Potential metal chelating ability of mycosporine-like amino acids: a computational research

Mycosporine-like amino acids (MAAs) are low-molecular-weight (< 400 Da) water-soluble secondary metabolites that are attributed many functions such as antioxidants, compatible solutes, nitrogen reservoirs and especially, photostable UV protectants. Recently, they are attracting attention due to their biotechnological and industrial potential for anti-aging and wound healing properties as well. In this study, we explored the metal chelating capacity of selected MAAs (4-deoxygadusol, mycosporine-glycine, mycosporine-taurine, palythine, poryphyra-334, shinorine, mycosporine-2-glycine and euhalothece-362) making use of density functional theory (DFT) calculations. We report model structures of ferrous and ferric ion–MAA complexes and their binding affinities in relation to their structural differences and multiple sites available for chelation on the MAAs. We also investigated calcium ion complexes for mycosporine-glycine, shinorine, porphyra-334 and mycosporine-2-glycine. Our findings support suggestions made to explain some experimental results obtained in previous studies on MAAs. Lastly, we briefly mention the findings in the context of early life and hence relevance to astrobiology. To the best of our knowledge, this is the first data report on MAAs metal chelation ability and ascribes them a new role as “metal chelators.”


Introduction
Mycosporine-like amino acids (MAAs), widely distributed in various organisms all over the globe (Shick and Dunlap 2002), were originally found in 1960s (Řezanka et al. 2004) and have attracted attention ever since. There are 1 3 many reviews written about their structures, biosynthesis (Sinha and Häder 2008;Singh et al. 2008;Pathak et al. 2019); roles such as antioxidant molecules, compatible solutes, nitrogen reservoirs; protection against UV radiation, desiccation or thermal stress, (Oren and Gunde-Cimerman 2007;Wada et al. 2013Wada et al. , 2015 as well as their biotechnological and industrial potential as natural sunscreens, antiphotoaging molecules, stimulators of skin renewal and functional ingredients of UV-protective biomaterials Sinha 2009, Chrapusta et al. 2017;Fuentes et al. 2019;Rosic 2019). MAAs are low molecular-weight (< 400 Da) water soluble compounds that absorb UV radiation with λ max 310 to 365 nm and high molar absorptivity. Therefore, they are generally considered to be UV absorbing pigments and regarded as "multipurpose" secondary metabolites due to their additional roles listed above (Oren and Gunde-Cimerman 2007). MAAs' structures consist of a cyclohexenone or cyclohexenimine ring substituted with a methoxy group at C2, a hydroxy group and a hydroxymethyl group at C5 and are conjugated to an amino compound (generally glycine) at C3 of the ring ( Fig. 1) (Sinha and Häder 2008;Simeonov and Michaelian 2017). A second amino acid (amino alcohol or enamino group) may be substituted at C1 of the ring.
MAAs may be extracellular as well as intracellular, and covalently bound to oligosaccharides in the glycan sheath of cyanobacteria (Böhm et al. 1995). A significant amount of organic matter bound to diatom frustules comprise of MAAs (Ingalls et al. 2010); the cyanobacterium Trichodesmium spp (Subramaniam et al 1999), dinoflagellate Lingulodinium polyedra (Vernet and Whitehead 1996), and dinoflagellate Prorocentrum micans (Tilstone et al. 2010) actively secrete MAAs into the surrounding water during surface blooms; shinorine is exclusively located in the extracellular matrix in cyanobacteria Microcystis aeruginosa PCC 7806 and does not play a major role in UV protection but may be involved in extracellular matrix formation and cell to cell interaction (Hu et al. 2015).
Cyanobacteria (CB) being very diverse and ubiquitous on earth (Shih et al. 2013) are the most studied organisms among those found to produce MAAs (i.e., prokaryotes, eukaryotic algae, invertebrates, dinoflagellates, vertebrates, bacteria, and yeast) (Jain et al. 2017) and most likely evolved from anoxygenic phototrophs (Shih 2019). The development of oxygenic photosynthesis by CB which flourished together with archaea in Earth's earliest ecosystems, 3.5 to 3.2 Ga. (Hickman-Lewis et al. 2020), caused the great oxygenation event (GOE) of the atmosphere during the early Proterozoic, 2.5-2.3 Ga, and led to evolutionary changes on earth (Schirrmeister et al. 2015). Consequently, they were the first taxa to adapt to oxidative stress with existing systems (Fischer A gene cluster for MAA synthesis, homologous to those identified from CB, was recently confirmed for two strains of actinobacteria (Miyamoto et al. 2014). Actinobacteria, similarly to CB are very common and diverse (Lewin et al. 2016), predate the GOE, and together with CB and Deinococcus share a common ancestor (Battistuzzi et al. 2004).
Photoferrotrophy, a form of anoxygenic photosynthesis (use of light as energy source and ferrous ion as electron donor), is one of the oldest photoautotrophic metabolisms on Earth (Camacho et al. 2017); yet mechanistically, the biochemical pathway for the conservation of energy coupled to the oxidation of iron is still poorly understood (Floyd et al. 2019). Photoferrotrophs also display mechanisms preventing cell encrustation with oxidized iron (Miot et al. 2009;Saraiva et al. 2012;Wu et al. 2014) and some anoxygenic phototrophic Fe (II) oxidizers form nanometer-sized grains of ferrihydrite loosely attached to the cell surfaces to protect themselves from UV irradiation as an early survival strategy in environments where Fe (II) was abundant (Gauger et al. 2015). Kolo et al. (2009) report iron encrusted "microbial like structures" on surfaces of hematite in a banded iron formation and state that "original cells" could have been Fe (III) reducing bacteria or "other" and that pitting found on surfaces might have resulted due to dissolution via iron chelators.
Iron amino acid chelates, such as iron glycinate chelates were developed to be used as food fortificants and therapeutic agents in the prevention/ treatment of iron deficiency anemia and are available commercially (Hertrampf and Olivares 2004). Iron amino acid chelates can be used as an alternative for Fe-EDTA to supply iron in nutrient solutions for plant growth (Ghasemi et al. 2012). MAAs, having at least one amino acid (usually two) and being antioxidants (Wada et al. 2015) seem to be ideal candidates to chelate iron.
There are a few quantum calculational theoretical studies carried out at the molecular level on MAAs. Klisch et al. (2007) in their experimental and calculational studies on porphyra-334 depict the absolute configuration at the stereogenic center of the ring (C5) as S; E configuration at the imino moiety; and an exceptionally high proton affinity (265.7 kcal mol −1 ). The photoprotection mechanism of palythine as elucidated by Sampedro (2011) is a very rapid deactivation of the excited state in which light energy is dissipated as heat. Losantos et al. (2015) describe a similar mechanism for highly photostable gadusol. Matsuyama et al. (2015) reveal the pH-independent charge-resonance mechanism of shinorine and related MAAs to be responsible of UV protection. Orfanoudaki et al. (2020), in a very recent study, report the absolute configuration of 14 MAAs and show that all the tested MAAs have the ability to inhibit collagenase. They refer to two studies that hypothesize MAAs could play a role due to chelation of iron and calcium for this unknown mechanism of inhibition. In the first paper, Volkman et al. (2006), state that the MAA "Euthalece-362 with its four hydroxyl groups together with the alanine carboxyl group, may act as an iron chelator." In the second paper by Tarasuntisuk et al. (2018) it is stated that mycosporine-2-glycine's inhibitory activity on protein crosslinking may have been due to the chelation of calcium ions.

Modeling and computations
Previous quantum calculational studies carried out on MAAs provide most of the structural information required to model MAAs at the molecular level, i.e., geometries, conformations, isomers, absolute configurations, zwitterions, etc. (Klisch et al. 2007;Sampedro 2011;Losantos et al. 2015;Matsuyama et al. 2015;Orfanoudaki et al. 2020). We modeled ferrous and ferric iron-MAA complexes both for the neutral forms and the anionic forms of the selected MAAs.
The numbering used throughout this work is shown on the figure for 4DG (Fig. 2) which represents the basic structure of the conjugated oxo-ring. The anion (enolate form) of 4DG, 4-deoxygadusolate (4DG-ate), dominates the acid-base equilibrium in water, delocalizes the charge at pH 7 as shown in Fig. 2 and is protonated to the neutral form under acidic conditions (Losantos et al. 2015;Matsuyama et al. 2015). C5 is the stereogenic center on the ring. The protonation of the oxygen on C1 produces the stereoisomer (mirror image) of 4DG-ate obtained by the protonation of the oxygen on C3 (the priority numbers of the groups surrounding the stereogenic center C5 change). Therefore, both stereoisomers are equally expected to be present under acidic conditions. Iron is separately placed in two different locations, Position 1 or Position 2 (Fig. 2), to test iron chelation of 4DG. These two positions become equivalent for 4DG-ate due to its charge-resonance delocalization.
MG, derived from 4DG, has a glycine moiety on C3, the absolute configuration S at C5 (White et al. 1989) and three positions in which iron may be tested for chelation. Position 1 is the same as that in 4DG; position 2 in between the methoxy oxygen and the carboxyl group, also in close proximity to the nitrogen; position 3 in between the carboxyl group and the two hydroxy groups on C5 (Fig. 3). MG is in its zwitterionic nature in the range pH 4-10 and delocalizes the charge over 5 atoms, from O to N ( Fig. 3) (Matsuyama et al. 15). It is not clear whether the neutral or zwitterionic forms of the MAA structures are active in vivo (Sampedro 2011); therefore, both the neutral forms and the anions of their zwitterionic forms are studied (Fig. 3). The neutral forms have a carboxylic acid group with a hydrogen and another hydrogen either on C1-O or C3-N. The iron ion may bind the one which does not have a hydrogen, either C1-O or C3-N to form an iron-MG chelate model (see Fig. 3). H + ion (either on C1-O or on C3-N, each separately) is removed from the zwitterion and an iron ion is placed in Position 1 (P1), Position 2 (P2) or Position 3 (P3) to form an iron-MGate chelate model (see Fig. 3). Also, each form is modeled both as a ferrous and a ferric ion chelate. This procedure is carried out for all the MAAs.
MT has the same core structure as MG but with a taurine moiety on C3. The longer chain in taurine compared to glycine (two -CH 2 − groups between the nitrogen and the sulfur) may allow the -SO 3 H (or -SO 3 − in the zwitterion) group to reach a larger conformational space. The absolute configuration of MT is undefined, but it is derived directly from 4DG therefore we adopted the S configuration at C5. PT structure is the same as that of MG with a replacement of the keto group with an imino group and represents the imino core ring (Fig. 4). SH and PR have serine and threonine moieties, respectively, on C1. They both have E configuration at the imino moiety, absolute configuration S at C5, and another chiral carbon (S) bound to the imino nitrogen with a carboxyl group on it (Klisch et al. 2007;Orfanoudaki et al. 2020). The carbon that bears the methyl group in PR has absolute configuration R (Klisch et al. 2007). PT, SH, PR and M2G are all in zwitterionic nature in the range  Orfanoudaki et al. 2020) and delocalize the charge in the zwitterionic form as shown in Fig. 4. EU has a 2,3-dihydroxyprop-1-enylimino group at C1 and alanine substituted at C3 (Volkmann et al. 2006) and its charge-resonance delocalization is extended to seven atoms ( Fig. 1). The absolute configuration at C5 being undefined, we chose to model the S form like the rest of the MAAs in this study. We then modeled MG, SH, PR, and M2G with the same strategy for their calcium ion complexing ability. Hybrid DFT methods offer excellent performance in the prediction of geometries of small and medium size molecules. Klisch et al. (2007), in their experimental and calculational studies on porphyra-334, used gas phase B3LYP/6-31G(d) methodology; Cardozo et al. (2008) studied structures of palythine with a similar level of theory, B3LYP functional with the 6-31+G(d,p) basis set, in their experimental and theoretical calculations. Our aim being the search of good geometries to model potential iron-chelates for a series of MAAs that have already been studied both experimentally and calculationally, our choice was the most cost-effective method, B3LYP in conjunction with 6-31G(d,p) basis set.
First, the geometry of each iron-MAA complex was fully optimized and a frequency calculation was carried out to ensure true minimum (no imaginary frequency) in gas phase at 298 K, 1 atm. Then, the set of MAA complexes of lowest energy were optimized again including polarizable continuum medium (PCM) calculations to study the solvent (water) effects. Lastly, we added double diffusion functions, B3LYP functional with the 6-31++G(d,p) basis set, and repeated the gas phase and PCM calculations for the complexes that bind iron covalently. Calculations were carried out making use of the Gaussian'09 program package (Frisch et al. 2009).
The metal affinities were calculated according to the following equations in which M is the metal ion = Fe 2+ , Fe 3+ , Ca 2+ ; MAA stands for neutral forms and MAA-ate stands for the anion of the zwitterionic form of MAAs.

Iron complexes
The procedure we applied to model iron-MAA complexes implies that the iron cation either replaces the H + ion removed at P1/P2 or prefers to bind at P3. To rephrase it for the iron-MAA-ate complexes, the carboxylic acid group releases the H + , the zwitterion forms, and the ferrous or ferric iron ion replaces the H + ion (at P1 or P2) or binds at P3. We tested different multiplicities for the ferrous/ferric complexes formed at different positions to find out if there are incidences at which the energy of the complex formed is lower than the one with the high spin. The calculated energies for the Fe 2+ complexes with multiplicity 5 and Fe 3+ complexes with multiplicity 6 were those with the lowest energies of all the obtained complexes. From here on, "Fe 2+ " stands for ferrous ion with multiplicity 5 and "Fe 3+ " stands for ferric ion with multiplicity 6. A concise summary of the theoretical work including complexes which are not among the most stable ones are presented in Table S1 (energies, metal affinities, complex forming bond lengths and important metal to closest atoms distances). The results and data for optimized structures of the most stable ferrous and ferric iron complexes are presented in Tables 1 and 2 and their 3D structures are provided in Figures S1 & 2. The 2D representations of the initially obtained gas phase most stable ferrous and ferric iron complexes are shown in Fig. 5.
All our MAA models are in accordance with the structure reported for palythine by Furusaki et al. (1980) "the six membered ring in an envelope form, C5 out of the plane formed by the other five carbon atoms, the -OH group on C5 in the axial and the -CH 2 OH group in the equatorial position". Figure 6 shows the 3D optimized structure of [Fe(MGate)] 2+ , MG-ate covalently complexing ferric ion at position This plane also accommodates the iron and the carboxyl O. C5 is 25° out of plane (dihedral angle C2-C3-C4-C5) and the methyl group of the methoxy substituent on C2 is 66° rotated from the axial OH group at C5 (with a tendency to move away from the metal ion). This structure can be a general model for [Fe (MAA-ate)] 2+ series shown in SI (Fig. S2).
4DG binds Fe 2+ at P1 with a covalent bond to the carbonyl oxygen. We were not able to obtain a ferric complex for 4DG. 4DG-ate binds both Fe 2+ and Fe 3+ at the only available position (P1 and P2 become equivalent as explained above). As seen in the first column in Fig. 5, MG, PT, M2G, EU all bind Fe 2+ at P2 with a covalent bond to the nitrogen on C3. MT binds Fe 2+ at P2 with a covalent bond to a sulfonic oxygen. SH and PR bind Fe 2+ at P1 with a covalent bond to the nitrogen on C1. MG binds Fe 3+ covalently to the oxygen of the carboxylic acid group at P3 but we could not model a covalently bound Fe 3+ complex for any of the other MAAs at this position.
MG-ate, PT-ate, M2G-ate, EU-ate all bind Fe 2+ at P2 (for M2G, P1 = P2) with a covalent bond to the carboxyl oxygen (see second column in Fig. 5). Our attempts to model a ferrous MT-ate complex at P2 resulted in a noncovalent complex with binding energy comparable to those of covalent complexes and distances comparable to those of covalent bond(s) to the nitrogen, methoxy oxygen and a sulfonic oxygen. SH-ate and PR-ate bind Fe 2+ at P1 with a covalent bond to the carboxyl oxygen. Similarly, MG-ate, PT-ate, M2G-ate bind Fe 3+ at P2 (for M2G, P1 = P2) with a covalent bond to the carboxyl oxygen and a second covalent bond to the nitrogen on C3 (see third column in Fig. 5). MT-ate binds Fe 3+ at P2 with two covalent bonds to sulfonic oxygens and a third covalent bond to the nitrogen on C3. EU-ate binds Fe 3+ at P2 with a covalent bond to the carboxyl oxygen. SH-ate, and PR-ate bind Fe 3+ at P1 with a covalent bond to the carboxyl oxygen and a second covalent bond to the nitrogen on C1. All attempts to covalently bind iron at P3 failed (except for [Fe(MG)] 3+ ) for MG-ate, MT, MT-ate, PT, PT-ate. Therefore, we did not search further at P3 for the rest of the MAAs.
These observations show that the best position for iron binding is P2. The carboxylic acid (-COOH or -COO − ) group of the MAA (or MAA-ate), the methoxy oxygen and the nitrogen on C3 act as three tethers for chelation when Fe 2+ is at P2. The methoxy oxygen seems to play a role through Coulombic forces. Unlike the rest of the series, P1 is the best binding position for SH, SH-ate, PR and PR-ate. The reason may be the lower pKa of the serine and the threonine moieties at P1 compared to pKa of glycine at P2 and/or the presence of the -OH group on serine and threonine acting as a fourth tether at P1. Coulombic forces with the methoxy oxygen seem to be influential both at P1 and P2 for Fe 3+ binding as well.
To summarize and generalize the gas phase calculations for iron complexes, 4DG, and all the MAAs studied bind Fe 2+ both in the neutral form and the enolate form (first and second columns in Fig. 5). They bind Fe 3+ only in the enolate form (only MG binds Fe 3+ in the neutral form too). All the MAAs studied bind iron with three tethers at P2 except SH and PR. SH and PR have amino acids substituted at C1 which are more acidic than glycine and have an extra -OH group on the amino acid that may act as a fourth tether at P1 (third column Fig. 5).
The effect of water on chelation resulting from PCM calculations show that neither any of the MAAs nor their enolate (MAA-ate) forms bind Fe 2+ covalently. All the complexes shown in the first and second columns in Fig. 5 became non-covalent chelates upon running PCM calculations. These ionic complexes have iron located at almost the same location with slightly longer bond lengths and less metal affinity. Nevertheless, all MAA enolate forms (except 4DG) bind Fe 3+ covalently. M2G-ate (P2 = P1) and PT-ate bind Fe 3+ at P2 with a covalent bond to nitrogen on C3 and another covalent bond to carboxyl oxygen; SHate and PR-ate, bind Fe 3+ at P1 with a covalent bond to nitrogen on C1 and another covalent bond to the carboxyl oxygen. EU-ate binds Fe 3+ at P2 with a covalent bond to the carboxyl oxygen. The last column in Fig. 5 undergoes changes only for MG-ate and MT-ate complexes. MG-ate binds Fe 3+ at P2 only with a covalent bond to the carboxyl oxygen and MT-ate binds Fe 3+ at P2 only with a covalent bond to the sulfonic oxygen. The order of affinities for ferric ion is MT-ate < MG-ate < PT-ate ≅ M2G < SHate < PR-ate < EU-ate. All the Fe-N and Fe-O distances display an increase except the Fe-O distance of the MT-ate complex for which the complexation has occurred from only one sulfonic oxygen therefore with a shorter distance (see PCM calculation results in Table 1).
The results of the second round of calculations with double diffusion functions displayed covalent ferric ion binding to the carboxyl oxygen (sulfonic oxygen for MT-ate) for all MAA-ate forms both in gas phase and water (except EU-ate). EU-ate does not bind Fe 3+ covalently in water, but interestingly has the highest binding affinity (Table 2). SH-ate and PR-ate complexes have the preference for P1 in gas phase, but they bind iron only at P2 in water. The order of affinities for ferric ion is MG-ate ≅ MT-ate < PTate < M2G-ate < SH-ate < PR-ate < EU-ate in gas phase and MT-ate < MG-ate < PT-ate ≅ M2G-ate < PR-ate ≅ SHate < EU-ate in water (Table 2).
Many bacteria in anaerobic environments couple organic carbon oxidation to reduce Fe 3+ to Fe 2+ , in other words, they respire iron. Free ferrous and ferric iron at physiological pH have very low solubility, therefore highly soluble iron chelators become important for such organisms when iron in the medium is scarce and/or iron levels fall intracellularly. These bacteria synthesize and excrete iron chelators into the medium to acquire ferric iron. After the iron transport, the electron flows through the microbial electron transport system where the final enzyme is ferric reductase, and the reduction causes the ferrous iron to be released (Neilands 1981). The solvent effect that we have observed for the MAAs studied fits very well with this scenario in which the [Fe(MAA-ate)] 2+ covalent complex carrying a ferric ion may release, after microbial reduction, the ferrous ion which becomes a noncovalent complex.
Our results support Volkman et al.'s (2006) suggestion that Euhalothece-362 may act as an iron chelator and Tarasuntisuk et al's (2018) observed chelating activity for M2G.

Calcium complexes
The results and data for the optimized structures of the most stable Ca 2+ complexes are presented in Table 3 and their 2D representations are shown in Fig. 7. The calcium complexes are non-covalent complexes. MG chelates Ca 2+ at P2 in close proximity to nitrogen, methoxy oxygen and carboxylic oxygen. M2G (P1 = P2) chelates Ca 2+ exactly like MG. SH and PR chelate Ca 2+ at P1 in between the nitrogen, methoxy oxygen, carboxylic oxygen and the hydroxyl group on serine or threonine, respectively (first column in Fig. 7). The Ca 2+ complexes of their enolate forms, MG-ate, M2G-ate, SH-ate, and PR-ate (second column in Fig. 7) have the same structures as their neutral forms. The effect of the fourth tether observed for iron complexes of SH and PR is observed for the calcium complexes as well.  MG-ate, M2G-ate, SH-ate, and PR-ate all chelate Ca 2+ at P3 by four oxygen atoms, namely the two oxygens of the carboxyl group and the two oxygens of the hydroxy and methylhydroxy groups on C5 (third column in Fig. 7). The common moiety in the pool for those tested at P3 is MG-ate. Comparison of iron chelating to calcium chelating behavior of MG-ate reveals that P3 may be preferable to P1/P2 for non-covalent chelation. The order of affinities for Ca 2+ are MG ≅ M2G < PR < SH and MG-ate < M2G-ate < SHate < PR-ate. Our results support the chelating ability of MAAs referred to by Orfanoudaki et al. (2020).

MAAs on early earth
The outcome of this computational research is highly relevant to life on early earth and hence to astrobiology. CB and actinobacteria share a common ancestor and predate GOE (Battistuzzi et al. 2004). MAAs could have been synthesized by their most ancient strains and could have served for iron acquisition during anoxic times. CB, being the first taxa to adapt to the oxidative stress with existing systems (Fischer et al. 2016) could have used MAAs as antioxidants as well as metal chelators. Photoferrotrophs' mechanisms preventing cell encrustation with oxidized iron (Miot et al. 2009;Saraiva et al. 2012;Wu et al. 2014) may have been due to extracellular MAAs. MAAs have the potential to serve bacteria as iron-chelators for the purpose of screening UVR (as colloid, on extracellular polymeric saccharides, or intracellularly). Furthermore, bacteria may also potentially utilize MAAs intracellularly or extracellularly for iron transport, solubilization, mobilization and storage purposes.

Conclusions
We discuss the potential iron chelating ability for selected MAAs in relation to their molecular structures investigated by DFT calculations first in gas-phase and then by inclusion of solvent (water) effect. Our gas phase results show all the MAAs studied bind Fe 2+ covalently both in the neutral and the enolate forms (first and second columns in Fig. 5). They bind Fe 3+ covalently only in the enolate form (third column in Fig. 5) except MG which binds Fe 3+ covalently in the neutral form too. Results of the PCM calculations show that all the Fe 2+ complexes both in the neutral and the enolate forms are non-covalent chelates while the Fe 3+ complexes bind covalently except for EU-ate which forms an ionic complex with high affinity ( Table 2). The preferred position for binding iron is P2.
Calcium MAA complexes are non-covalent. MG, M2G and their enolate forms chelate Ca at P2 with three tethers, SH, PR and their enolate forms chelate Ca at P1 with four tethers similar to the observations obtained for iron complexes. MG-ate, M2G-ate, SH-ate, and PR-ate all chelate Ca 2+ at P3 by four oxygen atoms. Comparison of iron chelating to calcium chelating behavior of MG-ate reveals that P3 may be preferable to positions P1/P2 for non-covalent chelation.