Functional characterization of a novel arachidonic acid 12S‐lipoxygenase in the halotolerant bacterium Myxococcus fulvus exhibiting complex social living patterns

Abstract Lipoxygenases are lipid peroxidizing enzymes, which frequently occur in higher plants and mammals. These enzymes are also expressed in lower multicellular organisms but here they are not widely distributed. In bacteria, lipoxygenases rarely occur and evaluation of the currently available bacterial genomes suggested that <0.5% of all sequenced bacterial species carry putative lipoxygenase genes. We recently rescreened the public bacterial genome databases for lipoxygenase‐like sequences and identified two novel lipoxygenase isoforms (MF‐LOX1 and MF‐LOX2) in the halotolerant Myxococcus fulvus. Both enzymes share a low degree of amino acid conservation with well‐characterized eukaryotic lipoxygenase isoforms but they involve the catalytically essential iron cluster. Here, we cloned the MF‐LOX1 cDNA, expressed the corresponding enzyme as N‐terminal hexa‐his‐tag fusion protein, purified the recombinant enzyme to electrophoretic homogeneity, and characterized it with respect to its protein‐chemical and enzymatic properties. We found that M. fulvus expresses a catalytically active intracellular lipoxygenase that converts arachidonic acid and other polyunsaturated fatty acids enantioselectively to the corresponding n‐9 hydroperoxy derivatives. The enzyme prefers C20‐ and C22‐polyenoic fatty acids but does not exhibit significant membrane oxygenase activity. The possible biological relevance of MF‐LOX1 will be discussed in the context of the suggested concepts of other bacterial lipoxygenases.

but also play important roles in human pathologies (Ackermann, Hofheinz, Zaiss, & Kronke, 2017;Chen, Sheller, Johnson, & Funk, 1994;Colakoglu, Tuncer, & Banerjee, 2018;Eckl et al., 2009;Haeggstrom & Funk, 2011;Harats et al., 2005;Kronke et al., 2009;Kuhn et al., 2015). In bacteria, LOXs have also been detected but here they occur at much lower frequency. A systematic search for putative LOX genes in the bacterial genomic sequences database (2013) revealed that the 3,700 deposited bacterial genomes (2013) involved 38 putative LOX genes (Hansen et al., 2013). Since certain bacterial species possess more than one LOX gene, it was concluded that <1% all bacterial species carry LOX genes. A more stringent search carried out in 2015 suggested that among the 13,000 bacterial genomes sequenced at this time some 60 species involved putative LOX genes (Horn et al., 2015). Although the vast majority of these potential bacterial LOXs has not been characterized, it was concluded that the presence of these enzymes may not be essential for bacterial life (Horn et al., 2015).
The first bacterial LOX was discovered in 1973 in the opportunistic human pathogen Pseudomonas aeruginosa (Shimahara & Hashixume, 1973). This protein (PA-LOX) was later on characterized as secreted arachidonic acid 15-lipoxygenating enzyme (Banthiya et al., 2016;Garreta et al., 2013;Lu et al., 2013;Vance, Hong, Gronert, Serhan, & Mekalanos, 2004;Vidal-Mas, Busquets, & Manresa, 2005). The crystal structure of PA-LOX was solved at a molecular resolution of 1.4 Å (Banthiya et al., 2016;Garreta et al., 2013) and it differed from other pro-and eukaryotic LOX isoforms in two major aspects: (a) While the polypeptide chains of most eukaryotic LOX are folded into a two-domain structure consisting of small N-terminal βbarrel domain and large helical catalytic domain (Choi, Chon, Kim, & Shin, 2008;Eek et al., 2012;Gilbert et al., 2011;Gillmor, Villasenor, Fletterick, Sigal, & Browner, 1997;Minor et al., 1996;Neau et al., 2009), the PA-LOX polypeptide folds into a single domain structure (Banthiya et al., 2016;Garreta et al., 2013). (b) Recombinant PA-LOX involves a bifurcated substrate-binding pocket consisting of two hydrophobic cavities and a joining lobby. These internal cavities harbor a phosphatidylethanolamine molecule. The subcavity containing the sn1 fatty acid of the endogenous ligand involves the catalytic nonheme iron (Banthiya et al., 2016;Garreta et al., 2013). More recently, the crystal structure of a LOX isoforms from Cyanothece sp. PCC 8801 was also solved (Newie et al., 2016). The biological activities of bacterial LOX have not been explored in detail. PA-LOX has been implicated in pathogen-host interaction  and in biofilm formation (Deschamps et al., 2016). More recently, the enzyme has been suggested as pathogenicity factor because of its capability of oxidizing membrane lipids of eukaryotic cells (Aldrovandi et al., 2018). A similar membrane lipid oxygenase activity has been suggested for a LOX isoforms from Cyanothece sp. PCC 8801 (Newie et al., 2016).
We recently rescreened the NCBI bacterial genome database and identified previously described LOX sequences in Myxococcus xanthus (WP_011551853.1, WP_011551854.1). The protein, which is encoded by the WP_011551853.1 gene, was identified as acidic LOX isoforms (Qian et al., 2017) and recombinant expression of the WP_011551854.1 gene also led to a catalytically active enzyme (An, Hong, & Oh, 2018). In addition, our database search identified two putative LOX genes in the genome of M. fulvus (WP_046712474.1 and SEU34910.1), which have not been characterized so far. To explore whether the WP_046712474.1 gene encodes for a functional LOX, we expressed the corresponding enzyme in different pro-and eukaryotic expression systems and characterized the recombinant protein with respect to its protein-chemical and enzymatic properties. Our results indicate that M. fulvus expresses an arachidonic acid 12S-lipoxygenating LOX isoform (MF-LOX1), which only shares a low degree (20%) of amino acid identity with a recently characterized LOX isoform from M. xanthus  and with other pro-and eukaryotic LOX isoforms.

| Database search and identification of putative LOX genes in the genome of M. fulvus
Lipoxygenases (ALOX isoforms) rarely occur in prokaryotes but in M. xanthus two functional ALOX genes (WP_011551854.1 and WP_011551853.1) have recently been identified Qian et al., 2017). In Myxococcus fulvus (M. fulvus), which differs from M. xanthus with respect to structural and functional characteristics, no LOX isoforms have been described so far. When we searched the NCBI bacterial genome database for potential ALOX sequences, we detected two potential ALOX sequences

| Recombinant expression of MF-LOX1 in E. coli
To explore whether the MF-LOX1 gene (WP_046712474.1) encode for a functional ALOX isoform, we expressed the corresponding enzyme as N-terminal his-tag fusion protein in Escherichia coli, purified the recombinant protein by affinity chromatography on Ni-agarose, and characterized it with respect to its protein-chemical and enzymatic properties. When E. coli cells were transformed with the recombinant expression plasmid, they express a his-tag fusion protein, which migrates in SDS-PAGE in the molecular weight range of 80 kDa (Figure 2a). No immunoreactive protein was observed when bacteria were transformed with an empty plasmid. To confirm the expression of a functional enzyme ALOX, activity assays were carried out ( Figure 2b). The RP-HPLC chromatograms and the UV spectrum of the major oxygenation product indicate the formation of a conjugated diene during the incubation period and this product comigrated with an authentic standard of 12-HETE (lower trace). This compound was not detected in control incubations (no enzyme). Since 12-and 8-HETE are not well separated under our chromatographic conditions, additional SP-HPLC was carried out to resolve the two product isomers. Here, the major oxygenation product comigrated with 12-HETE (data not shown). For more comprehensive characterization, we purified the recombinant enzyme by affinity chromatography on a Ni-agarose F I G U R E 1 Dual amino acid alignment of Myxococcus fulvus LOX1 (MF-LOX1, WP_046712474.1) and M. xanthus LOX2 (WP011551854.1). The putative iron ligands are framed in green, and the sequence determinants of the reaction specificity are color coded as follows: yellow-Coffa determinant, blue-Borngraber 1 determinant, red-Sloane determinants column. As indicated in Figure 2c, the his-tag fusion protein was bound at the affinity matrix and was recovered by eluting the column with increasing amounts of imidazole. The bulk of the recombinant protein was eluted in fractions 2, 3, and 4. To quantify the degree of purity of the final enzyme preparation, Coomassie staining of an SDS-PAGE was carried out and densitometric evaluation suggested that the enzyme was >95% pure (Figure 2d). The expression yield, the degree of purity, and the catalytic activities of the final enzyme preparations are summarized in Table 1.

| Recombinant expression of MF-LOX1 in a eukaryotic expression system
Evaluation of our activity assays suggested that the specific activity of the expressed MF-LOX1 was considerably lower than that of P. aeruginosa (PA-LOX) enzyme used in control incubations. To test whether the enzyme is more efficiently expressed in eukaryotic overexpression systems, we cloned the coding sequence into the pFastBac HT-B expression vector and expressed the enzyme F I G U R E 2 Bacterial expression and purification of recombinant Myxococcus fulvus LOX1 (MF-LOX1). MF-LOX1 was expressed as N-terminal hexa-his-tag fusion protein in Escherichia coli. (a) Immunoblotting: Bacteria were lyzed by sonication and cell debris was removed. Aliquots of the bacterial lysis supernatant were applied to SDS-PAGE, the separated proteins were transferred to a nitrocellulose membrane, and the membrane was probed with an anti-histag antibody. in Sf9 cells. Western blot analyses indicated that the recombinant enzyme is expressed ( Figure 3a) and activity assays (Figure 3c) confirmed this conclusion. Here again, conjugated dienes that cochromatographed in RP-HPLC with an authentic standard of 12S-HETE were formed during the incubation period but these products were absent in control incubations. The enzyme could also be purified by affinity chromatography on Ni-agarose ( Figure 3b) and the final enzyme preparations exhibited a high (>95%) degree of electrophoretic homogeneity. The expression yield, the degree of purity, and the catalytic activities of the final enzyme preparations are summarized in

| Protein-chemical characterization of MF-LOX1
The theoretical molecular weight calculated from the amino acid sequence of our MF-LOX1 construct including the hexa-his tag and the linker peptide was 79.800 Da. From SDS-PAGE of the purified enzyme, a molecular weight of 90.3 kDa was concluded ( Figure 4b). Isoelectric focusing ( Figure 4a) indicated several protein bands migrating in the pH region between 5.1 and 6.5 and these data suggest a structural microheterogeneity of our final enzyme preparation. Nevertheless, an IP in the pH region between 6.1 and 6.5 is consistent with the theoretical IP of 5.8, which was All LOX isoforms characterized so far involve either iron or manganese as catalytically active transition metal (Ivanov et al., 2010;Wennman, Karkehabadi, & Oliw, 2014). To explore whether MF-LOX1 carries manganese or iron as catalytically active constituent, we quantified the content of these transition metals in our purified enzyme preparation by atomic absorption spectroscopy. For this purpose, the enzyme preparation was desalted and aliquots of the desalting buffer were used for control measurements. We found that the manganese content of the enzyme preparation was lower than in the desalting buffer and these data suggest that MF-LOX1 does not involve manganese. Quantification of the iron content yielded significantly higher iron levels than in the desalting buffer (Table 2), but calculation of the iron load suggested that only 5.6% of wildtype MF-LOX1 expressed in E. coli carried an iron ion. Unfortunately, all attempts to improve the iron load (iron supplementation of the fermentation sample, mutagenesis studies of the iron ligands, in vitro iron incorporation into the purified protein after incorporation experiments) were not successful. To exclude the possibility that MF-LOX1 involves other transition metals as catalytically essential constituent, we next determined the copper and zinc concentrations in our enzyme preparation. However, both transition metals were below the detection limits. When we quantified the iron content of the Phe424Ile+Ile425Met double mutant, we observed a fourfold higher iron load. This observation is quite interesting since this double mutant exhibited a 4.7-fold higher specific activity (Table 4).
Thus, the major reason for the low catalytic activity of recombinant MF-LOX1 (Table 1) is its low iron load. The mechanistic basis for the unusually low iron load of MF-LOX1 remains unclear but possible scenarios are discussed later on in the manuscript.

| Enzymatic characterization of MF-LOX1
To explore the substrate specificity of recombinant MF-LOX1, we tested several omega-6 (linoleic acid, gamma-linolenic acid, arachidonic acid) and omega-3 (alpha-linolenic acid, eicosapentaenoic acid, docosahexaenoic acid) polyenoic fatty acids as substrate. Here, we found that MF-LOX1 most effectively oxygenated eicosapentaenoic acid ( Figure 5e). Docosahexaenoic acid, arachidonic acid, and alpha-linolenic were also well oxygenated. In contrast, no oxygenation products were detected when gamma-linolenic acid and linoleic acid were The reaction products were analyzed by RP-HPLC recording the absorbance at 235 nm. Authentic standards of 12-HEPE, 14-HDHA, and 12-HETE were prepared using recombinant human ALOX12 (Kutzner et al., 2017). (a) Product pattern of EPA oxygenation, (b) product pattern of DHA oxygenation, (c) product pattern of AA oxygenation, (d) product pattern of ALA oxygenation, (e) substrate specificity: The conjugated dienes formed during the incubation period were quantified and the product formation from EPA was set 100%. (f) Analysis of the enantiomer composition of the major oxygenation products formed by MF-LOX1 from C20 and C22-polyenoic fatty acids. The enantiomer composition was determined by chiral phase LC-MS (see Experimental Procedures). (g) Analysis of 12-HETE enantiomer composition formed by the Ala410Gly mutant used as substrate. To identify the chemical structure of the reaction products, we compared in RP-HPLC the reaction products formed by MF-LOX1 with those of human recombinant ALOX15. Here, we found that the major oxygenation product of EPA oxygenation (Figure 5a) comigrated with the minor oxygenation product formed from this fatty acid by the human enzyme, which is 12-HEPA (Kutzner et al., 2017). Similarly, the major oxygenation product formed from DHA and AA by MF-LOX1 (Figure 5b,c) cochromatographed in RP-HPLC with the minor oxygenation products formed from these substrates (14-HDHA, 12-HETE) by the human enzyme (Kutzner et al., 2017).
TA B L E 3 Enantiomer composition of major conjugated dienes formed from EPA (12-HEPE), DHA (14-HDHA), and AA (12-HETE) To define the degree of optical purity of the major oxygenation products, we carried out chiral phase LC-MS and the corresponding chromatograms are shown in Figure 5f. For this purpose, the chromatograms were followed at m/z 319 (12-HETE), 317 (12-HEPE), and 343 (14-HDHA). It can be seen that for all of these major oxygenation products, the S-isomer prevails and that the corresponding R-enantiomers are only formed in small amounts. Exact quantification of the S/R ratio is given in Table 3. Taken together, these data indicate a high degree of stereo-chemical control of the oxygenase reaction.
To obtain more detailed information on the reaction kinetic of MF-LOX1, we quantified its substrate affinity using eicosapentaenoic acid as model substrate. From Figure 6a, it can be seen that (eicosapentaenoic acid as substrate), a molecular turnover rate of (3.1 ± 1.2) x 10 −2 /s was determined. This value is more than two orders of magnitude lower than the turnover rate determined for linoleic acid oxygenation by human ALOX15 (Ivanov, Kuhn, & Heydeck, 2015) and four orders of magnitude lower than that of the PA-LOX (Banthiya et al., 2016). The molecular basis for the low specific activity of MF-LOX1 remains unclear but the low iron load of the recombinant enzyme may contribute.
On the other hand, wild-type P. aeruginosa LOX has a Km for ox- trations. In other words, MF-LOX1 exhibits a high oxygen affinity and this conclusion is consistent with experimental data obtained for different mammalian ALOX isoforms (Juranek et al., 1999).
LOXs exhibit different pH profiles but most isoforms have neutral or alkaline pH optima. Recently, an acidic LOX isoforms was identified in M. xanthus (WP_011551853.1), which showed a pH optimum of 3 (Qian et al., 2017). This enzyme shares a high degree (86%) of amino acid identity MF-LOX1 but it prefers linoleic acid over arachidonic acid. When we recorded the pH profile MF-LOX1 (Figure 6c), we observed the pH optimum at 9.5.
Finally, we determined the temperature dependence of EPA oxygenation by MF-LOX1 (Figure 6d). Here, we observed similar catalytic activities in the temperature range between 5°C and 15°C.
When we increased the reaction temperatures above 15°C, a steady decline of the catalytic activity was observed. The molecular basis for this unusual temperature dependence has not been explored but it may be related to a limited thermostability of the enzyme.

| Mutagenesis of catalytically important amino acids
For eukaryotic ALOX isoforms, several hypotheses explaining the mechanistic basis of their reaction specificities have been developed. The Triad Concept (Borngraber et al., 1999;Ivanov et al., 2015;Vogel et al., 2010), which was developed for mammalian ALOX15 cies, which exhibited an almost fivefold higher catalytic activity than the wild-type enzyme. However, the reaction specificity of this gainof-function mutant was identical to that of the wild-type enzyme (Table 4). These data suggest that the Triad Concept (Ivanov et al., 2015) might not be applicable for MF-LOX1.
Previous multiple sequence alignments have indicated that most S-LOXs carry an Ala at a critical position of their primary structures.
In contrast, most R-LOXs involve a smaller Gly at this position and Ala-to-Gly exchange increase the relative share of R-lipoxygenation products (Coffa & Brash, 2004;Coffa, Schneider, & Brash, 2005;Vogel et al., 2010). Although this concept was not applicable for   Brash, 1990). In Table 5, it can be seen that under our in vitro conditions the membrane oxygenase activity of rabbit ALOX15 was threefold higher than that of MF-LOX1. This difference was even more pronounced when we normalized the membrane oxygenase activity to the amounts of enzyme added as catalysts. For MF-LOX, we applied 7 mg/ml of purified enzyme but for rabbit ALOX15 7 µg/ ml. It should, however, been stressed that the specific fatty acid oxygenase activity of the rabbit ALOX15 is orders of magnitude higher than that of MF-LOX1 (Table 1).

| Occurrence of LOX isoforms in different myxobacterial species
In bacteria, LOXs rarely occur (Horn et al., 2015)  MF-LOX1 was expressed as N-terminal hexa-his-tag fusion protein in Escherichia coli and purified by affinity chromatography on Ni-agarose as described in Experimental Procedures. The sequence determinants of MF-LOX1 were identified by amino acid sequence alignment (Figure 1). Aliquots of the pooled Ni-agarose fractions (Figure 2d) were used for activity assays and the reaction products of arachidonic acid oxygenation were quantified by RP-HPLC (see Experimental Procedures). Note. Purified recombinant MF-LOX1 (7 mg/ml) and pure native rabbit ALOX15 (7 µg/ml) were incubated in PBS with beef heart submitochondrial membranes (1.2 mg/ml) for 15 min at room temperature. After the incubation period, the reaction products were reduced with NaBH4 and the pH was adjusted to 3.5 with acetic acid. Total lipids were extracted (Bligh & Dyer, 1959)  . In contrast, the enzyme appears to be more closely related to the poorly characterized M. xanthus LOX (Qian et al., 2017) since the amino acid identity score was close to 86% (Table 6). Similar identity scores were found when mouse (74%), rat (75%), and pig (86%) ALOX15 orthologs were compared with the human enzyme. Thus, it might well be that the M. fulvus LOX characterized here (WP_046712474.1, MF-LOX1) may constitute the functional equivalent of MX-LOX1 (WP_011551853.1), which has not been characterized very well (Qian et al., 2017).
However, the currently available functional data do not provide major evidence for such a close functional relation: (a) MX-LOX1 has been described as acidic LOX with a pH optimum of 3.0 (Qian et al., 2017). In contrast, we detected an alkaline pH optimum (pH opt of 9.5) for MF-LOX1 (Figure 6c). (b) MX-LOX1 prefers linoleic acid over arachidonic acid (Qian et al., 2017). In contrast, we found that linoleic acid is a poor substrate for MF-LOX1 when compared with arachidonic acid, eicosapentaenoic acid, and docosahexaenoic acid ( Figure 5e). (c) MX-LOX1 is rather stable and exhibits a temperature optimum of 30°C (Qian et al., 2017). In contrast, MF-LOX1 is unstable and exhibits an unusual temperature dependence with a T opt at 10°C. Although these kinetic parameters suggest that MX-LOX1 and MF-LOX1 may not be closely related, they still might constitute functional equivalents in the two different Myxococcus species. More detailed functional characterization of MX-LOX1 is needed to draw more definite conclusions.

| Low iron content and limited catalytic activity
Transition metal analyses of our final enzyme preparation indicated that recombinant MF-LOX1 does neither involve manganese nor copper and zinc as catalytically active transition metal. Unfortunately, we also found that the iron content was rather low so that an iron load of only 5.6% was calculated for the wild-type enzyme. This low iron saturation might be discussed as molecular basis for the low catalytic turnover rate ([3.1 ± 1.2] x 10 −2 s −1 ). If one calculates the putative catalytic activity for an enzyme preparation with 100% iron load, a molecular turnover rate of about 0.55 ± 0.21 s −1 results.
This value is still lower than the turnover rates determined for rabbit and human ALOX15 (Ivanov et al., 2015;Kühn et al., 1993), soybean LOX1 (Egmond et al., 1976;Maccarrone et al., 2001), and P. aeruginosa LOX (Banthiya et al., 2016;Garreta et al., 2013). The molecular basis for the low iron affinity of MF-LOX1 has not been explored in detail. However, iron supplementation of the fermentation sample, which was successful for the Cyanothece sp. LOX (Andreou, Gobel, Hamberg, & Feussner, 2010a), did neither improve the iron load nor the catalytic activity of MF-LOX1. Moreover, parallel expression of P. aeruginosa LOX led to a recombinant protein exhibiting an iron load of 100% (Banthiya et al., 2016). Since the expression levels of the two proteins were comparable, one can exclude that problems with the iron incorporating machinery are major reasons for the low efficiency of iron incorporation for MF-LOX1. These data rather suggest that MF-LOX1 protein has a low iron affinity and this property might be related to its 3D-structure. It might be possible that regular  Note. For the time being, two different LOX isoforms Qian et al., 2017) have been described in Myxococcus xanthus (MX-LOX1-WP_011551853.1, MX-LOX2-WP_011551854.1). When we screened the currently available bacterial genomes for LOX-like sequences, we detected two potential LOX genes in M. fulvus. We carried out dual amino acid alignments and observed variable degrees of amino acid conservation between the different myxobacterial LOX isoforms particular region of the primary structure are not very pronounced, alignment artifacts might be possible. On the other hand, both iron ligand clusters of MF-LOX1 align well with the corresponding amino acid of human and rabbit ALOX15, which makes alignment artifacts unlikely.

| Biological activity of bacterial LOX
Mammalian ALOX isoforms have been implicated in cell differentiation but also in the pathogenesis of various diseases (Haeggstrom & Funk, 2011;Kuhn et al., 2015). Unfortunately, much less is known on the biological relevance of bacterial LOXs but several scenarios have Since such kinetic properties are characteristic for sensor proteins (Berra et al., 2003), PA-LOX might be involved in oxygen sensing.
However, MF-LOX1 exhibits a rather high oxygen affinity, and thus, its suitability to function as oxygen sensor is limited. (e) PUFA toxicity: Unsaturated fatty acids are toxic for many bacteria (Greenway & Dyke, 1979;Raychowdhury, Goswami, & Chakrabarti, 1985). Oleic acid is toxic for Streptococcus pyogenes M49 but this bacterium expresses a fatty acid double bond hydratase to metabolize this toxin (Volkov et al., 2010). Although recombinant PA-LOX does not oxygenate oleic acid, it might contribute to detoxification of other unsaturated fatty acids.

| Chemicals
The chemicals were obtained from the following sources: arachidonic acid, linoleic acid, alpha-linolenic acid, gamma-linolenic acid, eicosa-

| Bacterial expression and purification of MF-LOX1
Genomic

| Eukaryotic expression of MF-LOX in the baculovirus-insect cell system
To improve the expression yield and the specific activity of the recombinant MF-LOX, we excised the coding region of the recombinant bacterial expression plasmid using the restriction enzymes

| Site-directed mutagenesis
Site-directed mutagenesis was performed using the Pfu Ultra II Hotstart 2XPCR Mastermix (Agilent Technologies, California, USA) as described before (Banthiya et al., 2016). After PCR aliquots of the reactions mixture were transformed into E. coli XL1-Blue competent cells (Thermo Fisher) and plated onto kanamycin agar plates.
Five clones were picked and plasmid DNA was prepared using NucleoSpin Plasmid Easy Pure (Macherey-Nagel). Nucleotide sequencing (Eurofins Genomics, Ebersberg, Germany) confirmed the sequences of the mutant plasmid clones.

| Fatty acid oxygenase activity assays
Fatty acid oxygenase activity of wild-type and mutant MF-LOX1 was determined by HPLC quantification of the reaction products formed during a 3-min incubation period. For this purpose, aliquots of the MF-LOX1 preparation were incubated in 0.5 ml of PBS containing different concentrations of fatty acids as substrates. After the incubation period, the hydroperoxy compounds formed were reduced with sodium borohydride and after acidification 0.5 ml of ice-cold methanol was added. The protein precipitate was spun down and aliquots of the clear supernatant were injected to RP-HPLC for quantification of the oxygenation products.

| HPLC analytics
HPLC analysis of the reaction products was performed on a Shimadzu HPLC system. Reverse phase-HPLC (RP-HPLC) was carried out on a Nucleodur C18 Gravity column (Macherey-Nagel; 250 x 4 mm, 5 μm particle size) coupled with a guard column (8 x 4 mm, 5 μm particle size). A solvent system of methanol/water/acetic acid (

| Activity measurements under normoxic and hyperoxic conditions
To judge the oxygen affinity of MF-LOX1, we carried out activity assays, in which the oxygen concentration in the reaction mixture was altered. For this purpose, variable volumes of oxygen saturated PBS (hyperoxic) were mixed with argon saturated PBS (anoxic). To obtain the hyperoxic solution, 50 ml PBS was flushed for 3 hr with pure oxygen. Similarly, 50 ml of PBS was flushed with argon to prepare the anoxic medium. Next, a photometric cuvette was filled with argon gas and aliquots of anaerobic PBS (0-0.7 ml) were added under argon atmosphere. The cuvette was closed with a plastic stopper containing two capillary wholes to add additives. Then, different aliquots of hyperoxic PBS (oxygen saturated) were added so that a final reaction volume of 0.7 ml was reached. Next, 10 μL of a partly anaerobized methanolic solution of eicosapentaenoic acid was added and the reaction was started with 50 μL of partially anaerobized enzyme solution.

| Membrane oxygenase activity assay
To quantify the membrane oxygenase activity of MF-LOX1, aliquots of the enzyme preparations were incubated for 15 min in PBS with beef heart submitochondrial particles (1.2 mg membrane protein/ ml), which constitute inside-out vesicles of inner mitochondrial membranes. After the incubation period, the reaction was stopped by the addition of sodium borohydride. Following acidification, the total lipids were extracted from the reaction mixtures (Bligh & Dyer, 1959), the solvent was evaporated, and the remaining ester lipids were reconstituted in 1 ml of a 1:1 mixture of chloroform and metha-  (Aldrovandi et al., 2018;Kuhn et al., 1990).

| Determination of the iron content and isoelectric focusing
The iron content of the purified MF-LOX1 was determined by atom absorbance spectroscopy on a Perkin-Elmer Life Sciences AA800 instrument equipped with an AS800 autosampler. For calculating the iron load of the enzyme, the iron concentration in the enzyme preparation was related to the protein content.
To determine the native isoelectric point of MF-LOX1, isoelectric focusing (IF) was carried out. For this purpose, a precasted IF gel (Bio-Rad, Munich, Germany) was employed and isoelectric focusing was run for 2.5 hr on a Bromma LKB 2197 high-voltage power supply electrophoresis system. Protein bands were stained with Coomassie blue and the following IF standards were used (phycocyanin, IP 4.45-4.75; β-lactoglobulin B, IP 5.1; bovine carbonic anhydrase, IP 6.0; human carbonic anhydrase, IP 6.5; equine myoglobin, IP 6.8/7.0; human hemoglobin A, IP 7.1; human hemoglobin C IP 7.5; lentil lectin IP 7.8/8.0/8.2; cytochrome c IP 9.6).

ACK N OWLED G EM ENTS
The authors would like to thank Dr. J. Zentek (Free University Berlin) for quantification of the transition metals. This work was supported by research grants of Deutsche Forschungsgemeinschaft to H.K. (Ku961/11-1, Ku961-12/1). We also acknowledge the support from the German Research Foundation (DFG) and the Open Access Publication Fund of Charité -Universitätsmedizin Berlin.

CO N FLI C T O F I NTE R E S T
The authors declare that they do not have any conflicts of interest with the content of this article.

E TH I C S S TATEM ENT
This paper does not involve any experiments with animals or humans, and thus, there is no ethical issue related to the content of this paper.

DATA ACCE SS I B I LIT Y
The authors declare that the experimental data published in this paper are made accessible upon request for interested readers.