estimated with RADDOSE-3D (Bury et al., 2018) to be ∼0.58 MGy (Gy = J•kg⁻¹) (Table 2) which lies well below the 20 MGy dose limit suggested by Henderson (Henderson, 1990) or the 30 MGy suggested by Owen and Garman (Owen et al., 2006), indicating that the structure is less affected by radiation damage.

The overall architecture of DdDyP is similar to that of the typical DyP-type peroxidase superfamily (Figures 4A–D) (Sugano et al., 2007; Chen et al., 2015; Rai et al., 2021). DdDyP contains a duplicated ferredoxin-like fold domain arranged as a β-barrel at the N- and C-terminals of the protein (Figures 4A, B). It contains 12 β-sheets and 13 α-helices formed by 185 residues of the full chain (306 residues), and the remaining 121 residues involved in the formation of loop structures that link these secondary structures. Similar to all other known DyP-type peroxidases, DdDyP contains α-helices with a unique β-sheet structure at the distal region of the heme plane (Sugano et al., 2007; Strittmatter et al., 2013). We determined the root mean square deviation (r.m.s.d.) between the Cα (1–306 residues) of the present structure with that resolved in complex with cyanide (PDB ID: 7O9L) to be 0.18 Å, indicating the striking similarity between the two structures (Figure 4C). Our PISA analysis predicted a stable dimer of DdDyP in solution with 33 residues contributing to the dimer interface, similar to previously reported DdDyP structures (Rai et al., 2021). These interfacial residues are distributed along the dimer interface from the N- to the C-terminal region. The dimeric structure reveals a solvent accessible area of 24,290 Å² and buried surface areas (BSA) of 5,330 Å², corresponding to about 22% of the total surface area of the protein. On the other hand, the BSA of the monomeric structure is 1,341 Å², corresponding to 9.6% of the total surface area of monomeric DdDyP. It should be noted that our PISA analysis favoured a tetramer oligomeric state for the native cyanide DdDyP (PDB ID: 7O9L) structure, displaying higher binding energy than that of the dimeric state. This indicates that DdDyP protein may exist physiologically in various oligomeric states. Indeed, several DyP-type peroxidases have been reported to exist in different functional oligomeric states ranging from monomeric to tetrameric state (Zubieta et al., 2007; Liu et al., 2011; Yoshida et al., 2016; Pfanzagl et al., 2020). The catalytic arginine residue, Arg239 in DdDyP, has been suggested to play a role in the protein oligomerization owing to its location and hydrogen bonding network with surface residues (Singh et al., 2012; Chen et al., 2015). In DdDyP, Arg239 is buried in the hydrophobic cavity of the heme binding pocket, excluding its contribution in DdDyP oligomerization. Moreover, our molecular replacement attempts aiming for a dimeric solution was not successful, thus we can reasonably conclude that our purified DdDyP favours a monomeric state in crystal. It has been previously reported, based on sedimentation velocity analysis with analytical ultracentrifugation, that dimeric DdDyP predominates in solution, which yielded a dimeric crystal structure (Rai et al., 2021). The crystal packing behaviour of the present structure (PDB ID: 8OHY) is significantly different from that described previously (Rai et al., 2021), likely due to a significantly low unit cell volume which exerts tight interactions between molecules in the unit cell (Supplementary Figure S1). Note that the molecular contact within the unit cell is contributed by similar regions in both forms that is primarily the loop and β-sheet of the ferredoxin-fold like domain II at the C-terminal region (Supplementary Figure S1).

Heme b in the DyP-type peroxidases, a protoheme IX, is either penta- or hexacoordinated (Sugano et al., 2007; Singh et al., 2012; Rodrigues et al., 2021). The native structure of DdDyP accommodates heme b in a hydrophobic binding pocket flanked by the unique β-sheet at the distal side of the heme plane, the α-helices of the ferredoxin-like fold domain II (Figures 5A–C, E) and a distinct long loop at its proximal side similar to previously reported DyP structures (Sugano et al., 2007; Zubieta et al., 2007; Liu et al., 2011). Our structural analysis shows that the native DdDyP heme is hexacoordinated, of which the pyrrole rings of porphyrin contributed to tetradentate chelation via their nitrogen atoms and via the conserved His222 at the proximal side with a distance of 2.11 Å. The sixth coordination is provided by a water molecule (wat-184) with a distance of 2.79 Å, which is shorter by ∼0.1 Å than that of the reported Fe(III)─CN distance, indicating a stronger coordination to Fe(III) (Figure 5C). This distance is typical for Fe(III) of the resting state, implying that the model is unaltered by radiation damage (Chen et al., 2015). Wat-184 forms a strong hydrogen bond (∼2.2 Å) with wat-182 and a weaker hydrogen bond with the catalytic residue Arg239 (Figure 5C). Intriguingly, the environment of our native DdDyP heme binding pocket is similar to that of the bacterial DypB and DtpAa peroxidases and that of the peroxidase-cyclooxygenase DdPoxA (Chen et al., 2015; Nicolussi et al., 2018; Ebrahim et al., 2019). The room-temperature serial femtosecond crystallography structure of the DtpAa revealed two water molecules (w1 and w2) in the catalytic vicinity with w1 ligated to the heme-Fe(III) with a distance of 2.32 Å (Ebrahim et al., 2019). We observed that wat-184 has slightly higher B-factor than wat-182, which may indicate its mobility and higher reactivity. On the other hand, wat-182 interacts via strong hydrogen bonds with the second catalytic residue Asp149 as well as Ser241, suggesting that these residues may act as proton acceptors to the H₂O₂ during the formation of compound I oxyferryl thereby contributing to its stabilization along with Arg239 (Figure 5C) as revealed in other A-type DyP peroxidases (Pfanzagl et al., 2019). The binding pocket is extensively occupied with water molecules which are in hydrogen bonding interaction with nearby residues that contribute to the heme stability (Figure 5C). Note that two of the water molecules (wat-150 and wat-159), which are in hydrogen bonding interaction with Asp149 and Arg137 near the heme access channel, substituted the 1,2-ethanediol molecule in the cyanide native structure that shifted Asp149 carboxylate group towards the cyanide, giving rise to increased B-factor of Asp149 comparing to its surrounding (Rai et al., 2021). This may indicate that this position is natively occupied by water molecules as demonstrated by our native structure.

The heme ligand is well resolved at a resolution of 1.95 Å as revealed by its 2mFo–DFc electron density map (Figure 5B; Supplementary Figure S2), indicating unambiguous incorporation and binding of the heme in the apo-protein. This is an important finding as crystallization with purified DdDyP proteins that have lower R _Z values were not successful, which might indicate that the heme on these purified proteins is not well accommodated in the binding pocket, affecting possibly their stability and hence the crystallization (see Table 1). The heme occupies 798 Å² surface area, corresponding to 5.7% of the total surface area of the native structure, similar to that of the cyanide native DdDyP structure and other DdDyP structures (Rai et al., 2021). PISA analysis indicates that the solvation free energy gain (ΔⁱG) of the natively incorporated heme is −22.5 kcal mol⁻¹ with 622 Å² interface area comparing to an average of −22.3 kcal mol⁻¹ for the reconstituted heme of the cyanide native structure (PDB ID: 7O9L) which has an interface interaction area of 609 Å², indicating similar heme binding affinity with slightly better properties for the natively incorporated heme. The van der Waals interactions as well as the hydrogen-bonding network provided by nearby residues and water molecules may contribute to the stabilization of heme binding (Figure 5C) (Sacquin-Mora and Lavery, 2006; Mogharrab et al., 2007). Our analysis shows that the heme is stabilized, along the plane, via its carboxylate oxygens by hydrogen bonding with several water molecules (wat-120, wat-79 and wat-166), and three residues, Glu152, Arg204 and Arg239. The Arg239 interacts weakly with the heme carboxylates oxygens via two hydrogen bonds with a distance of 3.3–3.4 Å, whereas Glu152 and Arg204 form hydrogen bonding with the heme via wat-120 and wat-166, respectively. These interactions indicate that the heme is well stabilized in our model (PDB ID: 8OHY), which confirms the correct geometry of its native incorporation, yielding comparable binding pocket geometry to that prepared with in vitro reconstitution (Chen et al., 2015; Rai et al., 2021). Furthermore, the superimposition with the cyanide native structure indicates a striking similarity (r.m.s.d. = ∼0.18 Å) around the heme binding pocket including the flanking loop at the proximal side of the heme plane (residues 204–220) (Figure 4D). A side-specific mutagenesis study in DypB found that this proximal loop may have significant role in the heme stability (Rodrigues et al., 2021). This loop has also been implicated in the stabilization of the substrate owing to its flexibility thereby facilitating the substrate/product turnover by flipping in and out around the heme binding site (Liu et al., 2011).

Further, we used the MOLEonline tool (https://mole.upol.cz/online) to analyse the cavities and tunnels nearby the heme binding pocket and those in long-range distances (Supplementary Figure S3) (Sehnal et al., 2013). Overall, 14 tunnels were identified, three of which are located next to the heme and perpendicular to each other with characteristics that might have an implication in the DdDyP catalysis—entry of substrates and exit of reaction products. These channels may serve as entry gates for H₂O₂ thereby facilitating the enzyme activation required for the oxidative catalysis (Figure 5A) (Chen et al., 2015; Yoshida et al., 2016; Habib et al., 2019). Two tunnels have average diameter of ∼3.0 Å, which is sufficient to facilitate the entry of H₂O₂ and perhaps the exit of reaction products of similar size. This diameter is about twice the diameter of the substrate channels in the DdPoxA, which is located roughly in a similar position at the distal side (Nicolussi et al., 2018). This may imply variation on the nature and substrate sizes between the two classes of heme peroxidases, the DyP-type and the peroxidase-cyclooxygenase. All tunnels are lined with hydrophobic residues in the middle of the channel as well as several key catalytic residues in the distal and proximal sides of the heme plane. In particular, Arg239, Asp149, His222, Glu152, Ser241, and Thr226 in addition to several hydrophobic residues where found in two proximal channels (Supplementary Figure S3), which are converged to form a main channel with a length of 42 Å and a diameter of ∼4.0 Å. The access gate of this channel is lined with charged residues as shown in Figures 5D, E, indicating its implication in the substrate entry. Further, we identified two major cavities at the distal side of the heme plane, of which one cavity (cavity 1) has a volume of 2,881 Å³, corresponding to 9.6% of the total surface of DdDyP and 4.7 times of the heme molecule. It is located at the heme binding pocket, accommodating the main channel at the binding pocket and extends to the proximal side of the heme plane, indicating a role for this cavity in the catalysis of DyP-peroxidases (Yoshida et al., 2011; Habib et al., 2019; Rai et al., 2021). The second cavity (cavity 2) with approximately half a volume of that of cavity 1 (1,411 Å³) is located at the N-terminal region distant from the heme binding pocket and in contact with cavity 2 near the molecular centre of DdDyP (Supplementary Figure S3). The presence of such cavities is important for accommodating wide-range of substrates thereby fulfilling the substrate broad specificity of DyP-type peroxidases (Pfanzagl et al., 2019; Silva et al., 2023).