Production of non-cytotoxic pyomelanin by a laccase: properties and chemical structure compared to bacterial and synthetic pigments

197 words Introduction → Conclusion : 4950 words Methods : 2130 words References number : 64


Introduction
Pyomelanin is a natural polymer of homogentisic acid (HGA, 2,5-dihydroxyphenylacetic acid) synthesized through the L-tyrosine pathway by bacteria, fungi, mammals and plants, and belongs to the heterogeneous group of allomelanins 1 . In living cells, the HGA 1,2-dioxygenase disruption or deletion leads to the pyomelanin accumulation 2 (Fig. 1S). The first property of the pigment is to protect microorganisms from UV light limiting free radicals and ROS generation 3 . The excitation of L-Dopa melanin by UV light produces cell-damaging ROS 4 , whereas the formation of ROS by light has never been reported with pyomelanin. As a consequence, pyomelanin increases resistance to light, for instance in Legionella 5 . Its antioxidant role has also been demonstrated in Pseudomonas aeruginosa 6 , Burkholderia cenocepacia 7 , Aspergillus fumigatus 2 , and many other strains. Pyomelanin also diminishes the oxidizing stress of the host microorganism, by its high tolerance to H2O2, as demonstrated in Ralstonia solanacearum 8 and clinical isolates of Pseudomonas aeruginosa from infected patients 9 . Moreover, pyomelanin is an antibacterial and antifungal agent against microbial outside attacks, especially pathogenic 10 , reduces biofouling 11 , chelates heavy metals 3,12 , and contributes to microbial pathogenesis as it is associated with virulence in a broad range of pathogenic fungi and bacteria 9,13 . Pyomelanin potentially reduces soluble FeIII in FeII 14,15,16 , FeII being essential in many bacteria such as Legionella pneumophila, that ensure homeostasis by an appropriate Fe 2+ /Fe 3+ ratio for their survival 17,18 . Consequently, in vivo or in vitro, pyomelanin may serve as a terminal electron acceptor, electron shuttle, or conduit for electrons, a complementary iron acquisition to the siderophore role. Several ways for pyomelanin synthesis may be explored. In microorganisms, two distinct pathways lead to HGA formation, route 1 through a 4-HPP dioxygenase (4-HPPD; EC 1.13.11.27), the most described from many bacteria mostly pathogen, and route 2 via a 4-HPA-1-hydroxylase (4-HPAH-1; EC 1.14.13.18) (Fig. 1S). To date, there are very few data on microbial pyomelanin production and yield reminds weak (max. 0.35 g/L 19 ), comparatively to the eumelanin pigment produced at 28.8 g/L by a tyrosinase overexpression in a recombinant Streptomyces kathirae 20 . The construction of a recombinant 4-HPPD enzyme is not an option because the substrate 4-HPP is expensive. Recently, as opposed to the plant enzyme-based assay, an optimized high-throughput screening assay using human 4-HPPD was constructed using the E. coli strain C43 (DE3) supplemented with L-Tyr in the culture medium, a useful tool to find new inhibitors against alkaptonuria disease 21 . However, no pyomelanin yield was reported. On the other hand, the 4-HPAH-1 enzyme responsible for HGA synthesis in Delftia acidovorans 22 and Azoarcus evansii 23 , had been partially characterized. In 2008, the sequence of the D. acidovorans enzyme had been given by several genomic approaches 24 . The enzyme contains two components, hpaH which codes for a flavoprotein NAD(H)-dependent oxidase that transforms 4-HPA in a non-identified metabolite called Z, the second hpaC catalyzes the conversion of Z in HGA. hpaH and hpaC were cloned together, however no HGA synthesis occurred from these constructions. To date, HGA is very expensive, at about $800/gram. Several methods of chemical synthesis have been developed but the majority not applicable for a large-scale process, the yield being comprised of 50-60% when indicated. Ultimately, the former procedure of DeForrest Abbott and Doyle-Smyth 25 through HGA-lactone as intermediate, remained the most convenient and proposed in two or three simple steps. For a long time, pyomelanin had been described as the result of the HGA autoxidation catalyzed by Mn 2+ or Cu 2+ , from neutral to alkaline conditions 26 . For 30 years, several authors hypothesized the action of oxidases, such as the multicopper-dependent laccases (EC 1.10.3.2) generally involved in the polymerization of dihydroxylated phenols 27 . Laccases have also been suggested in that of HGA in Vibrio cholerae 28 , Alcaligenes eutrophus (now Cupriavidus necator 29 ), Metarhizium anisopliae 30 , and their role definitively confirmed in Cryptococcus neoformans 31 . At this time, no marketed pyomelanin exists, and a method to furnish acceptable quantities of pigment became necessary. We have therefore developed a large-scale and low-cost production process by reconsidering the chemical synthesis of HGA on the one hand, and the use of laccase on the other. From this optimized enzymatic process, the resulting pigment (PyoENZ) and its chemical structure and properties had been examined. For comparison, bacterial pyomelanin (PyoBACT) from an induced Halomonas culture, the bacterial process, and that issued from the HGA autoxidation (PyoCHEM), the chemical process, were also extensively characterized. Furthermore, the mechanism of HGA polymerization was re-evaluated with a lot of care.

Results
Development of an enzymatic process (PyoENZ). To produce high quantities of pyomelanin, appropriate chemical synthesis of HGA has been combined with a polymerization step by a selected laccase in an on-line process (steps 1 to 5, Fig. 1). The first part of the synthesis of DeForrest Abbott and Doyle-Smyth 25 yielded pure HGA-lactone (steps 1 and 2, ~99% yield) as controlled by HPLC and GC-MS (see spectroscopic data in Methods). The alkaline hydrolysis of the lactone was followed by immediate buffering at pH 6.8 (step 3) and the addition of the rMt laccase that reacted in two successive stages, the formation of BQA and the polymerization (step 4). In the final step 5, the polymer was precipitated, washed, and dried. To lighten the process, step 1 could be omitted, giving preference to 2,5-DMPA as the starting molecule, and this despite its cost four times as much as that of 2,5-DMPAO according to various suppliers. Finally, the optimal yield was reached with 40 mM HGA (final conc.) or HGA-lactone equivalent, 17-22 U (~ 23-30 µL) of rMt laccase per mL of reaction volume, at pH 6.8 for 24 h agitation ( Fig. 2S and 3S). A temperature of 50°C for the polymerization did not improve the yield neither reduced the incubation time (Fig. 4S). In these conditions, at 30°C, the process yielded 1.25 ± 0.11 g of dried PyoENZ per g of 2,5-DMPA (mean of 3 experiments).

Production of pyomelanin (PyoBACT) by an induced wild Halomonas strain.
A bacterial strain able to compete with the enzymatic process (PyoENZ) in terms of production yield, was sought. The strategy consisted to select a Halomonas species among a large collection, similar to our previous studies 32,33,34 . Phenolic compounds in the medium were identified and controlled along with the growth of the induced cultures, the strains preferentially utilized the aromatic over glucose. These halophile bacteria easily grow and have been shown to produce dihydroxy phenols from 4-hydroxyphenylacetic acid (4-HPA), such as HGA in H. olivaria 32,34 , H. venusta, H alkaliphila, and 3,4-dihydroxyphenylacetic acid (3, in H. alkaliantartica, H. neptunia, H. sulfaedris (this work), and H. sp. HTB24 35 , through routes 2b and 3, respectively (Fig. 1S). H. titanicae is a -proteobacterium suggested to accelerate the destruction of the RMS Titanic, its genome now entirely available 36 , and has been selected for the most intense brownblack color from 5 mM 4-HPA supplemented cultures measured by the A400 nm (this work). The suspected presence of pyomelanin was first confirmed by identification of HGA (max 290 nm) only in the exponential phase, but not 3,4-DHPA or other dihydroxyphenylacetic derived compounds, by RP-HPLC-DAD and GC-MS of the TMS-derived metabolites and showing a matched fragmentation spectrum with that of the HGA standard. The strain could not grow in the presence of L-Tyr and was unable to metabolize 2-HPA or 3-HPA, hence suggesting that a 4-HPA-5-hydroxylase or a 4-HPA-6-hydroxylase were not implied, respectively. We concluded that H. titanicae was able to produce pyomelanin by direct conversion of 4-HPA to HGA through a 4-HPA-1-hydroxylase (4-HPAH-1, route 2b, Fig. 1S). Following this, pyomelanin production has been optimized. Because 5 mM 4-HPA was rapidly consumed in 2 days and served as an inducer of 4-HPAH enzymes 34,35 , successive additions of well-defined amounts of 4-HPA at different culture times were carried out by following an experimental design procedure (see Methods). Finally, pyomelanin was overproduced in a 500 mL medium by adding 5 mM 4-HPA at starting, then 10 mM after 3 days, in a total culture time of 6 days. In these conditions, H. titanicae was able to furnish 0.55 ± 0.09 g PyoBACT per Liter of culture, a mean of three independent experiments. Relative to the total amount of 4-HPA added, the recalculated yield was 0.241 ± 0.04 g PyoBACT per g of 4-HPA.
Production by chemical autoxidation (PyoCHEM). While pyomelanin issued from the HGA autoxidation has been commonly used 2,28 , the reaction has never been optimized to date. In presence of the transition metal Mn 2+ to enhance the catalysis 26 , an (HGA)/(Mn 2+ ) ratio of 20 for an optimal pyomelanin yield was obtained (see Methods). By applying this ratio, the developed on-line process (Fig. 1, steps 2-3-4'-5) provided 0.317 ± 0.031 g PyoCHEM per g of 2,5-DMPA (mean of 5 experiments), a yield four times lower than that of the enzymatic (PyoENZ) and higher than the bacterial (PyoBACT) process (summarized in Table 2). Despite this, the production of PyoCHEM remains interesting because the cheapest and easy to implement for use on a laboratory scale.
Structural data of the three pyomelanin. Better than 1 H-NMR, solid-state 13 C-NMR analyses of polymers can provide not only structural features through the resonances of the monomers but also the types of bindings. Pyomelanin issued from the three developed processes, PyoENZ, PyoBACT, and PyoCHEM were analyzed by solid-state 13 C CP-MAS at their optimal signal resolution in conjunction with FTIR experiments. Chemical shifts from the NMR spectra (cumulated in Fig. 2) were summarized in Table 1, the date included those of pure HGA analyzed in the same conditions. The three spectra exhibited common, typical, and prominent signals in slightly varied positions, at  172-173.4 ppm that corresponded to the unprotonated carbon in C-O/C=O of the carboxylic group, then at 149.3-149.4 of the unprotonated carbon (suggested C5) of the ring bearing the -OH group, with shoulders at 143.4 for the three pyomelanin (suggested C2), and at 118-119 ppm provided by the ethylenic and protonated carbons of the ring (-CH=C-) together in broadband. Less high signals at 33.0-34.8 ppm observed on the three structures represented the saturated aliphatic carbons (-CH2-) of the acetic acid moiety. PyoBACT and PyoENZ exhibited very similar FTIR and 13 C-NMR spectra, hence the study was focused on the PyoENZ and PyoCHEM absorptions (Fig. 5S) noting that the spectrum of PyoCHEM was better resolved in reason to its less high Mw (2,300 Da, Table 2). The peaks at the following wavenumbers and their corresponding structures included the bands for PyoENZ and PyoCHEM, respectively at (i) 3,401 and 3,278 cm -1 (broad) indicative of the -OH stretch of polymeric structures; (ii) two smaller bands for each compound at 2,960 (PyoENZ) and much more intense at 2,925-2,927 cm -1 (PyoCHEM), which corresponded to stretching vibrations of the aliphatic Car-H groups; (iii) 1,711 and 1,720 cm -1 quite resolved here and ascribed to carbonyl stretching (C=O) of the -COOH group, these bands were however absent on other microbial pyomelanin 29,37 ; (iv) 1,623 and 1,656 cm -1 absorptions that were described as typical for aromatic C=C conjugated with C=O groups (quinones), with a stronger response in PyoCHEM; (v) 1,384 and 1,385 cm -1 (both minor) would be assigned to the O-H bond of the hydroxyl groups attached to the ring; and (vi) strong bands at 1,197 (PyoENZ) and 1,222 cm -1 (PyoCHEM) of the phenolic-OH links. Especially, a reaction of substitution on the C4 position of the BQA ring by primary and secondary amines had been reported 38 , such substitutions might occur in biological systems where amino acids are present, like during PyoBACT and PyoENZ preparation. Here, the volume of the laccase extract added for the PyoENZ synthesis seemed insignificant, hence it remained difficult to look for amide or amine bonds from polymers, especially when they are minor. These C-N absorptions were generally encountered at  155-180 ppm (amides formed from the carboxylic moiety) and 135-145 ppm (aromatic amines) in 13 C NMR, mainly at 3,000-3,500 cm -1 (N-H stretching vibrations of aromatic amines) in FTIR, thus drowned in those of the major functional groups. Faced with this inability to detect traces of nitrogenous derivatives by NMR and FTIR, elemental analyses of the three pyomelanin were carried out and showed the presence of N in PyoENZ (2.75%) and PyoBACT (3.65%), as expected none in PyoCHEM, and higher in the indole-based melanin MelSYNTH and MelSEPIA ( Table 2). The presence of N in PyoENZ and PyoBACT is undoubtedly due to amino acids linked on C4 of the HGA rings and provided by the laccase extract and the components of the H. titanicae culture medium (yeast extract and/or amines formed during growth), respectively. From the three 13 C solid-state NMR spectra, the main differences are located in the region around  45-78 ppm, precisely at 52.5 and 67.9 (larger) ppm in PyoENZ, and 52.5 ppm alone in PyoBACT, whereas these two shifts are absent in PyoCHEM (Fig. 2). They suggested secondary reactions during the biological BQA polymerization that did not occur during autoxidation of HGA in abiotic and alkaline conditions. By comparison to standard molecules analyzed in parallel, the  52.5 ppm shift corresponds to the bi-protonated carbon of the ethanolic moiety (-CH2-O-) from 2,5-dihydroxybenzyl alcohol (gentisyl alcohol). Besides gentisyl alcohol, we also noted minor peaks at  190.5 and 191 ppm in PyoBACT and PyoENZ respectively, absent in PyoCHEM, and ascribed to an aldehyde group of the end-product 2,5-dihydroxybenzaldehyde (gentisaldehyde) ( Table 1). Gentisyl alcohol and gentisaldehyde resulted from a decarboxylation reaction extensively detailed in Fig. 1. With a lot of precautions because solidstate NMR was a semi-quantitative tool, the relative level of the decarboxylation products in the polymers was deduced from areas of the corresponding peaks ( Fig. 2), and approximately evaluated at 12-13% (gentisyl alcohol 10% + gentisaldehyde 2-3%), explaining these two compounds could not be identified by FTIR. Besides, low signals visualized at  17.2-23.1 (PyoENZ) and 17.4-23.9 ppm (PyoBACT) were attributed to lipid residues provided by the enzyme extract and the culture medium, respectively. At least, the broad signal at  67.9 ppm present in PyoENZ only had been assigned to the hydroxylated 13 C of a saccharide moiety (>C-OH) and also brought by the laccase extract, whereas it was absent in PyoBACT probably because the H. titanicae medium was not supplemented with glucose. Importantly, the area ratio of the 170/118 ppm resonances for each pyomelanin (Fig. 2, data framed) indicates a correct -CH2-COOH substitution for PyoENZ and PyoBACT with a value of 1/6, whereas a loss of the carboxylic moiety on PyoCHEM structure (ratio 1/10) has occurred. More information on the polymer assembly was necessary to elucidate the mechanism of polymerization and the types of linkage between the rings, i.e. Car-Car (aryl carbon) or/and Car-O-Car (aryl ether) linkages. Interestingly, the FTIR spectra showed absorption at 1,534 cm -1 strongly present in PyoCHEM (Fig. 5S, red) and absent in PyoENZ (blue) and PyoBACT. This resonance did not correspond to amides and was rather ascribed to aromatic Car-H. From this remarkable difference, it has been established that PyoENZ contains much less Car-H free, which means much more Car-Car linkages than PyoCHEM. As an important finding from the three pyomelanin 13 C NMR spectra, Car-O-Car (aryl ether) linkages were absent (Fig. 2), the related signal generally resonates at around  160-167 ppm 39,40 . Hence, the three HGA polymers were assembled by Car-Car linkages only. An alkaline-H2O2 oxidation treatment has also been tried on the three pyomelanin 41 (Table 2). While hydrolyzed MelSYNTH and MelSEPIA melanin lead to the two expected degradation products, pyrrole-2,3-dicarboxylic acid (PDCA) and pyrrole-2,5,5-tricarboxylic acid (PTCA) similarly to the literature 9,42 , any compound has been detected from the three hydrolyzed pyomelanin by RP-HPLC-DAD and GC-MS analyses. Thus, pyomelanin could not be hydrolyzed by such peroxide treatment, even after doubling or lowering the peroxide concentration. A pyrolysis-GC-MS coupling method had been developed to analyze pyomelanin from Penicillium chrysogenum 43 but reported too much heterogeneity to obtain uniform results between samples. This method utilizes heat to break the polymer into smaller fragments, such as 4-methoxybenzene acetic acid, 4-methoxybenzene propanoic acid, and other minor phenolic compounds, but not HGA. Unfortunately, this technique failed in PyoENZ, PyoBACT, and PyoCHEM with any identifiable compound.
Physicochemical properties (summarized in Table 2). All pigments are insoluble in neutral or acidic water as well as many usual organic solvents, entirely soluble in alkaline media such as NaOH (0.05 N minimal conc.). Exceptionally, the pigment is soluble in DMSO at a concentration that does not exceed 0.5 mg/mL after 24 h agitation. Solubility in H2O was not improved after 2 days at 80°C. As solid form and/or solubilized in alkaline solutions, PyoENZ, PyoBACT, and PyoCHEM were stable until 80°C for 3 days (max. tested) with no degradation products detected by RP-HPLC, and GC-MS analyses, near size modification by GPC/SEC. Molecular weights (Mw) of the three HGA-pigments were successfully determined and have been found at 5,400 Da (dispersity Ɖ 15.3) for PyoENZ, 5,700 Da (Ɖ 11.9) for PyoBACT, and a less high Mw at 2,300 Da (Ɖ 6.64) and 2,000 Da for PyoCHEM and MelSYNTH (Fig. 6S, Table 2), explaining why PyoCHEM and MelSYNTH were more rapidly solubilized in DMSO than the others. These Mw data were very close to those resulting from the elemental analyses (Table 2), indicating that these pigments were sufficiently purified by successive water and ethanolwashings.
Antiradical properties. The scavenging ROS activity was studied for PyoENZ, comparatively to the standards MelSEPIA and MelSYNTH. UVA induces damage by directly transferring energy or indirectly through ROS generated as primary and secondary radiolytic products 44 . Therefore, the protection by melanin pigment against UVA may be due to their ability to scavenging ROS in the cells. To prove this, a fluorescein-derived compound (DCFH-DA) was used to detect the generation and change of ROS in UVA-visible irradiated keratinocyte cells. Indeed, keratinocytes are a source of ROS that may affect neighboring skin cells, such as melanocytes, and influence the process of melanogenesis or contribute to the progression of vitiliginous lesions. Fluorescence measurements showed that PyoENZ effectively scavenged ROS generated by UVA-visible light in the test system with an IC50 of 82.2 ± 5.6 µg/mL, while IC50 of MelSYNTH (284.1 ± 12.3 µg/mL) was higher and that of MelSEPIA very far (Table 3). Thus, the amount of ROS in the cells decreased as the concentration of PyoENZ increased, much more efficiently than the concurrent pigment MelSYNTH. The DPPH-antioxidant activity was rarely reported due to the insolubility of the pyomelanin in organic solvents, and because the stable DPPH reagent reacts at slightly alkaline pH values. The assays were carried out on the three HGA-pigments, along with the two standards (MelSEPIA, MelSYNTH) and common antiradical agents such as Trolox, ascorbic acid, and propyl gallate, all prepared in DMSO. Fig. 7S-A and 7S-B indicated that PyoENZ (EC50 27.5 µg/mL) and MelSYNTH (EC50 25.9 µg/mL) have an antioxidant activity equivalent to that of ascorbic acid (29 µg/mL), as already reported for pyomelanin isolated from Pseudomonas stutzeri strain BTCZ10 and Pseudoalteromonas lipolytica BTCZ28 45,46 . The degradation of pyomelanin by enzymes or microorganisms has never been described to date, whereas ascorbic acid is rapidly metabolized and was thought to act as a pro-oxidant when the glutathione pool is depleted 47 , a feature that must also be controlled in the case of pyomelanin. Barely better than PyoENZ, PyoCHEM EC50 was 20.0 µg/mL, while EC50 PyoBACT was found much higher at 130.0 µg/mL (Fig. 7S-A). Whatever, PyoENZ, PyoCHEM, and PyoBACT exhibited much higher DPPHantioxidant activities than pyomelanin isolated from the Yarrowia lipolytica strain W29 (EC50 230 μg/mL 48 ), and far from eumelanin from Sepia officinalis (MelSEPIA, > 300 µg/mL, this work) and the synthetic butylated hydroxytoluene (BHT, EC50 722 μg/mL 49 ). Although the trihydroxylated benzoic ester, propyl gallate (EC50 4.2 µg/mL) (Fig. 7S-B), was one of the leading dietary antioxidants, it induced DNA damages 50 ; BHT was also found cytotoxic 51 , hence their use notably in the food industry became restricted.

Photostability of PyoENZ.
In NaOH at three diluted concentrations, the pigment has been exposed to drastic UVA-visible irradiation, 200 J/cm 2 for one hour, a dose much higher than the minima imposed by the European Medecines Agency (72 J/cm 2 ) for photostability studies of substances. It was concluded that PyoENZ was highly photostable because no change was observed in the UV-visible spectrum (200-700 nm) of the polymer, comparatively to the nonirradiated sample, nor in its molecular weight assessed by SEC.
Electron-transfer efficacy. By an adapted ferrozine assay, PyoENZ, PyoCHEM, and MelSYNTH exhibited equivalent and highest Fe 3+ -reducing activity among the five polymers tested (Fig.  3). From these data, the equivalent Fe 3+ -reducing activity of PyoENZ and PyoCHEM could not be explained, while PyoENZ and PyoCHEM have a different Mw (Table 2), thus none the same number of -OH and carboxylic groups, and even if gentisyl alcohol and gentisaldehyde (at ~13%) are present in PyoENZ structure only. Comparatively to MelSYNTH (100%), the reducing activity in decreasing order was PyoENZ (96), PyoCHEM (95), and to a less extent PyoBACT (54) and MelSEPIA (34). Because PyoENZ has the best production yield and is dedicated to potent applications, its Fe 3+ -reducing activity was evaluated at 1.73 µM per hour related to 50 µg of pigment, i.e. 5.30 ng Fe 2+ /h/µg.

Cytotoxicity.
For applications with pyomelanin as an ingredient for cosmetics or pharmaceutical preparations, cytotoxicity toward human keratinocytes has been evaluated by the vital dye NR penetration technique, from pigment prepared in alkaline solutions at dilutions which in no way modified the pH of the assay. Keratinocytes are the most abundant cells of the epithelial layer of the skin and are used as a part of the 3D skin model for the assessment of the toxic hazard of cosmetic ingredients. No reduction of the metabolic activity of the cells was observed as compared to the non-treated cells, thus formally postulating the absence of toxic effect on skin cell metabolic activity for PyoENZ, PyoBACT, PyoCHEM, MelSYNTH, and MelSEPIA, until 500 µg/mL (Table 2). Furthermore, using the normalized OECD protocol commonly used for cosmetology product evaluation, the three pyomelanin and the two standard melanins were found non-phototoxic (PIF < 2) ( Table 2).

Discussion
The laccase process is the most efficient provider of pyomelanin. Comparison of three realistic strategies by optimized production of HGA autoxidation (PyoCHEM), induced bacterial culture (PyoBACT), and for the first time using a recombinant laccase (PyoENZ) was undertaken. PyoENZ has been obtained at the highest level (1.25 g per g 2,5-DMPA), a procedure that meets all the criteria to design a large-scale prototype, high-efficient, cheapest, with mild conditions, and without sterility constraints. HGA-lactone was easily prepared from 2,5-DMPA, or even from 2,5-DMAPO by an additional reaction of Willgerodt-Kindler 25 . Despite the great number of extensive works on pyomelanin-producing microorganisms, to date there have been only three reported quantifications of the pigment, first with the wild yeast Yarrowia lipolytica that furnished 0.035 g/L of culture 52 , second 0.173 g/L culture of the Shewanella algae BrY strain supplemented by 2 g of L-Tyr/L 14 , and third 0.35 g/L by random mutagenesis of Pseudomonas putida 19 . In this work, an induced culture of H. titanicae was shown to convert 4-HPA to HGA by a 4-HPA-1-hydroxylase (4-HPAH-1) at the best microbial yield to date, 0.55 g/L culture, a feature confirmed by the presence in its genome of the related hpaH/C genes (unpublished). Such bioconversion generally occurred with less energy consumption, and for these reasons more efficiently. It seems reasonable to assume that the bacterial (PyoBACT) and the chemical (PyoCHEM) processes will never be able to compete with the laccase process (PyoENZ) in terms of production, except maybe by developing a recombinant overproducing microorganism. From these results, the ability of H. titanicae to synthesize pyomelanin and its property to reduce Fe 3+ , raise the question of the suggested role of the strain to accelerate the destruction of the RMS Titanic (rust formation?), in addition to the survival of the bacterium at 4,000 meters depth.
A few remarks are worth noting about the oxidation of HGA. Besides the biological implications of metal-catalyzed oxidations, true autoxidation of biomolecules does not occur in biological systems, instead this autoxidation is the result of transition metals bound to these biomolecules 53 . By analyzing the PyoCHEM structure, surprisingly the 13 C solid-state NMR spectrum revealed an unexplained loss (~40%) of carboxylic moiety during the alkaline Mn 2+autoxidation of HGA without observable by-products of this degradation, and hence contributes to the low pyomelanin yield. To date, the in vitro polymerization of HGA by a laccase has never been studied before. Here, the rMt laccase had been found to efficiently catalyze the HGA polymerization in terms of yield, and still confirmed the involvement of these oxidases in biological environments. It should be noted that pyomelanin-forming bacteria generally grow at pH 6-7, while the autoxidation is optimal at pH 8-9, one more element in favor of the laccase(s) action in living cells. The rMt enzyme is largely available and one of the cheapest in the market. Partial purification by ultrafiltration of the rMt extract would be an additional step unnecessary. Indeed Aljawish et al. 54 showed that, if the brown color decreases after UF (⁓90%), it eliminates only 2.5-fold of total proteins and the specific activity of the UF-enzyme increased by only 2.1-fold. Other laccases had also been assayed in parallel. At their optimal parameters, we evaluated that the Trametes versicolor enzyme furnished ~2 fold less pyomelanin than rMt, while the purified recombinant Pycnoporus cinnabarinus laccase, an unavailable enzyme, gave a quite similar yield, i.e. 1.1-1.2 g pyomelanin per g of substrate, at pH 5 in an acetate buffer. The molecular weight of pyomelanin rarely reported was first evaluated by GPC/SEC at 3,000 Da for the pigment of the bacterium Alcaligenes eutrophus, and at 1,700 Da for autoxidized HGA, however using unconventional PEG/PEO standards 29 . Turick et al. 14 estimated the size of Shewanella algae BrY pyomelanin ranging from 12 to 14 kDa, however by high-speed sedimentation and with proteins for calibration. In this work and as a suitable method in an alkaline eluent, the optimized processes led to close Mw of 5,400 and 5,700 Da for PyoENZ and PyoBACT, respectively (Table 2), a size much higher than those of laccase-synthesized polymers of catechol (Mw 1,268 Da), resorcinol (1,489 Da), and hydroquinone (1,157 Da) 27 .
Biological pyomelanin is a Car-Car assembly polymer that contains two decarboxylationissued products. Because alkaline-H2O2 hydrolyses and pyrolysis experiments failed, the chemical structure of the three pyomelanin was determined by 13 C solid-state NMR, and partly confirmed by FTIR analyses. Similar to the hydroquinone polymerization 27 , Car-Car bindings between the rings predominated in PyoENZ, PyoBACT, and PyoCHEM, a finding deduced from the absence of Car-O-Car (ether linkages) resonance in the NMR spectra of these polymers. The reactions that govern the polymerization of HGA by the rMt laccase were proposed in Fig. 4A and showed two main suggested assembly modes, C4-C6 (-bindings) and C3-C6 (-bindings), giving preference to the C3-C6 mode because of less subject to steric effects. Based on the NMR data, it was not possible to differentiate between the eight possibilities (Fig. 4A). The mechanisms of polymerization through radical reactions have also been proposed in Fig. 4B, still showing C3-C6 and C4-C6 linkages as the most probable structures for pyomelanin. In any case, analytical techniques are still not able to deliver the exact structure, this is the most problematic for all melanin and especially pyomelanin. Nonetheless, there is still to understand the mechanism of polymerization of HGA, particularly the relationships with the laccase structure. In this work, we notably reported a biological decarboxylation from BQA and caused by the action of the rMt laccase or suggested bacterial oxidase(s) (Fig. 1). Such a mechanism led to the formation of gentisyl alcohol and gentisaldehyde representing about 13% of the total components, and which polymerized together with BQA. Until today, HGA decarboxylation was attributed to an abiotic reaction from BQA at acidic pH (near [4][5], forming gentisaldehyde as the major product along with minor gentisyl alcohol 34 . In this work, PyoBACT and PyoENZ contained both products, but in reverse order of level, gentisyl alcohol (major) and gentisaldehyde (weak compound). No decarboxylation was observed during the abiotic and alkaline synthesis of PyoCHEM.
The pyomelanin PyoENZ for multiple applications. In addition to a DPPH-antioxidant activity equivalent to ascorbic acid, a high thermostability over time, a non-degradability in cells, PyoENZ efficiently scavenges ROS from irradiated human keratinocytes much better than the concurrent MelSYNTH (Table 3). Comparatively and with a similar technique, 400 µg/mL of L-Dopa melanin isolated from Pseudomonas maltophilia has been reported to almost totally scavenge ROS from UVA-induced fibroblast cells 55 . Human eumelanin and pheomelanin photogenerate ROS meanwhile they photoconsume O2 and are protective against skin cancer 56 . Nevertheless, they photochemically generate melanin degradation products that are responsible for sunlight-induced melanoma formation by inducing cyclobutane-pyridine dimers (CPDs) from DNA 57 . In contrast, strong irradiation of PyoENZ in solution did not generate degradation products, UV-visible spectroscopy, GPC/SEC, and RP-HPLC being reliable techniques to determine that PyoENZ is also a photostable polymer. At first glance, the two decarboxylation products in PyoENZ and PyoBACT structure did not seem to influence the Fe 3+ -reducing activity (Fig. 3). In any case, ferric-reducing activity may be considered as a marker of the redox cycling nature of pyomelanin, a property that might be used to conduct electricity like an electronicionic hybrid conductor. It was advanced that only a few femtograms per cell were assumed to be a sufficient amount for electron-transfer in bacterial systems 3,14,16 . Consequently, PyoENZ could be exploited as a hyperthermostable and Fe 3+ -reducing agent, and for bioelectronic applications better than melanin 58 . As an evident cosmetic ingredient and consistently with recent reports on microbial pyomelanin from Yarrowia lipolytica 48 and Pseudoalteromonas lipolytica BTCZ28 46 , all tested at a 100 µg/mL, PyoENZ was found non-cytotoxic and nonphototoxic on keratinocyte cells, until 500 µg/mL.

Chemicals and enzymes.
Solvents of mass spectrometry grade were supplied by Biosolve (Dieuze, France), media for human keratinocytes, and mouse fibroblasts cultures from Dutscher (Brumath, France), main chemicals including standard melanin from Sigma. Natural melanin (MelSEPIA, reference M2649) consists of purified eumelanin from the ink of Sepia officinalis. Synthetic melanin (MelSYNTH, M8631) is an L-Dopa melanin obtained from L-Tyr in presence of H2O2. HGA and HGA-lactone were used as standards for HPLC and the GC-mass data bank. The Aspergillus sp. laccase (SAE0050) consisted of a highly concentrated and brown misciblewater solution (density 1.15 g/mL, stored at 4°C) obtained by submerged fermentation of the recombinant Myceliophtora thermophile laccase expressed in Aspergillus oryzae. This enzyme originally furnished by Novozym under reference 51003, was re-named here 'laccase rMt'. Purified laccases from Pycnoporus cinnabarinus (recombinant, a gift of The Fungal Biodiversity and Biotechnology Laboratory, Marseille 59 ) and Trametes versicolor (Sigma, 38429) were also used.
Halomonas strain selected and growth conditions. The strain Halomonas titanicae was provided from the DSMZ collection (Germany), isolated and taxonomically characterized in 2010 43 , and compared in our laboratory among a large collection of Halomonas spp. 32 The strain was grown in a shaker (130 rpm) at 30°C in a basal medium containing (in g/L), yeast extract 1.0, NaCl 20, KH2PO4, K2HPO4 0.6, NH4Cl 1.0, MgCl2·6 H2O 10, CaCl2·2 H2O 0.1. The pH was adjusted to 7.0 with a 4 N NaOH solution. Aliquots (25 and 500 mL) were dispensed into flasks and sterilized by autoclaving at 120°C for 20 min. Ca-and Mg-chloride stock solutions were sterilized separately, and the accurate volumes were added to the medium. L-Tyr and 4hydroxyphenylacetic acid (4-HPA) stock solutions (250 mM in milliQ-H2O, neutralized, heated moderately, sterilized by 0.2 µm pore size filtration) were added before inoculation at 5 mM final concentration. The strain was pre-cultivated twice for 2 days each, in 25 mL of the same saline basal medium containing either L-Tyr or 4-HPA. The preculture served as inoculum at 10% (v/v) for the culture in 500 mL volume that was agitated at 150 rpm, until the A400 nm no longer changed. To overproduce pyomelanin, 500 mL cultures supplemented by repeated addition of 4-HPA amounts were carried out using the experimental design Azurad software (a company of Marseille). The 8 experiments resulted from the defined parameters, the response Yi (mass of pyomelanin per Liter of culture), and the 6 entry parameters, X1 for the 4-HPA concentrations added (2, 5, and 10 mM), and X2 for the time of supplementation (at 0, 2 or 3 days of growth). An experimental domain of cubic form was chosen and a second-degree polynomial model applied 60 .

Bacterial (PyoBACT) and chemical (PyoCHEM) pyomelanin preparation.
Bacterial cultures at the stationary phase were centrifuged (8500 g, 30 min), pyomelanin (PyoBACT) in the supernatant was precipitated by the addition of 2 N HCl (final conc.), and the solution left to rest for 24 h, at room temperature in the dark. After centrifugation (8,500 g, 20 min), the brown-black pellet was washed twice successively with milliQ-H2O (3 x) and ethanol (1 x), centrifuged, dried at 70°C for 2 days, weighed, and powdered as fine particles before storage in glass vials at room temperature. Chemical pyomelanin (PyoCHEM) was prepared by autoxidation of HGA. Routinely, 0.84 g of HGA (100 mM, see below for synthesis) and 50 mg MnCl2, 4 H2O (5 mM) were dissolved in 50 mL milliQ-H2O. The pH was adjusted to 9.0 (pH-meter control) with drops of NaOH 6 N, and the solution agitated in dark for 3 days with a barrel at 30°C. The pigment was precipitated with 6 N HCl (final conc.), left to sediment for 12 h, and centrifuged (8,500 g, 20 min). The light-brown colored supernatant indicated the presence of many oxidized compounds that could not precipitate, even by increasing the acid until 10 N. After washing and drying, the pelleted PyoCHEM was weighed and stored as for the bacterial pigment. To determine the optimal (HGA)/(Mn 2+ ) ratio, concentrations of HGA (1-300 mM) and MnCl2 (0.5-20 mM) were assayed in the same manner, the black-brown solutions diluted 50x in NaOH 0.1 N and absorbance read at 400 nm (A400 nm).
Process for the production of pyomelanin (PyoENZ) by the rMt laccase. The first part of the procedure consisted of the HGA synthesis 25 with modifications. The second part is the polymerization step by the rMt laccase. The starting compound 2,5-dimethoxyphenylacetic acid (2,5-DMPA) 5 g was solubilized in 40 mL of 48% HBr and refluxed gently for 4.5 h in a 100 mL bicol flask provided with a refrigerant maintained at 10°C. The resulting deeply red solution was evaporated to dryness in vacuo, the residue (3.80 g, 99.7% yield, 99.8% purity) identified as HGA-lactone following its UV spectrum (max 232, 289 nm, bands slightly lower than that of HGA), elution in RP-HPLC (retention time 4.2 min), and GC-MS analyses of the TMSderived compound (rt 16.3 min), similarly to the standard. In the second step of the procedure and typically, 1.0 g of HGA-lactone was dissolved by agitation in 130 mL hot milliQ-H2O (70°C), stayed 3-5 min and few drops of NaOH 2 N added until pH 9.3 (pH-meter) to hydrolyze the lactone into HGA (in HPLC-DAD, rt 2.7 min, max 290 nm), complete ring-opening was ensured by analysis of a 5 µL sampling diluted 10x in MeOH. Immediately after, 35 mL of Naphosphate buffer 0.3 M pH 6.8 were added, the concentration of HGA and buffer at this stage was 40 mM and 65 mM, respectively. Once the temperature of the solution has reached 30-40°C, 3-4 mL of concentrated laccase rMt were added (2250-3000 U in total), the enzyme activity was 750 U/mL (SD <5%) as determined by the syringaldazine assay (see Fig. 2S, Additional information). Then the mixture was agitated at 130 rpm in dark at 30°C for 48 h. The formed brown-black pigment was further precipitated by adding 34 mL of HCl 37% (2 N final concentration), agitated for 2 min, and stayed for 24 h, at ambient temperature in dark. The precipitated pyomelanin was centrifuged, washed with milliQ-H2O and ethanol, dried, and weighed as previously for PyoBACT and PyoCHEM (Fig. 1, step 5). The yield of the process was determined as a mean of three independent preparations from 2,5-DMPA. For optimal pyomelanin yield, laccase activity and HGA concentration to be used were determined in 4 mL glass vials tightly closed and containing 500 µL of Na-phosphate 100 mM buffer pH 6.8 (final 50 mM), 50 to 500 µL (5 to 50 mM) of HGA 100 mM stock solution in milliQ-H2O, 5 to 40 µL laccase rMt (3.75-30 U), and milliQ-H2O (qsp), in one mL total volume. Assays were incubated at 30°C or 50°C, at incubation time varying from 1 to 30 h, under 120 rpm agitation in dark. The enzyme activity was stopped by placing the tubes in a boiling water bath for 10 min, and pyomelanin content was evaluated by spectrophotometry (see below pyomelanin monitoring). Each point value resulted from triplicate tubes and a mean ± SD.
Homogentisic acid and gentisyl alcohol syntheses. Alkaline hydrolysis of 1.0 g of HGAlactone in hot 130 mL H2O was immediately acidified until pH 5 with drops of HCl 37%, followed by the addition of 6.5 mL of a saturated NaCl solution (5% v/v final conc.). Then HGA was extracted 3x in a funnel with AcOEt, the whole organic phase washed with milliQ-H2O, clarified with solid Na2SO4, filtered, and evaporated to dryness. To eliminate residual BQA, the dried HGA was dissolved in 0.1% HCOOH and applied on a glass column (20 mL Luer-lock tip syringe mounted with a vacuum flask and a pump) containing 10 cm 3 of Lichroprep RP18 (from Sigma) previously conditioned with MeOH and acidified H2O. After washing by 2 vol. of acidified H2O, HGA was eluted by a mixture of MeOH-acidified H2O (1:9, v/v), evaporated to dryness, and resulted in a 99.9% purity light grey HGA (yields ~70 wt%), as determined by RP-HPLC and GC-MS analyses, indicating that recrystallization was not necessary. As standard for NMR analyses, pure gentisyl alcohol was synthesized from gentisaldehyde by NaBH4-reduction in tetrahydrofuran (yield 49%) 61 , purity confirmed by GC-MS and 1 H-NMR in d6-DMSO.
Cytotoxicity. The viability of cells exposed to melanin was expressed as the concentrationdependent reduction of the vital dye Neutral Red (NR) uptake in intracellular lysosomes. Assays were carried out with the three prepared pyomelanin and the two melanin standards (Table 2), all prepared at 10 mg/mL in NaOH 0.05 N (stock solution). Human epidermal keratinocytes neonatal cells were maintained in a complete keratinocyte serum-free medium (Panserin 412, from Dutscher) supplemented with bovine pituitary extract (30 µg/mL), recombinant epidermic growth factor (rEGF, 0.2 ng/mL), and an antibiotic cocktail of 10 U/mL penicillin-100 µg/mL streptomycin. Precultures were seeded into 96-well plates (0.2 mL per well) at 1.10 5 cells/mL concentration. After incubation at 37°C (5% CΟ2) for 24 h until semi-confluent, the medium was decanted, replaced by 200 µL of complete medium containing the melanin (8 concentrations, 0-500 µg/mL), and cells were incubated again for 24 h. After removing the medium, cells were washed, placed into the NR medium (50 µg/mL NR in the complete medium), and incubated for 3 h (37°C, 5% CO2). The medium was removed, cells were washed three times with 0.2 mL of HBSS (Hank's Balanced Salt Solution, from Dutscher) to eliminate the excess dye, and 50 µL per well of a distaining solution (50% ethanol, 1% acetic acid, 49% milliQ-H2O) was added. The plates were shaken for 15 min at room temperature in the dark. The membrane damage degree, i.e. the increase of released NR, was determined by the A540 nm in an Infinite M200 Pro (Tecan, Swiss) reader. The results obtained for wells treated with the pigment were compared to those of untreated (100% viability) and converted to percentage values. Cell viability was calculated as Viability (%) = [A540 (test well) -A540 (blank)] / [A540 (negative control) -A540 (blank)]. The concentration of the pigment causing a 50% release of NR as compared to the control culture (IC50, in µg/mL) was calculated by non-linear regression analysis using the Phototox v2.0 software (Zebet, Germany).
Phototoxicity. The in vitro and normalized 3T3 NRU assay (OECD number 432) was used. Balb/c 3Τ3 mouse fibroblasts (3Τ3-L1, ATCC® CL-173™, from US type Culture Collection) were grown in DΜΕΜ supplemented with L-glutamine 4 mM and 10% of inactivated calf serum, seeded into two 96-well plates (0.1 mL per well) at 1.10 5 cells/mL concentration, and incubated (37°C, 5% CΟ2) for 24 h until semi-confluent. The medium was decanted and replaced by 100 µL of HBSS (see before) containing the appropriate pigment concentrations (8 concentrations, 0-500 µg/mL), then cells were incubated (37°C, 5% CO2) in the dark for 60 min. From the two plates prepared for each series of pigment concentrations and the controls, one was selected, generally at random, for the determination of cytotoxicity without irradiation (-Irr), and the other for the determination of phototoxicity with irradiation (+Irr). For each set of experiments, a negative control (in HBSS) and positive control (chlorpromazine final concentrations from 1 to 100 µg/mL (-Irr) and 0.01 to 1 µg/mL (+Irr), diluted in ethanol) were performed. The percentages of cell viability were calculated as previously (cytotoxicity). Irradiation was performed with a solar simulator Suntest CPS + (Atlas Material Testing Technology BV, Lochem, Netherlands) device equipped with a xenon arc lamp (1100 W), a glass filter restricting transmission of light below 290 nm, and a near IR-blocking filter. The irradiance was 750 W/m 2 corresponding to 4.5 J/cm 2 for one-min irradiation (0.41 J/cm 2 of UVA and 4.06 J/cm 2 of visible light). The Photo-Irritation-Factor (PIF) defined by the ration IC50 (-Irr) / IC50 (+Irr) was expressed to finalize the results. Based on validation studies (OECD 432 guideline), a test substance exhibiting a PIF < 2 predicts no phototoxicity, 2 < PIF < 5 a probable, and PIF > 5 a phototoxicity.

Photostability of PyoENZ in solution.
It was evaluated on 4 mL glass-closed tubes containing 3 mL each of PyoENZ solution at 0.05, 0.1, and 0.5 mg/mL NaOH 0.05 N. The tubes were placed horizontally and irradiated in the Suntest CPS + solar simulator, respecting the ICH Q1B guidelines (European Medecines Agency). A strong irradiance by the xenon lamp was maintained with light energy of 550 W/m 2 during 1 h (i.e. 200 J/cm 2 UVA-visible irradiation). Changes in the polymer structure were monitored by UV-visible spectroscopy from 200 to 700 nm and GPC/SEC (see Fig. 6S), comparatively to non-irradiated samples.
Metabolites identification, pyomelanin monitoring. Phenolic compounds along the three processes were identified by RP-HPLC-DAD and GC-MS according to our previous works 33,34,35 . To control the pigment formation during the bacterial culture and for optimization of the processes, the black-brown solution was diluted 20 and 50x in NaOH 0.1 N (qsp 1 mL), respectively, and absorbance (A400 nm) read against the same alkaline reference 2 .

Conclusions
Pyomelanin issued from the three processes has different properties, giving a large priority to PyoENZ that can now be produced in interesting yield and at low cost. The pigment efficiently scavenges ROS, exhibits high DPPH-antioxidant activity, is non-degradable, non-toxic, and can be stocked indefinitely without any precaution. As a representative pigment of microbial pyomelanin, PyoENZ becomes an available standard for laboratories, might be used for applications that require extreme conditions, as an electron-transfer agent, why not for energy storage, and exploited for skin protection, assuming that it cannot penetrate the blood skin vessels.

Additional Information
Supplementary information accompanies this paper at http://www.nature.com/xxxx

Author Contributions
J.L. initiated this study, designed the experiments, analyzed the data, supervised the research, and reviewed the manuscript. F.L. designed and executed the experiments, collected and analyzed the data, wrote the manuscript, and prepared the figures. F.Z. executed the NMR spectra and corrected the concerned part of the text. A.A. studied with J.L. the H. titanicae genome (hpaH/C genes), proposed this strain as a good candidate for producing pyomelanin, and helped in the culture experiments. C. Di G. managed the cytotoxicity, phototoxicity and ROS experiments with F.L. M.R. synthesized gentisyl alcohol and HGA. P.P. participated in the strategy and entirely supervised the financial aspects of this work. All authors have read and approved the final manuscript.

Funding
FL works were supported in part by an ANRT-CIFRE (Association Nationale Recherche Technologie-Conventions Industrielles de Formation par la Recherche, France) fellowship.