Synthesis and biological activity of iron(II), iron(III), nickel(II), copper(II) and zinc(II) complexes of aliphatic hydroxamic acids

Abstract Aliphatic hydroxamic acids (HA) with varying numbers of carbon atoms, C2, C6, C8, C10, C12 and C17, and corresponding Fe(II), Fe(III), Ni(II), Cu(II) and Zn(II) complexes have been synthesized and characterized by various methods, including structural determination by single crystal X-ray diffraction and theoretical calculations. The biological activities of HA and their complexes have been assessed on a panel of pathogens, including eight bacteria strains and one fungus. The C12 aliphatic HA displayed the lowest minimum inhibitory concentrations (MIC) towards several microbial strains and a selective antifungal activity. This antifungal selectivity was further improved considering the Ni(II), Cu(II) and Zn(II) complexes of C12 HA showed higher IC50 values, thus less impact on the SiHa human cell line viability. This work warrants further investigation to understand the underlying mechanism of action and potential biological applications of HA and the derived metal complexes.


Introduction
The first hydroxamic acids (HA) were discovered by Lossen in 1869 (oxalohydroxamic: HONH-CO-CO-NHOH), but this class of organic compounds was only thoroughly studied from the nineteen eighties [1]. HA are among some of the most well studied compounds due to their significance in so many different applications in modern society [2].
HA also display antimycobacterial activity against Mycobacterium tuberculosis [14] and Mycobacterium smegmatis [15]. They are powerful inhibitors of enzymes such as hydrolase [16], urease [17], thermolysin [18], aminopeptidase [19,20], tyrosinase [21] and matrix metalloproteinases (MMP) [22] by chelation of metal co-factors, in particular Fe(III) and Zn(II) [23,24]. Desferrioxamine is used clinically to treat metal poisoning, highlighting the strong chelating properties of the HA group [24,25]. Systematic biological evaluation of large series of such compounds however remains scarce and the variability of methods used in previous work renders global interpretation of the results rather tedious. Therefore, a comparative study of the biological activity of HA with various alkyl chain lengths in combination with various metal centers could help draw conclusive results.

Chemical and physical methods
All the syntheses of HA were followed by TLC on GF 254 silica chromatographic plates. The revelations were made by a UV lamp (254 nm), iodine or a solution of phosphomolybdic acid 10% in ethanol. The solvents were evaporated using a rotary evaporator. 1 H and 13 C NMR spectra were recorded with a JEOL JNM-ECZ 400 R/S3 type NMR spectrometer (Tokyo, Japan), respectively, at 400 MHz and 100 MHz by dissolving about 10 mg of product in 500 mL of deuterated solvents (DMSO-d 6 or CDCl 3 ). The chemical shifts are given in parts per million (ppm) using the residual solvent peak as reference (2.50 and 7.26 ppm for DMSO-d 6 and CDCl 3 , respectively), and the coupling constants are expressed in Hertz (Hz). The IR spectra were recorded on a Perkin Elmer FT-IR-MIR/NIR (Beaconsfield, UK) spectrophotometer equipped with an ATR Diamond KRS-51 module using 2-3 mg of each product within a wavenumber range 400-4000 cm À1 . Raman spectra were taken with a Bruker FT-Raman spectrometer RFS 100/S (Massachusetts, USA) equipped with a Michelson interferometer for Fourier transform. The source is a near-IR Nd 3þ : YAG laser at 1064 nm wavelength, while the Ge detector is cooled with liquid nitrogen. The Raman spectra were recorded with power ranging from 50 to 250 mW depending on the nature of the sample, as a compromise between best quality spectra and decomposition of the sample under the laser beam. The highest possible power was used for each sample by increasing gradually the power within this range during each recording. The solid samples were introduced into the cavity of a small metallic sample holder using a small funnel and pestle. The resolution was fixed at a value of 4 cm À1 , and for each sample 48 scans were recorded in the 0 to 4000 cm À1 range. Elemental sulfur is used as a reference to verify all experimental parameters before analyzing research samples. The UV spectra were recorded with a spectrophotometer Shimadzu UV-1800 (Kyoto, Japan). 5 mM solutions were prepared by dissolving the appropriate amount of product in MeOH. Then 100 lL of this solution were taken and diluted to 5 mL with MeOH (final solution at 100 lM). With Cu(II), Zn(II) and Ni(II) complexes, a second dilution was carried out to have a final concentration of 50 lM. The mass spectra (MS) were obtained with a Q-TOF-6520 spectrometer from Agilent (Santa Clara-California, USA). The MS of complexes are given with a m/z range from 100 to 1700 in positive mode. MS were obtained by dissolving 1-1.5 mg of complexes in 1 mL of MeOH, then diluted 1/100 in a mixture of 0.2% formic acid in water -MeOH (2: 8). Electron paramagnetic resonance (EPR) spectra of some Fe(III) and Cu(II) complexes and some HA were recorded on a Bruker Magnettech ESR5000 CW benchtop spectrometer operating at 9.4 GHz (X-band). Spectra of the Fe(III) complexes were recorded at 93.15 K and, for HA, at room temperature. EPR measurements of the Cu(II) complexes were not recorded (C % 0.3 mol/L, insufficient solubility in CHCl 3 for the EPR analysis). For these measurements, 20 lL of a freshly prepared solution of the HA (C % 1 mol/L) or Fe(III) complexes in CH 3 OH (C % 0.1 mol/L) was filled in a 3 mm EPR tube.
Elemental analyses have been determined on a Perkin Elmer 2400 elemental analyzer equipped with a sixty (60) sample autosampler. The results processing system were managed by EA 2400 data manager software. Melting points (uncorrected) were determined on a B€ uchi B-545. Stability of complexes in water was assessed by UV spectroscopy in the range 200-800 nm [32]. Methanolic solutions of complexes (20 mM) were further diluted to 1 mM with the same solvent. Subsequently, MilliQ water was added to obtain 50 and 100 mM solutions (5 and 10% MeOH). The UV spectra were recorded on a spectrophotometer Cary 60 (Agilent), including one spectrum each 5 min during 30 min and one spectrum each 30 min during 24 h.
The amount of water in the complexes was determined according to the Karl Fischer method with a 756 KF Coulometer from Metrohm (Antwerpen, Belgium). 25.0 mg of complexes were weighted into an Eppendorf tube and dissolved in 500 mL of MeOH. Then, 200 mL of this solution was injected through the septum of the device via a syringe. Two reference salts were used to validate this experiment (H 2 C 2 O 4 Á2H 2 O, MgSO 4 Á7H 2 O). All experiments were repeated twice for each sample. A blank analysis (methanol) was done for each sample. The complexes were kept 24 h in a dry atmosphere before analysis to avoid hygroscopic water.
X-ray diffraction data were collected on a MAR345 image plate detector (Hamburg, Germany) using Mo Ka radiation generated by a Rigaku Ultra X18S rotating anode (Tokyo, Japan). Crystals for structure determination were obtained for the Fe(III) complexes, giving red-orange plate like crystals after evaporation of a MeOH/DCM solution (1: 9). The collected images were integrated and reduced by CrysAlis PRO and the implemented absorption correction was applied. Structure was solved by SHELXT and refined by full-matrix least squares against F 2 using SHELXL 2014/7. Quantum mechanical methods for density functional theory (DFT) calculations have been achieved using Orca 4.2.1 [33], performed in the gas phase. Geometries were fully optimized with the triple-f quality, low-cost method B97-3c from Grimme and co-workers [34]. Potential energy surface minima found upon optimization were confirmed by frequency calculations and free energies were corrected to account for the zero-point energy. Optimized geometries were verified as minima (i.e. no imaginary frequencies).
The cyclic voltammetry (CV) experiments were performed with an Epsilon potentiostat BASi (west Lafayette, USA) connected to a carbon screen-printed electrode (DRP-110) from Dropsens (Asturias, Spain). The cSPE constituted of a miniaturized cell with 3 electrodes all in one includes a carbon working electrode, a carbon counter electrode and a silver pseudo reference electrode. The software Epsilon-EC-version 1.50.69_XP was used for instrument control. For CV experiments, stock solutions of the samples (HA, salts and complexes) were prepared in MeOH to obtain a concentration of 10 mM. These solutions were subsequently diluted with a 1 M KCl aqueous solution as support electrolyte and with 5% MeOH (in order to avoid any risk of reprecipitation of samples in aqueous solution) to a final concentration of 100 lM. A volume of 50 mL of this solution was carefully deposited on all surfaces of the miniaturized cSPE. Between each CV measurement, a new cSPE was used to avoid the so-called "memory" effects and solutions were deoxygenated with nitrogen for at least 5 min before each measurement to ensure no interference with dissolved oxygen. The potentials ranged from either À0.2 V or À0.4 V to þ1.2 V. Then, the scanning was reversed toward the initial value. The cyclic voltammograms were carried out in one scan and multiple scans (10 cycles) at a scanning rate of 50 mV/s. Partition coefficients (logP) were determined for all synthesized compounds [32]. 10 mg of each compound were partitioned between 1 mL of n-octanol and 1 mL of water. After stirring for 1 h, 25 mL were taken in each phase and then diluted to 2.0 mL with the corresponding solvent. The absorbances of the different solutions were measured in the organic and aqueous phases by UV spectroscopy (Ɛ in both solvents were not significantly different). The spectra were recorded in the range 200-700 nm using a Shimadzu UV-1800 instrument. For each compound three measurements were performed. The concentrations of compounds dissolved in water (C water ) and octanol (C octanol ) were calculated by applying the Beer-Lambert law (Eq. 1). The logP was calculated for each experiment (Eq. 2). The average logP was obtained from three independent experiments.
Magnetic measurements were carried out in a superconducting quantum interference device magnetometer (Quantum Design SQUID VSM MPMS3). Hysteresis loops were measured on dried samples in plastic capsules at 300 K with a maximum field of 10 kOe. The magnetization was normalized to the powder mass determined by subtracting the weight of the capsule from the total weight of the sample. A reference measurement using a Pd reference sample was conducted to quantify the remanent moment from the superconducting magnet, which allowed correction of the field values in the hysteresis loops. A linear diamagnetic background contribution from the sample holder was subtracted from the magnetic moment data of the complexes ( Figure S19). The effective magnetic moment (m eff ) which gives the effective number of l B per molecule was calculated using the following equation [35]: v M : molar magnetic susceptibility ¼ mass susceptibility Â molar mass, T: temperature (300 K) and lB is the Bohr magneton.
Here v M is the value of the molar magnetic susceptibility in units of m 3 mol À1 which equals the mass susceptibility multiplied by the molar mass, T is the value of the temperature in units of K, and lB is the Bohr magneton.
After measuring the electrical conductivity of the compounds using the TetraConV R 325 conductivity meter, the molar conductance was calculated (Eq. 4). Stock solutions at different concentrations were prepared in 2 mL of MeOH (C 0 ) and then diluted 20 times in the same solvent. MeOH (blank) was used as reference.

Biological methods
. Antimicrobial assay. The antibacterial, antimycobacterial and antifungal activity assays were performed using the microdilution method [36]. The stock solutions were at 200 mM in DMSO for HA, Fe(II/III) complexes (HAnFe2, HAnFeCl and HAnFe3) and metal chloride (FeCl 2 , FeCl 3 , CuCl 2 , ZnCl 2 and NiCl 2 ) and in MeOH for Cu(II), Zn(II) and Ni(II) complexes (HAnCu2, HAnZn2 and HAnNi2). Volumes of 5 mL and/or 25 mL of stock solution were added to the MHB to a final volume of 1000 mL to obtain a final concentration of 1 mM and/or 5 mM (solutions containing 1% and/or 5% of DMSO or MeOH, v/v). The aqueous stock solutions of antibiotics/antiseptic were 69 mM (100 mg/mL), 0.6 mM (0.2 mg/mL) and 12 mM (10 mg/mL) for vancomycin, cetrimide and rifampicin, respectively. These solutions were subsequently diluted in MHB to achieve a final concentration of 7 mM (10 mg/mL), 6 mM (2 mg/mL) and 12 mM (10 mg/mL), respectively. These samples were subsequently diluted within 2-fold dilution series in 96-well plates giving a final volume of 100 mL. Compound concentrations were rounded to the nearest hundredth.
The precultures of bacteria, mycobacteria and fungi were diluted to obtain a turbidity of 0.5 McFarland corresponding to approximately 1.5 Â 10 8 colony forming units (CFU) per mL measured with a nephelometer. Then the diluted microbial suspensions were further diluted 100 times to achieve 1.5 Â 10 6 CFU/mL and 100 mL of bacteria/mycobacteria/fungi were inoculated into 96-well plates containing 100 mL of diluted HA, complexes, salts or antibiotics/antiseptic. The plates were incubated at 37 C for 24 h (bacteria), at 30 C for 48 h (mycobacteria), at 30 C (B. subtilis) for 48 h, and at 22 C (C. albicans) for 48 h.
The next day, 30 mL of MTT dissolved in MilliQ water (80% m/v) were added to the plate and incubated for 30 to 60 min. The minimum inhibitory concentrations (MIC) were obtained by visual observation and corresponded to the lowest concentration of drug/antibiotic/antiseptic able to inhibit 100% microorganism growth compared to control wells without drugs. In addition, the minimum bactericidal/minimum fungicidal concentration (MBC/MFC) were determined by transferring well samples with drug concentrations ! MIC on TSA plates. MBC/MFC represented the lowest concentration of tested drug/antibiotic/antiseptic required to kill 99.9% of microorganisms, recorded by visual analysis on the second or third growth day, when untreated bacteria were giving colonies [37]. If the MBC/MIC or MFC/MIC ratios were 1 or 2, the effect was considered as bactericidal/fungicidal. In contrast, if this ratio was equal to 4 or 16, the effect was defined as bacteriostatic/fungistatic [38].
2.2.2.2. Cytotoxicity assay. The cytotoxicity assays on SiHa cells were carried out as previously described [39]. For 24 h, the SiHa cells were grown in Dulbeccos's Modified Eagle Medium (DMEM) with 10% (v/v) fetal bovine serum (FBS) at 37 C with 5% CO 2 . The cells were diluted to a concentration of 7.6 Â 10 5 cells/mL and further grown for another 24 h at 37 C in humidified atmosphere containing 5% CO 2 in 96-well plates (100 mL/well). The drug stock solutions were 20 mM in DMSO (HA6Fe3, HA8FeCl, HA8Fe3, HA10FeCl, HA10Fe3, HA12, HA12Fe2, HA12FeCl, HA12Fe3) or in MeOH (HA12Cu2, HA12Zn2, HA12Ni2) and diluted hundred times in DMEM to a final concentration of 200 mM (0.2% solutions of DMSO/MeOH). These samples were subsequently diluted in DMEM within 2-fold dilution series in 96-well plates giving a final volume of 100 mL. A positive control was set up without drugs. The plates were incubated at 37 C in a humidified atmosphere containing 5% CO 2 for 72 h. Then, 20 mL MTT (0.5 mg/mL in RPMI-1640) were added to each well and after 4 h incubation at 37 C with 5% CO 2 , the supernatant was removed and cells were washed with phosphate buffered saline (PBS) several times. Subsequently, 100 lL DMSO were added to each well and shaken for 10 min. The absorbances at 570 nm (the maximum formazan absorbance wavelength) and 630 nm (the background noise wavelength) were recorded using a spectrophotometer (Synergy HT, Bio-Tek). The average of the relative activity was calculated from the difference of the absorbances obtained at 570 and 630 nm for each test. The raw data were transferred to a Microsoft Excel sheet and analyzed. The percentage inhibition of cells using Eq. (5) and the cell viability was calculated according to Eq. (6) [40,41].
A 0 : absorbance of the cells without drug (the positive control) and A 1 : relative absorbance of the cells with drug.
The 50% inhibitory concentration (IC 50 ) values were also measured. The reduction of cell viability at each concentration was plotted as a dose response curve. The IC 50 values were calculated from the dose response curves by nonlinear regression to fit data to the cell viability (y)decimal logarithm of the concentration (logC) [42]. The selectivity index (SI) was calculated by Eq. (7). A SI greater than or equal to 10 indicates that the test compound can be applied at a concentration that is ten-fold higher than the MIC value without exhibiting cytotoxicity [43].
3. Statistical analysis. The antimicrobial results (MIC) were analyzed using t-test (Excel V R ) at the statistical significance level of p 0.05. Two sets of data were used for this purpose (bilateral distribution). The first series concerns the MIC results obtained with HA and complexes and the second concerns the MIC obtained with vancomycin, cetrimide and rifampicin from the three independent experiments carried out (see SI).

Formulas
Codes of HA and complexes with carbon chains C 2 , C 6 , C 8 , C 10 , C 12  Zn(II), Ni(II), di-hydroxamato chloride Fe(III) or tri-hydroxamato Fe(III) complexes (Scheme 2). Minor impurities remained in some compounds, as evidenced by elemental analysis (see Appendix B). However, a further purification step of selected complexes (see section 3.3.1. Antimicrobial effect) led to identical physico-chemical data.

Characterization data
The HA and their complexes were obtained in 60 to 86% yields ( Figure S1) and characterized by FTIR, stability analyses, mass spectrometry, Raman, UV-Vis, 1 H/ 13 C NMR and EPR spectroscopic methods (Figures S2-S14). KF titration was used for water determination and CV for redox properties. Single crystals suitable for X-ray structure determination were only obtained for HA8Fe3. Melting points and elemental analyses are given in table S1 and chloride ions of Fe(III) complexes (HAnFeCl) were identified by precipitation of AgCl using AgNO 3 ( Figure S15).

IR spectroscopy
The IR absorbances are shown in Table S2 (for spectra, see Figure S2). Bands of medium (m) and sharp (sh) intensity in the 3200 cm À1 region can be attributed to the stretching vibration of the O-H group. Medium to strong (s) bands that are specific of the N-H bond stretch are found around 2800 cm À1 . The strong bands at 1650 cm À1 are characteristic of the carbonyl group (C ¼ O), while bands at 1560, 1470 and 1370 cm À1 can be assigned to elongation of the C-N bond.
The main spectral change in the complexes is the disappearance of the hydroxyl m(OH) peak due to the replacement of the hydrogen atom by transition metal ions. IR spectra of the complexes HAnFe2, HAnFeCl and HAnFe3 showed absorption bands between 3160-3200 cm À1 . The bands at 1520 and 1595 cm À1 arise from the carbonyl group (C ¼ O), while signals between 1320 and 1470 cm À1 come from the C-N bond. Finally, strong absorption bands observed in the low energy region at 520 and 438 cm À1 are attributed to the Fe-O stretching vibrations. The Cu(II), Zn(II) and Ni(II) complexes displayed characteristic bands located between 3190-3250 cm À1 . Unlike HAnFe2, HAnFeCl and HAnFe3 complexes, HAnCu2, HAnZn2 and HAnNi2 complexes show narrower stretching bands. The medium to strong intensity resonances around 2900 cm À1 arise from stretching of the N-H bond. The strong bands at 1539 and 1599 cm À1 are attributed to the carbonyl group (C ¼ O). Between 1340 and 1470 cm À1 are found the frequencies from the C-N bond. Finally, medium to strong absorption bands at 550-490 cm À1 could be attributed to the Cu-O, Zn-O and Ni-O bonds. Raman spectra for all HA, HAnFe2, HAnFeCl, HAnFe3 and HAnZn2 complexes are shown in Figure S3.

UV-Vis spectroscopy
Comparison of the UV-Vis spectra from HA10Fe2, HA10FeCl and HA10Fe3 helped attributing the HA-metal stoichiometry ( Figure 1A). The maximum of absorption for the solutions of these complexes is found at 425 nm, with different extinction coefficients depending on the oxidation state (Fe(II) or Fe(III)). The highest Ɛ is observed for HA10Fe3 (Ɛ ¼ 2350 M À1 cm À1 ), followed by that of HA10Fe2 (Ɛ ¼ 1600 M À1 cm À1 ), HA10FeCl having the lowest one (680 M À1 cm À1 ). The UV-Vis spectra of HA8Cu2, HA8Zn2 and HA8Ni2 ( Figure 1B) show maxima of absorption around 250 nm with the highest Ɛ obtained for the HA8Cu2 (7300 M À1 cm À1 ), followed by the HA8Ni2 (Ɛ ¼ 2280 M À1 cm À1 ) and HA8Zn2 (Ɛ ¼ 1040 M À1 cm À1 ). Similar results were observed under the same conditions with the other complexes having the same type of coordination metal ( Figure S4).

EPR spectroscopy
EPR spectra have been recorded for Fe(III) complexes in MeOH solution at 0.1 mol/L (HA6Fe3, HA8Fe3 and HA12Fe3). A broad signal is observed under the magnetic field strength 165 mT for the three complexes ( Figure S13). As expected, compounds HA6, HA10 and HA12 do not show any signal in EPR (data not shown).

Karl Fischer titrations
The water content was determined by the KF coulometric method. All HA2-containing complexes HA2Fe2, HA2FeCl, HA2Fe3, HA2Cu2, HA2Zn2 and HA2Ni2 were all obtained without water molecules. The Fe(II/III) complexes with an aliphatic chain longer than C 2 comprised one water molecule. Their Zn(II) and Ni(II) homologues crystallized with 0.5 water molecule. No water molecules were found in Cu(II) complexes. Therefore, the water content in these compounds varied depending on the type of metal but also the synthetic pathway.
The oxidation peak tends to disappear when multiple scanning is carried out. During the reverse scan, a reversible redox couple appeared with a broad reduction peak (Ec 2 ) starting at À0.2 V and a broad oxidation peak (Ea 2 ) at þ0.4 V. These were also observed when starting at 0.0 V towards negative values, indicating a reduction. It should be noted that the intensity of this redox couple (Ea 2 /Ec 2 ) increased slightly with multiple scanning (Figure 3A-D).
A reversible redox process at the metal center was observed for Fe(II/III) and Cu(II) chlorides. However, Zn(II) and Ni(II) chlorides remained inert ( Figure S5). For all Fe(II/III) complexes, a quasi-reversible peak within À0.2 V to þ0.5 V was found. These processes  are identified by a red circle ( Figure 4A-C). No redox peak at À0.2 V/þ1.2 V was observed from the Cu(II) complexes ( Figure 4D). While Zn(II) and Ni(II) complexes showed two broad redox peaks, i.e. a reduction starting at À0.2 V and an oxidation at þ0.4 V, their intensities differed ( Figure 4E,F).

Partition coefficients
The observed partition coefficients (Log P) were measured for all compounds and tend to increase with chain length ( Figure S6).

X-ray diffraction
Attempts to obtain suitable crystals for X-ray diffraction analysis were unsuccessful for all complexes except for HA8Fe3, crystallographic parameters and refinement data are listed in Table 2. All heavy atoms were refined anisotropically and H-atoms were placed  in calculated positions with isotropic temperature factors fixed at 1.2 Ueq of the parent atoms (1.5 for methyl and O-H hydrogens). Although high resolution data were recorded, a resolution limit of 1.15 Å was imposed during refinement, beyond which poor data statistics were observed. The mobility of the alkyl chains is visible, showing an increase of the thermal ellipsoids moving along the alkyl chains towards the methyl ends. Rigid bond restraints on the thermal ellipsoids were applied on the last five atoms of all three alkyl chains. For every alkyl chain, the bond distances were restrained to be similar, no bond angle or torsion angles were restrained. One lattice water molecule is present ( Figure 5) which corroborates the results obtained by KF analysis. Water stabilizes the molecular packing by hydrogen bonding between neighboring complexes ( Figure S7.A). The hydrogen atoms of this water molecule were first located in the density maps, after which the water molecule was idealized and refined as a rigid group allowed to rotate around the oxygen atom. The conformation of each alkyl chain is different from the two others. A first one is fully elongated with all torsion angles eclipsed. A second chain starts with a gauche conformation, with the remaining torsion angles eclipsed. The third chain has three torsion angles in gauche conformation of which two are found consecutively. The hydrogen bonds propagate along the b and c axis forming a sheet-like structure, flanked by the apolar alkane chains ( Figure S7.B).
A search in the CSD database reveals only one related structure (REFCODE SUXREI) [27]), equally a Fe complex with the shortest possible alkyl chain (methyl). The HA8Fe3 complex is arranged as a distorted octahedron (distortion parameter R ¼ 80.6 ) in the mer configuration, while in the crystal structure of SUXREI both the fac and mer configurations are present. The octahedral geometry of HA8Fe is comparable to the mer isomer of SUXREI (R ¼ 86.3 , RMS deviation is 0.124Å 2 for the metal ion and the first coordination sphere, 0.217 Å 2 for all atoms in common ( Figure S8).
Since crystals obtained for X-ray diffraction analysis exhibited no suitable diffraction patterns (except for HA8Fe3), DFT calculations were employed to uncover the local Figure 5. Thermal ellipsoid plots of complex HA8Fe3. All ellipsoids drawn at the 50% probability level. Colors are as follows: black, C; blue, N; red, O; grey, H; dark blue, Fe. CCDC 2079080 contains the supplementary crystallographic data for this paper. These data can be obtained free of charge from the Cambridge Crystallographic Data Centre via www.ccdc.cam.ac.uk/structures. geometry around the metal centers. The DFT-optimized structures showed tetrahedral geometries for HA12Fe2, HA12Cu2 and HA12Zn2, and a square-planar geometry for HA12Ni2 ( Figure S9).
CCDC 2079080 contains the supplementary crystallographic data for this paper. These data can be obtained free of charge from the Cambridge Crystallographic Data Centre via www.ccdc.cam.ac.uk/structures.

Stability
Stability analyses of HA by 1 H NMR showed no significant spectral modifications. Only NH and OH peaks disappeared for compounds HA2 and HA6 after 24 h ( Figure S10). This should be due to proton exchanges of NH and OH with water in DMSO-d 6 . Stability assays of complexes were monitored by UV-Vis spectroscopy. Compounds were dissolved in MeOH and water was added to reach the desired concentration for obtaining proper spectra. Significant changes with isosbestic points around 400 nm were observed after 10 minutes for HA2Fe3 ( Figure S11 With other compounds, increased absorptions occurred during the test. This could be interpreted as the resolubilization of precipitated complex from the MeOH solution upon water addition rather than a degradation process. These results showed that the stability of compounds is influenced by the type of coordination metal and the length of the alkyl chain. The change in absorbance over time at maximum wavelengths of HA6Fe3 showed hypsochromic and hyperchromic effects (k at 350 and 380 nm). The appearance of isosbestic point after ten minutes could be due to an equilibrium between two species (Figure 7).

Magnetic susceptibility
Magnetic susceptibilities of the complexes were examined by measuring the magnetization of a dried sample at 300 K in a SQUID magnetometer (Table 3). All compounds show paramagnetic behavior at room temperature, thereby variation in the mass susceptibility can be ascribed to the differences in the paramagnetic part in agreement with the nature of the compounds ( Figure 8A-D).

Antimicrobial effect
The MIC and the MBC/MFC were determined using Gram positive bacteria (methicillinsensitive/resistant S. aureus, C. glutamicum, B. subtilis), Gram negative bacteria (E. coli, P. aeruginosa, K. pneumoniae), mycobacteria (M. smegmatis) and fungi (C. albicans). The concentration ranges of tested samples were between 16 mM À 3 mM. The antibacterial activity has been defined as significant when the MIC is below 20 mM (between 2-20 mg/mL depending on the compound), moderate when 20 < MIC < 200 mM, low when 200 MIC <500 mM and insignificant when MIC > 500 mM, range in mg/mL depending on the chemical compound [49] (Table S3).
Selected compounds were further evaluated for bactericidal/fungicidal activities (Table S4 and Figure S17). Their MFC or MBC were generally obtained at concentrations reaching two-three times the MIC. The fungicidal activity (MFC of 31 mM) of HA12 and HA12Zn2 on C. albicans was particularly relevant.
The most active compounds HA12, HA12Fe2, HA12FeCl, HA12Fe3, HA12Cu2, HA12Zn2 and HA12Ni2 were further purified and re-tested (see Appendix A). Results of biological re-evaluations were similar to those obtained during the initial screening.
The SI of the antimicrobial activities of these compounds with regards to their human cell line cytotoxicity were determined by calculating their IC 50 /MIC ratio. Compounds having a high IC 50 and a low MIC were considered the most interesting ones, as they show together low cytoxicity and high antimicrobial activity. The ratios were rather similar when considering the antimicrobial activities against S. aureus MSSA, C. glutamicum, B. subtilis, E. coli and M. smegmatis strains (0.26) ( Table 6). The best antimicrobial selectivities were observed for HA12Cu2 (SI ¼ 5.11), HA12Zn2 (SI ¼ 3.74) and HA12Ni2 (SI ¼ 1.60) on C. albicans (Table 6), having higher IC 50 compared to their MIC.

IR spectroscopy
Spectral analyses confirmed the identity of the HA and their complexes. IR spectra revealed the presence of HA in their deprotonated form in the complexes. Water molecules within the coordination sphere of Fe(II/III) complexes was identified by broad bands in the region 3450-3250 cm À1 due to O-H stretch [9]. With Zn(II) and Ni(II) complexes, the bands observed in the same region are very weak and even absent for Cu(II) complexes, which corroborate the results of the water determination by the KF titration. A sharp band at 1619-1662 cm À1 with HA, in the region 1520-1595 cm À1 for the Fe(II/III) complexes, and at 1539-1599 cm À1 for those of Cu(II), Zn(II), Ni(II) is attributed to the shifted vibration frequency of C ¼ O upon complexation. This band undergoes a shift to lower wavenumber as described elsewhere [31,48], which is consistent with the chelation by the carbonyl oxygen atom. The C-N bands are found at 1459-1499 cm À1 upon complexation. Regarding the M-O stretching vibrations we observed several bands between 400 and 600 cm À1 , depending on the metal, that confirmed the chelated structure [31,48,[50][51][52].

UV spectroscopy
In the UV spectra (Figure 1), a slight bathochromic and medium hyperchromic displacement characteristic of the tri-coordinated Fe(III) complex (HA10Fe3) appears compared to the di-coordinated Fe(II) complex (HA10Fe2, DƐ ¼ 750 M À1 cm À1 ). Between the complexes of HA10FeCl and HA10Fe3, a large hyperchromic displacement of  - 60 50 1670 M À1 cm À1 is observed. This variation of Ɛ is influenced by the nature of the coordinating metal. Moreover, the variation of absorbance is an important factor in determining the stoichiometry of the complexes. The wavelength corresponding to the maximum absorption depends on the mode of coordination of the metal ion and stoichiometry (1 metal for 2 HA or 1 metal for 3 HA) and is independent of the nature of the HA [53]. The electronic transitions n!p Ã (C ¼ O) observed show a bathochromic shift towards a higher wavelength. This indicates that complexation has occurred between HA and the corresponding metal chloride [54]. The absorption maximum at 425 nm characterizes the ligand-metal charge transfer bands of complexes HAnFe2, HAnFeCl and HAnFe3 [55], those of HAnNi2 and HAnCu2 at k max ¼ 250 nm (Figures 1 and S4).

Cyclic voltammetry
Additional characterization by CV allowed us to obtain the electrochemical features of our synthesized molecules. CV of HA showed a first irreversible anodic peak (Ea 1 ), assigned to a one-electron oxidation of the hydroxamato group leading to formation of a transient nitroxide radical. It can be seen that the longer the alkyl chain, the lower this oxidation potential Ea 1 . This could be due to the donor effect of the alkyl chain, although this effect can be considered as weak when the chain is longer than five methylene groups. Then, different oxidation pathways have been suggested with either the formation of an acyl radical intermediate with the loss of one electron or the formation of an acyl nitroso, both hydrolyzing into their respective carboxylic acids with the release of a molecule of NHO (Scheme 3A) [56,57]. This latter is unstable and dimerizes leading to formation of N 2 O and H 2 O (Scheme 3B) [56][57][58]. According to previous publications [58][59][60], this oxidation pathway is even more complex and can lead to formation of the respective N,O-diacylhydroxylamine by recombination of either the acyl radical with the nitroxide radical or by reaction between the acyl nitroso with the deprotonated form of HA (Scheme 3C). The hydroxamate group may also be reduced, although this reduction is hardly described in the literature. According to the CV of aliphatic HA and by comparing them with data published on aromatic HA, it could be postulated that the mechanism is closer to the reduction of N-hydroxy-2-oxo-2-phenylacetamide (PhCOCONHOH) into 2-hydroxy-2-phenylacetamide (PhCHOHCONH 2 ) and 2,3-dihydroxy-2,3-diphenylsuccinamide ([PhC(OH)CONH 2 ] 2 ) [61], and that may explain the appearance of the reversible redox couple (Ea 2 /Ec 2 ) observed at À0.2 V and þ0.4 V with HA.
Regarding the Cu(II) complexes, the absence of electrochemical processes could be explained by their greater electrochemical stability in KCl aqueous solution (1 M) in 5% MeOH on the carbon screen-printed electrode (cSPE) surface than the other studied complexes (Figures 4 and S11). For the Zn(II) and Ni(II) complexes, the redox process can be assigned to the electrochemical phenomenon on the HA itself meaning that those complexes are more unstable in those conditions. The CV of Fe(II/III) complexes reveal a quasi-reversible redox process with absorption phenomenon involving likely the HA itself and the central metallic atom. The well-defined reversible peaks at Ea 1 (þ0.1 V) and Ec 1 (-0.1 V) are due to the redox process of the Fe center [62]. Compared to the reversible redox peak at þ0.9 V for iron chloride ( Figure S5), those peaks are shifted to a lower value of potential, indicating a facilitated electron transfer in those complexes. Finally, the broad anodic peak at þ0.3 V and the broad cathodic peak at À0.2 V are linked to the electrochemical reaction of the HA.

Stability of compounds
Concerning the stability of HA in water, hydrolysis into carboxylic acid and hydroxylamine is the main degradation pathway described in the literature. This hydrolysis is very slow in a neutral aqueous solution compared to an acidic medium, by several orders of magnitude [63,64]. The non-catalyzed hydrolysis of HA in water is extremely slow and insignificant [64], hence HA are expected to be stable. Ligand-water molecule exchange is the expected reaction that could occur with complexes. The stability study of our complexes using UV-visible spectrophotometry revealed that only HA2Fe3 and HA6Fe3 are clearly unstable in water. After about 20 minutes, strong changes appeared in the UV-visible spectra. According to previously published work, the degradation products could be carboxylic acid and NH 2 OH through hydrolysis, and Fe 3þ by decomplexation. The hydrolysis rate of Fe(III) complexes is greater than that of their corresponding HA [63,64]. Data found in the literature highlighted two structural elements related to the stability of the HA complexes: (i) the type of coordination metal and (ii) the length of the alkyl chain [65,66]. Regarding the metal, the size of the cation influences the stability: the smaller the cation, the higher stability (following the Irvings-Williams order of stability: Fe 2þ < Ni 2þ < Cu 2þ < Zn 2þ ). The chain length is cited as a factor influencing the stability but with no explanation.

Magnetic susceptibilities and electrolytic characteristics of complexes
The susceptibility of the different complexes at room temperature ( Figure 8A-D) show paramagnetic behavior. Although a diamagnetic component in the susceptibility does not change much, it could also be present in the compounds. As seen from Table 3, Cu(II) and Zn(II) complexes show lower susceptibility values compared to Fe(II/III) and Ni(II) complexes, from which HA12FeCl complex demonstrates the largest value of mass susceptibility (34.64 Â 10 6 emu g À1 ) in agreement with the previous report on ferric compounds [67]. The ratios between mass susceptibilities for Fe(II/III), Ni(II), Cu(II) and Zn(II) complexes are comparable with those between molar susceptibilities for related inorganic compounds, i.e. v m ¼ þ13450, þ6145, þ1080, À55 Â 10 À6 cm 3 mol À1 for FeCl 3 , NiCl 2 , CuCl 2 , ZnCl 2 , respectively [35]. It can be seen in Figure 8B,C that a dominant ferromagnetic/superparamagnetic-like contribution is observed for some Cu(II) complexes as well as HA12HNi2 complex at room temperature. Hence a small room temperature coercive field (H c ¼ 63 Oe) and remanence (M r ¼ 8 Â 10 À4 emu g À1 ) for the HA12Cu2 complex ( Figure S19) are observed, probably an effect of magnetic contamination. Interestingly, a paramagnetic dependence is observed for Zn(II) complexes ( Figure 8D) for one would expect prevailing diamagnetism featured in Zn inorganic compounds.
The highest magnetic moment (m eff ) values were obtained with the Fe(II/III) complexes which are between 4.41-7.12 BM (  [68,69]. As for the Zn(II) complexes, the values were between 0.89-1.69 BM. This is in accord with data published, for instance with the triazole complexes of Fe (m eff ¼ 5.06 BM), Ni (m eff ¼ 2.97 BM) and Cu (m eff ¼ 1.79-1.81 BM) [5]. These values indicate that these complexes are paramagnetic. The values of m eff for triazole Zn(II) complexes is close to 0 BM, corresponding to diamagnetic complexes [5,70]. These data allow correlating the magnetic properties of the complexes with the configuration of the d 6 , d 8 , d 9 (paramagnetic: Fe(II/III), Ni(II), Cu(II), respectively) and d 10 orbitals (diamagnetic: Zn(II)). Overall, the magnetic properties of the investigated complexes correlate well with the values of molar magnetic susceptibilities of their inorganic counterparts and effective magnetic moments of similar complexes.
In the literature, molar conductance values between 11-19 X À1 cm 2 mol À1 indicate an absence of ionic form of the complexes in solution. For instance, a conductivity value of 98 X À1 cm 2 mol À1 obtained with a chromium(II) complex with a triazole ligand was considered to be a confirmation of its ionic character (electrolyte) [70]. Results obtained with our Fe(II/III), Ni(II) and Cu(II) complexes versus FeCl 2 , FeCl 3 , NiCl 2 and CuCl 2 thus corroborate their non-ionic nature. As for the Zn(II) complexes, the value of their molar conductance is close to that of ZnCl 2 , suggesting an ionic character (Table 4). In another published work, complexes of Ni(II), Cu(II) and Zn(II) with Schiff bases (sulfanilamides) have shown molar conductance values in dimethylformamide (DMF) between 135-142 X À1 cm 2 mol À1 , indicating the electrolytic nature of these complexes [71].

Antimicrobial effect and cytotoxicity
Vancomycin, rifampicin and cetrimide (reference antibiotics and antiseptic) showed a greater inhibitory effect than HA and Fe(II/III), Ni(II), Cu(II) and Zn(II) complexes.
Although the antimicrobial effects observed are weaker than those of standard antibiotics, a higher inhibition than other compounds described in the literature is observed. The antimicrobial results obtained with compound HA12 (13.44 mg/mL) are more significant than those obtained with N-(2-hydroxyethyl)-5-nitrosalicylaldimine (MIC ¼ 250 mg/mL) on the same strains (SASM and E. coli). Although the HA and complexes of the present work were inactive against K. pneumoniae and P. aeruginosa (MIC > 0.5 mM), previously described Cu(II) complexes with Schiff base ligands showed an inhibitory effect against these two bacterial strains (MIC ¼ 2.4-5.8 mM) [72]. The HA12Ni2 (MIC ¼ 15 mg/mL), HA12Cu2 and HA12Zn2 (8 mg/mL) complexes which were the subject of our study have a more significant antifungal effect than the Fe(II) and Ni(II) complexes of N-(2-hydroxyethyl)-5-nitrosalicylaldimine (MIC ¼ 250 mg/mL) against C. albicans [73].
Although HA displayed antimicrobial activities [4,75,76], the effect of the length of the alkyl chain has not been studied in the literature [77]. In our study, compounds which have a significant antimicrobial effect on the microorganisms tested are those having a logP ! 3 (HA12 and corresponding complexes) ( Figure S6). This points out the hydrophobic character as an important factor for antimicrobial activity thanks to a better diffusion through the cell membrane [78,79]. A correlation between the alkyl chain length and the antibacterial or antifungal effect of the hydrophobic compounds HA10 and HA12 has been observed (Table S3) with a two-fold lower MIC for HA12 on C. albicans compared to HA10. However, HA12 is moderately active on E. coli with MIC at 63 lM (13 lg/mL) and HA17 is inactive against the same strain, highlighting that water solubility is also a significant parameter. Consequently, there is a need for a good balance between these two properties. For instance, it was reported that the Nacetoxy palmitoylhydroxamic acid has no significant effect on E. coli [80], probably due to low water solubility.
Compared with HA6 and HA8, their derived complexes with Fe(II/III) or Zn(II) (HAnFe2, HAnFeCl, HAnFe3, HAnZn2) displayed stronger inhibition against S. aureus and E. coli, in agreement with the report that the antimicrobial activity of HA complexes was higher than the corresponding free HA [8]. Cell wall penetration in Gram negative bacteria is lower than in Gram positive bacteria, due to the fact that the outer membrane of the Gram negative bacteria limits the diffusion of compounds inside the cytoplasm [78,79]. Lipophilic ligands could facilitate the penetration of the metals through the hydrophobic lipid layer, explaining thus the increase in antimicrobial activities for most lipophilic molecules [81]. The increased activity of the complexes compared to their respective HA is often explained by Overton's concept and Tweedy's chelation theory [72]. According to the Overton's concept of cell permeability, the lipid membrane surrounding the cell favors the passage of only fatsoluble materials. During chelation, the polarity of the metal ion is reduced due to the neutralization of the positive charge with the negatively charged donor groups of the ligands. This improves the lipophilic character of the metal atom in the complex compared to the metal alone. Penetration through the lipid layers of the cell membrane is facilitated by this lipophilicity [82].
Compounds HA10 and HA12 displayed antimicrobial effect on C. albicans, showing a fungistatic activity for HA10 and fungicidal activity for HA12. HA12, HA6FeCl, HA6Fe3, HA8FeCl, HA8Fe3, HA10Fe2, HA10FeCl and HA10Fe3, HA6Zn2, HA8Zn2, HA10Zn2, HA12Zn2 and zinc chloride also demonstrated bactericidal activity against S. aureus MSSA, C. glutamicum and E. coli. HA12 had a broad spectrum of antimicrobial activities, being active not only against fungi, Gram positive and negative bacteria but also against drug resistant MRSA and M. smegmatis. On the opposite, its derived complexes, HA12Cu2, HA12Zn2 and HA12Ni2, were mainly active against C. albicans. As previously published, Fe(II/III), Ni(II), Cu(II) and Zn(II) complexes of HA displayed various antibacterial and antifungal activities [8][9][10][11]. For the moment, the mechanisms of action of these compounds are not known but some hypotheses could be proposed, based on the chemical properties of HA and the metals used in this study.
Compounds such as HA could inhibit microorganisms through metal chelation involved in enzymatic reactions but also through their ability to release HNO and NO, as shown in CV analysis [83]. Fosmidomycin, an antibiotic from Streptomyces, derived of N-hydroxyhexanamide, is a good example of a natural HA targeting a metal containing enzyme in bacteria. Indeed, it is able to inhibit DXP reductoisomerase whose activity depends on Mn 2þ , Co 2þ or Mg 2þ ions [84]. In the case of metal complexes, either small amount of metal arising from complex dissociation or the metal complexes themselves could produce deleterious biochemical effects on microorganisms. It remains difficult to determine precisely the mechanism of action of such species in complex biological systems, but it is logical to consider that metal ions could combine with the chemical groups present in proteins, mainly SH, COOand imidazoles, such as those present in ribosomes [84]. Other hypotheses on the mode of action include inhibition of DNA synthesis, cell wall or membrane disruption or inhibition of RNA synthesis based on a disturbance in the energy metabolism of the cell. Finally, disruption of the membrane could affect the electron transport chain and the oxygen consumption of microorganisms. However, some compounds could also act outside the cytoplasm. Pathogens secrete various molecules (small ones or proteins) to achieve some metabolic pathways at the membrane periphery, as exemplified by bacterial transpeptidases that are inhibited by beta-lactam antibiotics.
Some compounds were mostly inactive on a drug resistant S. aureus strain. This could be due to the higher cell wall hydrophobicity of this strain [85]. The broad antimicrobial spectrum activity of HA12 could be due to the fact that this compound could chelate different metals in the various tested microorganisms [77]. The increased antifungal activity of HA12 derived complexes could be due to their ability to trigger cell death when they get attached to the negatively charged microbial cells and can cause irreversible cell membrane damage [86].
HAnZn2 and ZnCl 2 have stronger inhibitory effects than other complexes (HAnFe2, HAnFeCl, HAnFe3, HAnCu2 and HAnNi2) on E. coli. HAnZn2 displayed growth inhibition with a MIC of 78 mM, higher than the MIC of ZnCl 2 (< 78 mM) on E. coli (Table  S3), indicating that complexation is not always necessary for the metal to exert its biological activity. A good example is ZnO nanoparticles that can cause membrane disorganization and, subsequently, destruction of intracellular components of E. coli [87]. The HA12 complexes showed a particular antifungal activity on C. albicans, with stronger activity for HA12Cu2 and HA12Zn2 than HA12Ni2, probably due to the difference in standard reduction potentials. Indeed, the CV assay values and UV stability studies indicated that HAnCu2 are more stable than those of HAnZn2 or HAnNi2 (Figures 4  and S11). Our hypothesis is that Zn(II) or Ni(II) could be reduced in fungi's cytoplasm and lose more easily their ability to bind the targets. In other words, reactivity of the complexes could improve their antifungal activity [12]. The lipid, carbohydrate or protein contents in microorganism membranes, and consequently, their physicochemical properties, could explain the specificity of the HA derived complexes on the different microorganisms. The hydrophobicity of the compounds, along with the nature of the metal in the complexes, is thus only one of the factors that could explain specific antimicrobial properties [77,88].
Selectivity towards human cells was a major concern for a potential use of the complexes in medicine. In this view, SI were determined. Antifungal compounds HA12Cu2, HA12Zn2 and HA12Ni2 showed better selectivities compared to HA12 ( Table 6). The broad activity spectrum of HA12 could be due to its pleiotropic action on microorganism and human cells based on metal complexation. However, Fe(II/III) complexes with a shorter alkyl chain (hydrophilic) were more cytotoxic on SiHa cells than their higher lipophilic homologues. The complexes HA6Fe3 and HA8Fe3 showed the highest cytotoxicities with SI of 0.04 and 0.10, respectively. These hydrophilic complexes could be more easily localized in the cell nucleus or in the mitochondria and therefore induce cell cycle arrest by triggering DNA damage. Previous work on HA reported that their toxicity could result from iron ion withdrawal by chelating, which could greatly reduce DNA synthesis and cell growth [89]. They could also be competitive inhibitors of transferrin [90].
The results on the antimicrobial effect of some compounds provide interesting data, especially the HA12, HA12Ni2, HA12Cu2 and HA12Zn2 showing significant fungicidal activity against C. albicans. Further studies must be undertaken to assess their potential medical interest (as disinfectant for instance) and to determine their mechanism of action.

Conclusion
HA and their Fe(II/III), Ni(II), Cu(II) and Zn(II) complexes were synthesized, characterized and evaluated for their antimicrobial properties against various microorganisms. As expected, a correlation was found between the lipophilicity of the HA and derived complexes (HAnFe2, HAnFeCl, HAnFe3, HAnNi2, HAnCu2 and HAnZn2) (expressed as their logP) and their antimicrobial effects and cytotoxicities [81,91]. A similar correlation was found between their molecular weights and antimicrobial activities. The lipophilic complexes HA12Cu2, HA12Zn2 and HA12Ni2 have interesting antifungal activity, considering their selectivity index. On the opposite, HA12 has a broad spectrum of antimicrobial activities but is also equally toxic on human cell lines. Altogether, our results suggest that HA and in particular, their complexes, could be used for the development of novel antimicrobial and particularly antifungal compounds, as surface coating agents or as disinfectants.