Mammalian sulfite oxidase (SO) is a dimeric enzyme consisting of a molybdenum cofactor- (Moco) and haem-containing domain and catalyses the oxidation of toxic sulfite to sulfate. Following sulfite oxidation, electrons are passed from Moco via the haem cofactor to cytochrome c, the terminal electron acceptor. In contrast, plant SO (PSO) lacks the haem domain and electrons shuttle from Moco to molecular oxygen. Given the high similarity between plant and mammalian SO Moco domains, factors that determine the reactivity of PSO towards oxygen, remained unknown. In the present study, we generated mammalian haem-deficient and truncated SO variants and demonstrated their oxygen reactivity by hydrogen peroxide formation and oxygen-consumption studies. We found that intramolecular electron transfer between Moco and haem showed an inverse correlation to SO oxygen reactivity. Haem-deficient SO variants exhibited oxygen-dependent sulfite oxidation similar to PSO, which was confirmed further using haem-deficient human SO in a cell-based assay. This finding suggests the possibility to use oxygen-reactive SO variants in sulfite detoxification, as the loss of SO activity is causing severe neurodegeneration. Therefore we evaluated the potential use of PEG attachment (PEGylation) as a modification method for future enzyme substitution therapies using oxygen-reactive SO variants, which might use blood-dissolved oxygen as the electron acceptor. PEGylation has been shown to increase the half-life of other therapeutic proteins. PEGylation resulted in the modification of up to eight surface-exposed lysine residues of SO, an increased conformational stability and similar kinetic properties compared with wild-type SO.
Molybdenum enzymes catalyse key redox reactions in the global carbon, sulfur and nitrogen cycles . Their overall reaction is characterized by the transfer of an oxygen atom to or from a substrate in a two-electron transfer reaction . Molybdenum is bound to the pterin-based molybdenum cofactor (Moco) of those enzymes, which in most cases harbour additional prosthetic groups for intramolecular electron transfer (IET). Most molybdenum enzymes are found in bacteria; in eukaryotes only five molybdenum enzymes are known so far .
In mammals, the most important molybdenum enzyme is sulfite oxidase (SO), which is mainly found in liver, where it catalyses the oxidation of sulfite, which is generated throughout the catabolism of cysteine . Animal SO is a dimeric enzyme, which harbours a cytochrome b5-type haem domain in addition to a Moco domain . The catalytic cycle of SO involves electron transfer from sulfite to Moco, followed by two electron-transfer steps via the cytochrome b5 domain to the terminal electron acceptor cytochrome c . The orthologous plant SO (PSO) lacks the haem domain and thereby constitutes the simplest eukaryotic molybdenum enzyme . As a result, electrons derived from sulfite oxidation are passed directly to molecular oxygen, a process that generates hydrogen peroxide (H2O2), which, in the presence of the enzyme catalase, is converted further into water and oxygen . The latter explains the peroxisomal localization of PSO , whereas animal SO is localized in the mitochondrial inter-membrane space, where it uses cytochrome c as the terminal electron acceptor .
SO from chicken was the first reported crystal structure of a eukaryotic molybdenum enzyme and has been fundamental in understanding the relationship between the molybdenum and haem domains in animal SO . Within the crystal structure of chicken SO, a large distance of 32 Å (1 Å=0.1 nm) was found between the molybdenum and haem domain , which would not support the high electron transfer rates measured between both domains using laser flash photolysis methods . Thus it has been suggested that animal SO may undergo conformational change that brings the molybdenum and haem domains in close proximity in order to allow fast electron transfer. On the basis of the decrease in IET as a function of viscosity  and residue deletions within the tether connecting the molybdenum and haem domains , it was confirmed that animal SO undergoes dynamic conformational changes, which guide the haem domain in docking through electrostatic interactions to the molybdenum domain thus allowing efficient electron transfer.
The reaction mechanism of SO can be divided into a reductive and oxidative half-reaction . In the reductive half-reaction, sulfite binds at the MoVI centre and is oxidized to sulfate by the transfer of two electrons to the molybdenum centre yielding the reduced MoIV species. According to the respective reduction potentials, one electron is transferred via IET to the haem domain creating a paramagnetic MoV intermediate state, which can be detected using EPR spectroscopy . The oxidative half-reaction is initiated with the transfer of one electron from haem to the final electron acceptor cytochrome c. Then the second electron can leave the MoV centre by a second IET step via haem to a second cytochrome c yielding the fully oxidized form of the enzyme . The crystal structure of PSO from Arabidopsis thaliana depicted the close resemblance of the molybdenum domain between animal and plant SOs  suggesting a similar reductive half-reaction during the catalytic cycle of sulfite oxidation for animal and plant SOs . However, the absence of the haem domain in PSO implicates a different oxidative half-reaction from that in animal SO. In fact, it has been shown that the molybdenum domain of PSO reacts directly with oxygen leading to the formation of superoxide ions as the immediate product of the oxidative half-reaction, which is spontaneously dismutated to H2O2 . The PSO reaction with oxygen is very unusual as none of the molybdenum enzymes characterized so far are able to react at significant rates with oxygen . Furthermore, plant and animal SO share a structurally very similar molybdenum centre and residues important for substrate binding are strictly conserved . Thus the factors responsible for PSO reactivity towards oxygen remain unknown. The impact of mutations on IET between the molybdenum and haem domains have been investigated using the human SO enzyme [18,19]. Interestingly, the authors showed also a reactivity of human SO with ferricyanide as the electron acceptor, which, unlike cytochrome c, can directly be reduced by the molybdenum centre. However, oxygen reactivity studies were not performed with human SO .
Defects in any step of the biosynthesis of Moco lead to Moco deficiency (MoCD), a rare inherited metabolic disorder resulting in the loss of activity of all molybdenum enzymes, and affected patients usually die in early childhood . Mutations in the SUOX gene result in SO deficiency (SOD); approximately 30 cases have been reported so far . MoCD and SOD are clinically very similar, which qualifies SO as the most important molybdenum enzyme in humans . Both deficiencies are characterized by a severe neurodegenerative phenotype  resulting from the accumulation of sulfite and other toxic metabolites such as S-sulfocysteine . MoCD can be grouped into three types according to the underlying genetic defect . Type A deficiency affects two-thirds of all patients and is caused by mutations in the MOCS1 (molybdenum cofactor synthesis 1) gene . Type B patients accumulate the first Moco intermediate cPMP (cyclic pyranopterin monophosphate)  due to defects in the MOCS2 gene . Type C deficiency affects the GPHN (gephyrin) gene  with only two cases reported to date [29,30].
Until the last few years, no effective therapy was available for MoCD, and death in early childhood has been the usual outcome . The ability to purify the first Moco intermediate cPMP  was the starting point for the establishment of the first treatment approach towards MoCD type A and, since 2010, several replacement therapies with cPMP have been reported for MoCD type A patients [31,32]. Treated patients were exposed to repetitive intravenous injections of cPMP, which resulted in the restoration of molybdenum enzyme activities and the normalization of all disease biomarkers [31,32]. However, cPMP is the only reported stable Moco intermediate and similar therapies for MoCD type B and C are not feasible . Knowing that the clinical symptoms observed in MoCD are mainly caused by the loss of SO activity, a possible replacement therapy with purified SO may be considered. However, such a therapy is limited by the fact that SO catalytic activity requires mitochondrial translocation, which will result in unfolding of the protein and loss of Moco.
In the present study, we investigated mammalian SO from two different aspects. Given the high similarity of PSO and animal SO, we first aimed to identify factors that restrict the reactivity of animal SO towards oxygen. We investigated the production of H2O2 by different mammalian SO variants with altered or defective IET between the molybdenum and haem domains and found a significant reactivity towards oxygen for IET-restricted SO variants with nearly comparable rates between PSO and haem domain-deficient mammalian SO. Secondly, we investigated the potential use of oxygen-reactive SO proteins in a replacement therapy towards MoCD as such an enzyme may use blood-dissolved oxygen as an electron acceptor, thus overcoming mitochondrial translocation. As high immunogenicity and a reduced protein half-life are common for protein-based therapeutics, we explored the possibility to use PEG-based protein modification (PEGylation) as a tool to increase the half-life of a protein in a biological environment without loss of activity. PEGylation of proteins has been shown in previous studies to effectively increase the serum half-life of proteins and decrease their immunogenicity [33,34]. Furthermore, PEGylated proteins have been approved by the U.S. FDA (Food and Drug Administration) for human administration and are successfully used in many diseases such as hepatitis and cancer [35,36]. We used the PEGylation of surface-exposed lysine residues of SO by the N-hydroxysuccinimide (NHS) ester method and found a significant impact on the oligomerization state of the protein, whereas kinetic properties and oxygen reactivity were only moderately altered. Finally, cell-based assays showed that PEGylated oxygen-reactive mammalian SO in combination with catalase were effective in preventing sulfite-mediated toxicity, suggesting a dual replacement therapy as a possible novel therapeutic route towards the treatment of MoCD and SOD.
Expression constructs for human and murine SO molybdenum domain (HSOMo and MSOMo) were generated by cloning the coding sequence for the dimerization and Moco domains [human SO (HSO), GenBank® accession number AY056018.1, residues 110–488; murine SO (MSO), GenBank® accession number BC027197.1, residues 168–546] into pQE80L (Qiagen) using SalI and HindIII restriction sites. The same restriction sites were used for cloning of HSO wild-type (wt). HSO deletion variants HSOΔKVATV and HSOΔKVAPTV were generated by fusion PCR and cloned into the pQE80L vector using SacI and SalI restriction sites. MSO (wt) and MSOΔhaem were generated as described previously . For catalase expression, the coding sequence of human catalase (GenBank® accession number BC110398.1) was PCR-cloned into pQE80L using SalI and HindIII restriction sites. For recombinant expression of PSO, the previously described rAt-SO construct was used .
Protein expression and purification
All SO proteins were expressed in Escherichia coli TP1000  as described previously . Human catalase was expressed in E. coli BL21(DE3) cells. Expression was induced with 0.1 mM IPTG at an A600 of 0.1 and continued for 15 h at 30°C. All His-tagged proteins were purified by Ni2+-nitrilotriacetic acid (Ni-NTA) affinity as recommended by the manufacturer (Qiagen). For PSO, a second purification step consisting of an anion-exchange chromatography was performed as described previously . All purified proteins were exchanged into the same buffer (20 mM Tris/HCl, pH 8.0, and 50 mM NaCl) and stored at −80°C.
Determination of Moco saturation
Moco saturation was determined by denaturing 500 pmol of protein using acid iodine oxidation and alkaline phosphatase treatment resulting in the formation of the stable Moco oxidation product FormA-dephospho, which was quantified further using reverse-phase HPLC as described in .
Determination of SO activity
For PSO, HSOMo and MSOMo, activities were either measured using the sulfite:ferricyande or sulfite:cytochrome c assay. Sulfite:ferricyanide activity was measured by monitoring the reduction of ferricyanide [Fe(CN)6] at 420 nm (ε420=1020 M−1·cm−1) . The assay included the following components: 140 μl of 100 mM Tris/acetate (pH 8), 20 μl of protein solution and 20 μl of 4 mM Fe(CN)6, and the reaction was started by the addition of 20 μl of sodium sulfite (various concentrations). Activities were measured at an enzyme concentration of 50 nM and 500 nM for the plant and mammalian SO proteins respectively.
Sulfite:cytochrome c activity was determined for HSO and deletion variants HSOΔKVATV and HSOΔKVAPTV by monitoring the absorption change of cytochrome c at 550 nm (ε550=19630 M−1·cm−1) . Briefly, equal proteins concentrations (10 nM) were incubated in a 200 μl final volume with a mixture containing 50 mM Tris/acetate (pH 8) and sodium sulfite (various concentrations), and the reaction was started by adding 12 μl of cytochrome c (10 mg/ml). SO activity in crude protein extracts was determined in a similar way with the following modifications: 50 μg of protein crude extract was used and the assay buffer mixture contained 50 mM Tris/acetate (pH 8), 0.2 mM deoxycholic acid, 0.1 mM potassium cyanide and 0.5 mM sodium sulfite. All activities were measured at room temperature (25°C) using a 96-well plate reader (BioTeK).
Quantification of H2O2 is based on the formation of a complex between Xylenol Orange and ferric ions (Fe3+), which is produced by the peroxide-dependent oxidation of ferrous iron (Fe2+). The method was performed using a commercial kit (National Diagnostics) and detection was carried out colorimetrically following the protocol of the manufacturer. Quantification was carried out after an incubation time of 30 min at room temperature (25°C) by measuring the absorption at 560 nm using a 96-well plate reader.
Oxygen consumption was measured using an Oroboros Oxygraph 2k Instrument. First, a 2 ml solution containing 100 mM Tris/HCl (pH 8.0) and 250 nM purified enzyme was introduced into the Oxygraph chamber and maintained at 37°C for approximately 15 min to equilibrate oxygen concentration in the chamber. The reaction was started by addition of 60 μM sulfite and the slopes corresponding to the linear change in oxygen concentration of the plots were used to calculate the oxygen-consumption rates. Data were analysed using DatLab4 software (version 4.3).
Western blots were performed on mouse crude extracts of liver, kidney and brain derived from three different animals. The protein concentration was determined using the Bradford method and 50 μg of each crude extract was separated by SDS/PAGE (10% gel). Primary antibodies used were anti-sulfite oxidase (Eurogentec) and anti-actin (Santa Cruz Biotechnology). Secondary antibodies coupled to horseradish peroxidase (Abcam) were visualized using Super Signal West Pico Chemiluminescent Substrate (Thermo Scientific), and analysis of the protein bands was performed with the Chemoluminescence DeVision HQ2 camera system and the Gel-Pro Analyzer software (Decon Science Tec).
For PEGylation of plant and mammalian SO proteins, the NHS method was used . Three different PEG reagents (mPEG NHS ester; Celares) different in size and chemistry were applied (for details, see Supplementary Figure S3). Briefly, the SO proteins were exchanged into PBS and concentrated to 5–10 mg/ml. PEG reagent was added to the protein solution at a 20-fold molar excess and the solution was incubated for 30 min at room temperature (25°C). Finally, excess of non-reacted PEG was removed by size-exclusion chromatography using a HR16/30 Superdex 200 column (GE Healthcare).
Cell viability studies were conducted in human embryonic kidney (HEK)-293 cells using the MTT assay (Promega). Briefly, 80 μl of HEK-293 cell suspension (containing 2×104 cells) were dispensed into each well of a 96-well tissue culture plate and incubated overnight at 37°C in a humidified 5% CO2 atmosphere. Next, 10 μl of SO and/or catalase proteins were added to each well at a final concentration of 0.5 μM and allowed to incubate for 30 min at 37°C. Then, 10 μl of sulfite (various concentrations) was added to each well and incubated for 15 h at 37°C in a humidified 5% CO2 atmosphere. H2O2 toxicity was also investigated in the absence and presence of catalase using a concentration range of 0–0.5 mM enzyme. Finally, cell viability was evaluated using the MTT dye according to the supplier's protocol, and absorption at 570 nm (reference 650 nm) was recorded using a well plate reader (Tecan).
Electron transfer between Moco and haem determines SO reactivity towards oxygen
SO from chicken (Gallus gallus) and plant (Arabidopsis thaliana) share 47% sequence identity despite the lack of the haem domain in PSO. However, only PSO has been reported to react with oxygen, a process involving the formation of superoxide ions, which are further dismutated to H2O2 . Given the similar kinetic properties in sulfite oxidation between mammalian and plant SOs and their different domain structure, we asked to what extent the haem domain affects oxygen reactivity in mammalian SO. Therefore we first determined H2O2 production as a function of sulfite oxidation in both animal and plant SOs. For this purpose, a colorimetric method that quantifies all organic peroxides including superoxide ions and H2O2 was used, and the exclusive production of H2O2 was probed by the addition of recombinantly expressed and purified human catalase. We first determined sulfite-dependent H2O2 production (for 30 min) of wt MSO and HSO, and compared this with that of PSO. H2O2 formation was low in the presence of MSO showing 15 μM H2O2 formation with 75 μM sulfite (Figure 1A), whereas HSO did not lead to any significant H2O2 production (<2 μM, Figure 1B). In contrast, when using PSO, H2O2 formation correlated with the increase in sulfite in a linear manner, suggesting a stoichiometric turnover (Figure 1C). In all experiments, no signal was detected upon the addition of catalase (Figure 1), suggesting that, under our experimental conditions, only H2O2 is formed and detected.
The haem domain affects H2O2 formation in mammalian SO
Next, we used two different approaches to investigate the impact of the haem domain on oxygen reactivity in mammalian SO. First, a haem-deficient MSO variant (MSOΔhaem) was generated by replacing both haem-co-ordinating histidine residues with alanine (H119A/H144A), thus resulting in a loss of haem binding . Disruption of haem binding did not result in any perturbation of the folding of MSO, as assessed by CD spectroscopy (Supplementary Figures S1A–S1C). In contrast with wt MSO, MSOΔhaem showed a linear sulfite-dependent production of H2O2 resulting in more than 50 μM H2O2 formed from 75 μM sulfite (Figure 1D), demonstrating that, in the absence of a functional haem domain, the reactivity of MSO towards oxygen is favoured.
In a second approach, we investigated the impact of deletions of the tether connecting the molybdenum and haem domains on oxygen reactivity of HSO. In mammalian SO, IET between the molybdenum and haem domains is essential to complete the catalytic cycle. The IET process was investigated extensively in HSO using laser flash photolysis, which enables the measurement of the IET rate constants between both redox centres in HSO and identified the importance of the tether linking the molybdenum and haem domain during the IET process . Accordingly, deletion of five residues within the tether in the HSOΔKVATV variant resulted in a strong decrease in the IET rate constant (467±19 s−1 in wt HSO and 5.59±0.03 s−1 in HSOΔKVATV) owing to restricted domain mobility .
Therefore we generated the same HSO deletion variant (HSOΔKVATV) and a second deletion variant (HSOΔKVAPTV) by additionally deleting a proline residue (Pro111) within the tether between the molybdenum and haem domains . Similar to MSOΔhaem, both HSO deletion variants did not display structural changes in comparison with wt HSO as assessed by CD spectroscopy (Supplementary Figures S1D–S1F). However, both tether deletions in HSO resulted in a dramatic decrease in sulfite:cytochrome c activity as documented by a 3-fold and 15-fold decrease in kcat for HSOΔKVATV and HSOΔKVAPTV respectively compared with wt HSO (Supplementary Figure S2A). When using HSOΔKVATV, which is not deficient in haem but shows a reduced IET rate constant  between Moco and haem, again a linear dependence of H2O2 production on sulfite concentration was observed (Figure 1E). However, the molar ratio of H2O2 formed per sulfite, fell from 0.72 to 0.41 compared with MSOΔhaem (Figure 1D). The additional deletion of Pro111 in HSOΔKVAPTV further reduced sulfite:cytochrome c activity (Supplementary Figure S2A), whereas the rate of H2O2 formation per mol of sulfite increased to 0.54 (Figure 1F).
Mammalian SO molybdenum domains show oxygen reactivity similar to PSO
In order to structurally mimic PSO, bacterial expression constructs of the Moco and dimerization domains of murine (MSOMo) and human (HSOMo) SO corresponding to the C-terminal 378 residues of both mammalian proteins were generated, recombinantly expressed and purified. First, we characterized those mammalian SOMo variants by steady-state kinetics using the sulfite:ferricyanide assay (Figures 2A and 2B) and compared their activities with that of PSO (Supplementary Figure S2B). Mammalian SOMo variants revealed two major differences from PSO. First, SOMo variants showed no inhibition at high substrate concentration, which was in contrast with PSO. Secondly, the determined kcat values for MSOMO and HSOMO were 2.4- and 4.7-fold lower respectively than for PSO (Figures 2A and 2B, and Supplementary Figure S2B).
Haem deletion enables oxygen reactivity of mammalian SO
Sulfite-dependent H2O2 formation was demonstrated for both SOMo variants, confirming their ability to use molecular oxygen as the electron acceptor (Figures 2C and 2D). In contrast with PSO, which showed a nearly stoichiometric (0.96) correlation between H2O2 formation and sulfite concentration (Figure 1C), H2O2 formation mediated by MSOMo and HSOMo showed a molar ratio between H2O2 and sulfite of 0.68 and 0.71 respectively (Figures 2C and 2D). Knowing that H2O2 is able to oxidize sulfite non-enzymatically  (Supplementary Figure S2C), we propose that the reduced catalytic activity of the SO haem-deleted variants measured using the sulfite:ferricyanide assay (Figures 2A and 2B) favoured the non-catalytic sulfite oxidation by H2O2 (compare Figures 1D–1F), thus leading to an overall reduction in H2O2 accumulation. Additionally, H2O2 was almost not detected at low (below 10 μM) sulfite concentrations using the mammalian SOMo variants (Figures 2C and 2D). This is due to the fact that, under aerobic conditions and low concentrations, sulfite is susceptible to air oxidation, which, combined with the non-catalytic sulfite oxidation mediated by H2O2, might explain the detection limit of H2O2 at low sulfite concentrations.
Reduced IET increases oxygen consumption in mammalian SO
Next, we determined the oxygen consumption of SO variants by using an Oxygraph instrument, which allows direct determination of sulfite:oxygen activity. Addition of 60 μM sulfite to 0.25 μM PSO induced a rapid decrease (within 5 s) in the oxygen concentration from 190 μM to 124 μM, which correlated well with the amount of sulfite added, thus suggesting a stoichiometric conversion of sulfite into H2O2 (Figure 3A). The resulting velocity of 14 μM·s−1 (Figure 3B) converts into a rate of 56 s−1, which is comparable with the rates reported previously . However, under our assay conditions, inhibition at substrate concentrations higher than 80 μM was observed for the PSO-catalysed reaction (Supplementary Figure S2B). Therefore, under saturating sulfite concentrations, a direct comparison of the oxygen-consumption rates of plant and mammalian SO variants is not feasible.
Oxygen consumption by plant and mammalian SO variants
We measured oxygen-consumption rates for all mammalian SO variants (Figure 3B). The activity of MSO and HSO remained very low (1–2 μM·s−1), supporting the previous results of H2O2 formation (Figure 3B). Comparison of MSO with its haem-deficient and domain variants or HSO with its deletion and domain variants, showed an increase in oxygen consumption, which correlates with the respective decrease in IET between Moco and haem in the HSO deletion variants (measured by sulfite:cytochrome c activity). Consequently, oxygen reactivity was highest when the haem domain was completely absent (using haem-deficient or SOMo domain variants; Figure 3B). In conclusion, either the absence of the haem cofactor or a reduced IET results in a strong increase in oxygen reactivity of mammalian SO, thus suggesting that oxygen is competing with the haem cofactor for the electrons derived from sulfite oxidation.
Mouse brain lacks capacity for sulfite oxidation
Knowing that SO is the most important molybdenum enzyme in mammals and that both SOD and MoCD are mainly characterized by sulfite-induced neuronal cell death, we asked to what extent sulfite is oxidized in the brain. We determined SO expression levels and SO activities in mouse liver, kidney and brain crude protein extracts. SO protein was not detectable in the brain, whereas liver and kidney showed a strong expression (Figure 4A). Furthermore, SO activity was nearly undetectable in the brain, whereas liver and kidney revealed high activities (Figure 4B) as expected from the SO expression levels (Figure 4A). Knowing that liver is characterized by a high capacity for sulfur amino acid catabolism , it is not surprising that sulfite is primarily generated in the liver . Accordingly, a treatment option for MoCD and SOD could rely on removing sulfite from the circulation, which would be favourable in preventing sulfite-induced brain damage.
SO expression and activity are low in mouse brain in comparison with liver and kidney
PEGylation of SO induces changes in stability and oligomerization
Up to now, a replacement therapy for MoCD and SOD using mammalian SO has not been feasible due to the requirement for SO translocation into mitochondria where its native electron acceptor, oxidized cytochrome c, is localized. The ability of mammalian SOMo domain variants to use oxygen as the electron acceptor offers the possibility for the development of an enzyme replacement therapy towards MoCD and SOD, in which SO variants can use dissolved oxygen in blood for sulfite oxidation. To increase the stability of SOMo proteins and to suppress potential immune responses, we investigated protein PEGylation as a modification method, which has been successfully applied with other protein- and peptide-based therapeutics .
For PEGylation we used NHS-activated PEG molecules that were covalently coupled to surface-exposed lysine residues of different SO variants. This method is based on the reaction of activated PEG esters with primary amino groups of the protein, which lead to the formation of stable amide bonds between lysine residues and PEG molecules . Thus the number of PEG molecules attached on the surface of the protein is directly proportional to the number and the accessibility of surface-exposed lysine residues. As a result, the molecular mass of the modified protein increases according to the number and molecular mass of the attached PEG molecules.
We first PEGylated PSO and investigated the effects of parameters such as incubation time and structure of PEG molecules on the efficacy of modification (Figures 5A and 5B). When using a 4.2 kDa PEG molecule, approximately 50% of PSO was PEGylated within 2 min, as depicted by a shift in molecular mass, and, within 30 min, PEGylation was nearly completed (Figure 5A). Next, two PEG molecules differing in size and structure were used to PEGylate PSO: a branched (4.2 kDa) and a linear (5 kDa) PEG (Supplementary Figure S3). SDS/PAGE analysis showed an increase in molecular mass of all PEGylated proteins, demonstrating that the PEG molecules were covalently bound (Figure 5B). The exact molecular mass of the PEGylated proteins could not be determined by SDS/PAGE, as they displayed a heterogeneous distribution pattern in the gel (branched PEG) or the apparent molecular mass exceeded the resolution range of the gel (linear PEG). Therefore MS was applied, and the number of added PEG molecules was determined on the basis of the apparent mass increase. We found that four and eight PEG molecules were coupled to PSO by using the branched or linear PEG, respectively (Supplementary Table S1).
PEGylation induces changes in stability and oligomerization of SO proteins
Owing to the higher coupling efficacy of PSO with the linear 5 kDa PEG, MSOMo and HSOMo were PEGylated using two sizes of linear PEGs (0.5 and 5 kDa, Supplementary Figure S3). Modified proteins were separated from excess PEG molecules by size-exclusion chromatography (Figures 5C and 5D). SDS/PAGE analysis showed again a shift in the molecular mass of all PEGylated proteins (Supplementary Figure S4A). PEGylation with the 5 kDa PEG resulted in proteins displaying a heterogeneous band pattern, which was due to the polydisperse nature of the PEG used. When using the monodisperse 0.5 kDa PEG, modified proteins displayed a single sharp band after SDS/PAGE (Supplementary Figure S4A). Furthermore, PEGylation of SO proteins did not change their Moco content (Supplementary Figure S4B).
Size-exclusion chromatography confirmed the corresponding increase in molecular mass of PEGylated PSO (Figure 5C and Supplementary Online Data). HSOMo displayed a heterogeneous elution profile depicted by the presence of three peaks corres-ponding to monomers, dimers and high-molecular-mass oligomers (Figure 5D). However, PEGylation of HSOMo variants resulted in the formation of a single homogeneous peak in the size-exclusion chromatogram (Figure 5D) and similar results were obtained using the murine SOMo variant (Supplementary Figure S5). Dimerization of molybdenum enzymes is expected to follow the incorporation of Moco , and it has been shown that mammalian SO dimerization and maturation is independent of haem binding . Neither deletion of the haem domain in the HSOMo variant nor PEGylation of HSOMo resulted in major structural changes in comparison with wt HSO as assessed by CD spectroscopy (Supplementary Figures S6A and S6B). Furthermore, stability of wt HSO, HSOMo and PEGylated HSOMo proteins were not affected by Moco saturation and sulfite:cytochrome c activities remained stable over 10 h (Supplementary Figure S6C and S6D). Thus we conclude that the impact of HSOMo PEGylation observed using size-exclusion chromatography does not rely on an increased stability, but is rather due to an increased solubility due to the amphiphilic character of the PEG molecules used (Supplementary Online Data).
PEGylation of SO preserves its catalytic activity
Following PEGylation, a change or loss in catalytic activity may occur if the attached PEG molecules hinder substrate-binding, conformational flexibility and/or catalysis. Therefore sulfite:ferricyande steady-state kinetics of PEGylated PSO, MSOMo and HSOMo were first determined. We found that PEGylation of SO proteins did not result in major changes in the corresponding kinetic parameters (Figures 6A–6D, Table 1 and Supplementary Figures S7A and S7B). PEGylation of PSO with 5 kDa PEG did not alter its catalytic turnover (kcat of 483 min−1 compared with 511 min−1), whereas a slight increase in kcat was observed when the 0.5 kDa PEG was used (kcat of 650 min−1 compared with 511 min−1). On the other hand, an approximately 2-fold increase in Km was determined for both PEGylated PSO proteins compared with native PSO (Table 1).
|Protein||Km (μM)||kcat (min−1)||kcat/Km (min−1·μM−1)||Ki (mM)|
|PEG–PSO (0.5 kDa)||68±6||651±29||9.5±4.7||1.38±0.18|
|PEG–PSO (5 kDa)||63±7||483±28||7.7±3.5||2.42±0.60|
|PEG–MSOMO (0.5 kDa)||149±11||303±7||2±0.7|
|PEG–MSOMO (5 kDa)||58±5||175±4||3±0.8|
|PEG–HSOMO (0.5 kDa)||102±7||111±2||1.1±0.3|
|PEG–HSOMO (5 kDa)||110±5||134±2||1.2±0.4|
|Protein||Km (μM)||kcat (min−1)||kcat/Km (min−1·μM−1)||Ki (mM)|
|PEG–PSO (0.5 kDa)||68±6||651±29||9.5±4.7||1.38±0.18|
|PEG–PSO (5 kDa)||63±7||483±28||7.7±3.5||2.42±0.60|
|PEG–MSOMO (0.5 kDa)||149±11||303±7||2±0.7|
|PEG–MSOMO (5 kDa)||58±5||175±4||3±0.8|
|PEG–HSOMO (0.5 kDa)||102±7||111±2||1.1±0.3|
|PEG–HSOMO (5 kDa)||110±5||134±2||1.2±0.4|
PEGylated SO retains catalytic activity and oxygen reactivity
The influence of PEGylation on the activity of mammalian SOMo proteins was similar to that observed for PSO, resulting in an approximately 2-fold increase in Km with either similar or even increased kcat values compared with those of the non-modified enzymes (Figures 6C and 6D, Table 1 and Supplementary Figures S7A and S7B). Furthermore, PEGylation of PSO and mammalian SOMo variants did not alter their ability to generate H2O2. All PEGylated SO proteins showed similar levels of H2O2 production to those of non-modified proteins (Figures 6E–6H and Supplementary Figures S7C and S7D). Interestingly, also here a very low H2O2 concentration was detected at low sulfite concentrations (below 10 μM) for the PEGylated proteins, which was more visible using mammalian SOMo variants, attesting to a higher contribution of the non-catalytic sulfite oxidation mediated by H2O2 in mammalian SOMo variants in comparison with PSO (Figures 6E–6H and Supplementary Figures S7C and S7D).
SO is able to catalyse sulfite oxidation using dissolved oxygen in cell culture
Our previous experiments showed that mammalian SO variants are able to react with oxygen in vitro. Furthermore, we showed that PEGylation as a masking method, consisting of shielding the protein by the covalent attachment of PEG molecules to the surface-exposed lysine residues of the protein, did not cause loss of activity. We next evaluated the capacity of both plant and human SOs to use oxygen as an electron acceptor, thus generating H2O2 in a cell-based assay. For this purpose, we exposed HEK-293 cells to SO-dependent H2O2 toxicity. Cell viability was determined using the MTT cell proliferation assay. Even very low amounts of H2O2 induced toxicity in HEK-293 cells, which was prevented if purified human catalase was co-incubated with the cells exposed to H2O2 (Supplementary Figures S8A and S8B).
Next, we probed the role of PSO in H2O2 generation. Exposing HEK-293 cells only to sulfite concentrations up to 0.5 mM was not toxic, given that those cells express functional SO (Figure 7A). However, when HEK-293 cells were co-incubated with 0.1 mM (and higher concentrations) sulfite and PSO, a significant cell death with only 20% cell viability was measured (Figure 7B). PSO-mediated toxicity was prevented when purified catalase was added (Figure 7C), being in line with the PSO-dependent formation of H2O2 in the presence of sulfite. On the basis of these results, we next used 0.5 mM sulfite and investigated the sulfite-dependent H2O2 formation of non-modified and PEGylated PSO and HSOMo as well as full-length HSO. Similar to PSO, HSOMo as well as the PEGylated proteins were equally effective in inducing cell death resulting in 20% cell survival compared with control (Figure 7D), and toxicity was again prevented if purified catalase was added (Figure 7E). In contrast, full-length HSO was not able to induce cell toxicity, which again confirms its inability to accept oxygen as a suitable co-substrate (Figure 7D).
SO prevents sulfite-dependent H2O2 toxicity in HEK-293 cells
Mammalian and plant SO are localized in different cellular compartments catalysing the oxidation of sulfite by coupling electron transfer either to mitochondrial respiration or peroxidation. As a result, mammalian SO requires a haem domain-mediating electron transfer to cytochrome c, whereas PSO consists only of a single catalytic domain, which passes electrons directly to molecular oxygen. Neither vertebrate, nor bacterial forms of SO were reported to react with oxygen at any appreciable rates [17,46], whereas PSO uses molecular oxygen as electron acceptor for sulfite oxidation and consequently produces H2O2 . Using a colorimetric detection method of all organic peroxides together with the application of catalase confirmed a stoichiometric sulfite-dependent H2O2 formation by PSO. In contrast, with mammalian SOs, H2O2 formation was very low. However, using a MSO variant with deficient haem binding, two human SO variants with hinge truncations impairing haem domain mobility, or haem-deficient murine and human variants, consistently demonstrated that reducing or abolishing IET between the molybdenum and haem domains enabled oxygen reactivity of mammalian SOs. Highest levels of oxygen reactivity were found for the haem-deficient SOMo variants as well as MSOΔhaem, whereas intermediate activities were found for the hinge deletion variants of HSO. In particular, HSOΔKVATV has been shown in previous studies to exhibit an almost 100-fold reduction in the IET rate , which, in contrast, resulted in only 3-fold decreased sulfite:cytochrome c activity (Supplementary Figure S2A). An additional deletion of Pro111 (conserved residue in animal SO) further reduced the steady-state activity in HSOΔKVAPTV, suggesting a further diminished IET, which correlated well with the increased reactivity towards oxygen.
What is the mechanism of haem-dependent inhibition of sulfite-dependent oxygen reduction in mammalian SOs? The catalytic cycle of mammalian SO involves the oxidation of MoVI to MoIV followed by two sequential one-electron transfer steps to haem, passing by the paramagnetic MoV intermediate state . In the absence of cytochrome c and presence of excess sulfite, the enzyme will end up in a two-electron reduced state with a ferrous haem and a MoV. The lack of reactivity towards oxygen in mammalian SO suggests that the MoV state is not able or is less able to react with oxygen. Interestingly, studies of the MoV centre of PSO using EPR spectroscopy suggested that depending on the conformation of Arg374, the active site of PSO may exist in a ‘closed’ or ‘open’ form that differs in the degree of accessibility of the molybdenum centre to substrate and water molecules . In this study, deletion of the entire haem domain resulted in an increased oxygen reactivity of mammalian SO similar to that of PSO. We therefore propose that the presence of haem in mammalian SO controls the conformational change allowing the enzyme to pass from a ‘closed’ to an ‘open’ form that only reacts with oxygen if the IET rate between the molybdenum and haem domains is decreased or abolished.
Impaired sulfite oxidation is the major cause of neuronal cell death in MoCD and SOD. Furthermore, the majority of sulfite in MoCD and SOD originates from peripheral tissue, as we could not detect any significant SO protein as well as activity in brain extracts, which is in agreement with previous studies . This assumption is supported by effective treatment of MoCD in mice following the restoration of Moco biosynthesis in liver . We therefore conclude that sulfite is primarily generated in liver and kidney and is transported to the brain via the vascular system. Consequently, we assume that enzyme-replacement therapies targeting sulfite removal from the blood will be effective in preventing sulfite toxicity. Our finding that mammalian SO variants can function outside mitochondria by transferring electrons to oxygen in the absence of a functional haem domain provides a novel approach towards the treatment of the sulfite-toxicity disorders SOD and MoCD.
Enzyme-replacement therapies are based on either systemic distribution of enzyme without any particular subcellular targeting or the use of cell-surface receptors such as the mannose 6-phosphate receptor to promote internalization of enzymes acting within the vesicular compartment [50,51]. To our knowledge, enzyme-replacement therapies targeting mitochondria have not yet been reported. Given the fact that cellular maturation of mammalian SO is dependent on a highly orchestrated co-transport of apo-SO and Moco into the mitochondrial inter-membrane space , targeting of full-length and active SO into mitochondria using an enzyme-replacement approach is not feasible. Therefore we investigated SOMo PEGylation as a potentially suitable method to increase stability of haem-deficient SO. PEG is a non-toxic molecule that has been approved by the FDA for human administration and was successfully used in the treatment of diseases such as hepatitis and cancer [35,36].
PEGylated mammalian SOMo variants appeared as homogeneous dimeric proteins in contrast with the non-modified proteins, which showed a high degree of oligomeric heterogeneity, probably due to the lack of haem domain and subsequent exposure of hydrophobic surface patches. PEGylated SO proteins showed only minor changes in their activity as demonstrated by sulfite:ferricyanide steady-state kinetics. More importantly, H2O2 production and thereby oxygen reactivity was preserved for all PEGylated proteins.
HEK-293 cells were used as a cellular model system to study sulfite-dependent H2O2 formation. Whereas sulfite (up to 500 μM) was well tolerated, SO-dependent H2O2 formation severely affected cell survival. Using this assay, we could confirm that mammalian haem-deleted SO variants catalyse sulfite oxidation by using oxygen from the culture medium as the electron acceptor. However, high levels of H2O2 produced by SO and the resulting cellular toxicity clearly indicate that a catalase is required to efficiently remove produced H2O2. Alternatively, in the presence of low amounts of SO, the clearance rate of sulfite would be low, which in turn would enable the non-enzymatic oxidation of sulfite by H2O2, formed in the first case.
Recently, an SO-deficient patient with a mild phenotype has been reported , showing a positive response following a methionine- and cysteine-restricted diet suggesting the presence of residual SO activity in this patient. Given the fact that the disease-causing mutation affects one of the two haem-co-ordinating histidine residues (H143A), we conclude that the sulfite:cytochrome c activity of SO is abolished. On the basis of the findings of the present study, we argue that the SO H143A variant is able to use oxygen as the electron acceptor, thus retaining some degree of residual SO activity and thereby prohibiting the development of the typical and severe symptoms of SOD in the patient. Furthermore, this finding would also suggest that H2O2 formed by the SO H143A variants was effectively cleared.
In conclusion, in the present study, we (i) demonstrate that mammalian SOs are able to transfer electrons from sulfite to oxygen, but only with reasonable rates in the absence of efficient haem reduction, (ii) found a stoichiometric formation of H2O2 by haem-deleted mammalian SOs, and (iii) show that PEGylation of haem-deleted SO variants preserves their catalytic activity. These findings therefore provide a novel concept in the development of an enzyme replacement therapy for the treatment of MoCD and SOD. The well-established Mocs1 animal model of MoCD can provide a starting point for a proof-of-concept study. Given that MoCD result also in the loss of xanthine oxidase, MoCD mice would probably suffer from kidney failure. Future studies should consider the use of whole-body or organ-specific SO knockouts, which would on one hand enable the study of isolated SO deficiency and on the other hand could serve as a model for the development of the enzyme-replacement therapy. In both models, PEGylated SOMo could be applied by intravenous application and react with oxygen, thus lowering the levels of circulating sulfite. Knowing the toxicity of H2O2, it will be important to ensure low levels of H2O2 by either co-application of catalase or by using a dosing scheme that allows sulfite-dependent non-catalytic clearance of H2O2.
Abdel Ali Belaidi designed and performed experiments, analysed results and wrote the paper. Juliane Röper, Sita Arjune, Sabine Krizowski and Aleksandra Trifunovic performed experiments and analysed results. Guenter Schwarz designed the study, analysed results and wrote the paper.
Technical assistance by Simona Jansen and Joana Stegemann (University of Cologne, Cologne, Germany) and mass spectrometry by Tobias Lamkemeyer (CECAD Cologne) are gratefully acknowledged.
This work was supported by the Center for Molecular Medicine Cologne [CMMC grant D05].
cyclic pyranopterin monophosphate
Food and Drug Administration
human embryonic kidney
human SO molybdenum domain
intramolecular electron transfer
molybdenum cofactor synthesis
murine SO molybdenum domain
Present address: The Florey Institute for Neuroscience and Mental Health, The University of Melbourne, 30 Royal Parade, Parkville, Victoria 3052, Australia.
Abdel A. Belaidi and Guenter Schwarz are co-inventors on a patent application related to the use of PEGylated human sulfite oxidase for enzyme substitution therapy in molybdenum cofactor deficiency.