Synergetic Antimicrobial Activity and Mechanism of Clotrimazole-Linked CO-Releasing Molecules.

Several metal-based carbon monoxide-releasing molecules (CORMs) are active CO donors with established antibacterial activity. Among them, CORM conjugates with azole antibiotics of type [Mn(CO)3(2,2'-bipyridyl)(azole)]+ display important synergies against several microbes. We carried out a structure-activity relationship study based upon the lead structure of [Mn(CO)3(Bpy)(Ctz)]+ by producing clotrimazole (Ctz) conjugates with varying metal and ligands. We concluded that the nature of the bidentate ligand strongly influences the bactericidal activity, with the substitution of bipyridyl by small bicyclic ligands leading to highly active clotrimazole conjugates. On the contrary, the metal did not influence the activity. We found that conjugate [Re(CO)3(Bpy)(Ctz)]+ is more than the sum of its parts: while precursor [Re(CO)3(Bpy)Br] has no antibacterial activity and clotrimazole shows only moderate minimal inhibitory concentrations, the potency of [Re(CO)3(Bpy)(Ctz)]+ is one order of magnitude higher than that of clotrimazole, and the spectrum of bacterial target species includes Gram-positive and Gram-negative bacteria. The addition of [Re(CO)3(Bpy)(Ctz)]+ to Staphylococcus aureus causes a general impact on the membrane topology, has inhibitory effects on peptidoglycan biosynthesis, and affects energy functions. The mechanism of action of this kind of CORM conjugates involves a sequence of events initiated by membrane insertion, followed by membrane disorganization, inhibition of peptidoglycan synthesis, CO release, and break down of the membrane potential. These results suggest that conjugation of CORMs to known antibiotics may produce useful structures with synergistic effects that increase the conjugate's activity relative to that of the antibiotic alone.

In the wake of the rapidly expanding studies on carbon monoxide (CO) biology and therapy, which revealed a broad area of applications,[1] some of us decided to investigate the influence of CO on bacteria. For this purpose, we used gaseous CO and several metal-based CO-releasing molecules (CORMs) that had been established as therapeutically active CO donors in in vivo disease models. We found that several of those CORMs had significant antibacterial activity toward both Gram-positive and Gram-negative bacteria, whereas others were weaker or even innocuous.[2] The readily available and most widely used Ru(II)-based CORM-2 and CORM-3 (Figure ) were among the more potent CORMs in this seminal study, and thus, they were selected for further studies by us and others. Much attention was devoted to the understanding of the mode of action (MoA) of this new class of antimicrobials. Despite serious efforts, the MoA still is not completely understood, and a full account of its details has been given.[3,4] There are data clearly supporting the action of CO and data favoring the dominant action of Ru(II) ions, as well as data countering each of these possibilities. This is due to the very labile nature of both CORM-2 and CORM-3 in biological media and their unusual and very complex chemistry.[5,6] In any case, the potential utility of these CORMs, and certainly of other new antimicrobials overcoming bacterial resistance, became evident in three studies showing that CORM-3 offers protection to rodents infected with Pseudomonas aeruginosa;[7] CORM-2 potentiates the effect of antibiotics, for example, by resensitizing a metronidazole-resistant Helicobacter pylori strain;[8] and CORM-2, in combination, increases the efficiency of tobramycin against P. aeruginosa.[9]

Figure 1

Selected CORMs with antimicrobial activity.

In order to better evaluate the role of CO as a bactericidal agent, Schatzschneider and co-workers introduced the structurally stable CORM [Mn(CO)3(Tpa-κ3N)]Br (Figure ) and studied its bactericidal action on Escherichia coli. The compound requires photoactivation to release CO. The proposed overall mechanism is complex and includes the destruction of the bacterial membrane by HO• radicals formed in the reaction of the partially CO-depleted fragment [Mn(CO)2(Tpa-κ3N)]+ with H2O2 generated in the cytoplasm.[10,11] In agreement with the results obtained with CORM-2 and CORM-3, this photoCORM synergized with doxycycline and colistin to better kill a multidrug resistant avian pathogenic E. coli strain.[10,12]

Also moving to more structurally stable metal carbonyl compounds, Metzler-Nolte and co-workers studied a family of complexes of the general formula [Re(CO)3(DpaR)]+, where R represents a series of substituents on the central amine of the tripodal di-(2-picolyl)amine (Dpa) ligand (Figure ). Several of these complexes displayed high activity against Gram-positive bacteria, but activity against Gram-negative bacteria was not detected.[13,14] In an elegant structure–activity relationship (SAR) study, they established that high activity (minimal inhibitory concentration, MIC ≈ 2–4 μg/mL) was only met for certain types of R chains, based upon peptide nucleic acids (PNAs), and that the [Re(CO)3]+ metal fragment was necessary but not a sufficient condition to impart bactericidal activity to the (DpaR) ligand structure. The MoA study indicated that the compound affects the membrane potential and membrane integrity, as well as cell wall integrity, thereby inducing cell membrane blebbing. Similar compounds were shown to have bactericidal activity in the dark and more strongly under light irradiation, killing both Gram-positive and Gram-negative bacteria.[15]

In a slightly different approach, Schatzschneider and co-workers reported that the conjugated CORMs [Mn(CO)3(Bpy)(azole)][PF6] (where the azole was miconazole, ketoconazole, or clotrimazole (Ctz)) had much higher activity against a panel of Gram-positive bacteria and parasites than the azole alone. In particular, the [Mn(CO)3(Bpy)(Ctz)][PF6] derivative (Figure ) showed the highest activity (up to 30-fold higher than that of Ctz alone) against the eukaryotic parasites Leishmania major and Trypanosoma brucei.(16) Interestingly, the lowest MIC values of 0.6 μM are close to those reported for the active [Re(CO)3(DpaR)]+ molecules, and no mention of any CO participation in the mechanism is made in both studies. More recently, phosphopyricin, a phosphine derivative of the W(CO)5 fragment, W(CO)5{PPh2(C3-pyrrole)}, has also shown activity in the same MIC range, with the property of losing activity upon exposure to light.[17]

Metzler-Nolte and co-workers also reported potent bactericides against Gram-positive bacteria based on the [Re(CO)3(BpyRR)X]0,1+ scaffold where BpyRR represents substituted bipyridyl and X represents N-heterocyclic carbene (NHC) ligands derived from benzimidazole.[18] Along these lines, Zobi and collaborators have demonstrated the enormous antimicrobial potential of [Re(CO)3(L–L)X]0,1+ complexes in two outstanding screening studies published while this work was ongoing.[19,20] A number of highly potent derivatives of the [Re(CO)3(BpyRR)X]0,1+ framework (X = pyridine-based ligands) were found to be active against Staphylococcus aureus, methicillin-resistant Staphylococcus aureus (MRSA), and Candida albicans in the sub-micromolar or low micromolar range. Importantly, these molecules (Figure ) were devoid of important toxicity, presented high therapeutic indexes, and even outperformed linezolid and vancomycin.[19,20]

It is now clear that the [M(CO)3] fragment, which is present in all of these complexes as well as in W(CO)5{PPh2(C3-pyrrole), enables the assembly of a range of metal carbonyl complexes with potent bactericidal activity. However, the true mechanistic MoA of these complexes has not been studied in detail, and it may differ among the different types of scaffolds.

The [Mn(CO)3(Bpy)(azole)]+ conjugates are very interesting species since different types displayed important synergies with other antibiotics, to the point of reverting acquired antibiotic resistance features in some bacteria. In this work, we sought to better understand the MoA of these CORM conjugates as well as modulate and improve their activity as antimicrobials. For this purpose, a SAR study based upon the [Mn(CO)3(Bpy)(Ctz)]+ lead structure was carried out, producing several conjugates with varying metals and ligands, which were tested for their antimicrobial activity and MoA.

Results

Design, Synthesis, and Characterization of CORM–Clotrimazole Conjugates

Considering that the [M(CO)3] fragment is a potential determinant of the antibacterial activity,[16] we used the structure of [Mn(CO)3(Bpy)(Ctz)][PF6] (Bpy = 2,2′-bipyrydine and Ctz = clotrimazole) as the lead to this SAR study intended to gain mechanistic insights into the MoA of these type of metal carbonyl complexes. For this purpose, several analogues of [Mn(CO)3(Bpy)(Ctz]PF6] were synthesized with the following modifications: (i) replacement of Mn with Re; (ii) introduction of a chemical substitution on the Bpy ligand; (iii) replacement of bipyridine ligands with other diimine ligands; and (iv) introduction of nondiimine bidentate ligands with different electronic and steric requirements.

We started from Mn(CO)5Br (aka ALF0212), and the conjugates were prepared, along with a series of useful intermediates using the stepwise approach depicted in Scheme . In this way, several isostructural and isoelectronic conjugate complexes were prepared (Figure ). Since all these compounds have a common set of four ligands (3xCO + Ctz) as well as the same counterion, triflate, we will identify them by the symbols of the metal and the bidentate ligand, thus facilitating their identification when comparing properties, for example, [Mn(CO)3(Bpy)(Ctz)]OTf = MnBpyCtz and [Mn(CO)3(Bpy)Br ] = MnBpyBr.

Scheme 1

Synthesis of the Clotrimazole–CORM Conjugates

M = Mn and Re; X = Br and I; OTf = O3SCF3 (triflate); L–L = Bpy, Biq, PyBzim, Bpydinon, PyNHC, and bisNHC; and Ctz = clotrimazole. OTf (O3SCF3, triflate).

Figure 2

Conjugates prepared by combining the fac-{M(CO)3}d6 fragment with bidentate ligands and clotrimazole.

These clotrimazole conjugates are all 18-electron monocationic species. As solids, they are stable against laboratory manipulation in air and under ordinary light for short periods. They are routinely kept in the dark under nitrogen and, when necessary, are dissolved in deoxygenated solvents except for the biologic tests. The synthetic procedures are described in the Materials and Methods section along with the spectroscopic data. All compounds were characterized using elemental analysis, Fourier transform infrared (FTIR) spectroscopy, 1H NMR, and, when necessary, by 13C NMR and high-resolution mass spectrometry (HRMS). FTIR data on νCO stretching vibrations (Supporting Information Table S1) confirm the presence of facial fac-{M(CO)3}d6 fragments in all the prepared complexes.

Compound [Mn(CO)3(Bpy)(Ctz)][OTf] (MnBpyCtz) (Figure ) was prepared as previously described.[16] The rhenium analogue ReBpyCtz was synthesized from the known fac-[Re(CO)3(Bpy)Br] (ReBpyBr). The νsymCO stretching vibration (Supporting Information Table S1) is slightly lower for ReBpyCtz (2029 cm–1) than for MnBpyCtz (2038 cm–1), which is the complex with the highest νsymCO value of all the complexes shown in Figure ,indicating the more efficient π-acceptance of this ligand in this set of compounds. The 2,2′-biquinoline (Biq) complex, MnBiqCtz, derived from well-characterized fac-[Mn(CO)3(Biq)Br],[21] extends the ligand bulk on the equatorial plane as well as the π cloud of the bidentate ligand. The latter extension is reflected in the red color of MnBiqCtz, which contrasts with the yellow or yellowish color of all the other complexes. The values of the νsymCO stretching vibrations are intermediate between those of ReBpyCtz and MnBpyCtz. Conjugate MnPyBzimCtz extends the bulk of the N–N bidentate ligand also on the equatorial plane of the complex but in an asymmetric manner. Its νsymCO stretching vibration is clearly lower (ca. 14 cm–1) than that of MnBpyCtz, in agreement with a lower π-acceptance of the pyridine-benzimidazole ligand relative to the Bpy ligand. Complex MnBpydinonCtz contains a Bpy ligand modified by the presence of two C9-alkyl chains in positions 4 and 4′ of the heterocyclic rings. Such chains impart a marked stereochemical bulk on the distal side of the [Mn(CO)3] face. This substitution produces a ca. 10 cm–1 decrease in the νsymCO stretching vibration when compared to MnBpyCtz, which agrees with a reinforced π-back donation due to the electron donating effect of the alkyl substituents that increase the electron density at the metal.

A new subset of complexes with other small, bidentate ligands was also assembled. In this smaller set, the ligands have higher σ-donating and lower π-accepting ability. In particular, we used imidazolium-derived NHC ligands, which are very strong carbon σ-donors that have been extensively used in catalysis. The bidentate ligand in MnPyNHCCtz features one N-bound pyridine ring and one C-bound NHC. From the structural point of view, this ligand is flat, making the fragment [Mn(CO)3(PyNHCMe)]+ very compact, similar to the Bpy analogue. The νsymCO for MnPyNHCCtz is similar to that in MnBiqCtz and MnBpydinonCtz, indicating a fairly efficient π-acceptance. The bisNHCMe ligand featured in MnbisNHCCtz is an even more powerful electron donor to the [Mn(CO)3(Ctz)]+ fragment, and this is reflected in the lowest νsymCO stretching vibration (2013 cm–1) of all shown conjugates. However, in contrast to the PyNHCMe ligand, the BisNHCMe ligand is not flat with both methyl substituents on the N2C3 rings pointing to the same side of the equatorial plane.[22] Therefore, complex MnbisNHCCtz is the first in this series without a flat equatorial ligand. Nevertheless, the stability imparted by the BisNHCMe ligand in MnbisNHCCtz is very high, judging by its air stability and the robust catalytic activity of its precursor [Mn(CO)3(BisNHCMe)Br] in the catalytic reduction of CO2 to CO.[22]

Taken together, this small library of complexes contains species that have the same type of geometrical structure (octahedral around the metal with the same relative arrangement of the three CO and Ctz ligands), the same positive charge (1+), the same counterion (triflate), a continuum of electron density at the metal (thus reactivity), as judged by the range of νCO values, and a few small structural variations concerning the bulk of the equatorial plane containing the bidentate ligand and two CO ligands. Below, we show how these small differences influence the antibacterial activity of the CORM complexes.

Bactericidal Properties of CORM–Clotrimazole Conjugates

We first used three model bacterial pathogens to characterize the antimicrobial activity of the synthesized CORMs, namely, two Gram-negative species E. coli and Salmonella enterica, and Gram-positive MRSA. Several compounds among our set of CORM–Ctz conjugates demonstrated single-digit micrograms per milliliter MIC values against S. aureus, and, notably, several of them outperformed free clotrimazole (Table ). ReBpyCtz was 4 times more active than its Mn analogue MnBpyCtz, while both of their clotrimazole-depleted precursors, namely, ReBpyBr and MnBpyBr, respectively, did not show antibacterial activity against these bacteria up to 32 μg/mL (Table ). Cultures of E. coli and S. enterica still showed residual turbidity after overnight incubation with CORMs (Table ), and most compounds were not active enough to inhibit growth sufficiently for an MIC readout. Nonetheless, several compounds reduced the growth rate also for the Gram-negative species (Supporting Information Table S2).

Table 1

MICs of Conjugate CORMs (A) and of ReBpyCtz and Its Synthetic Precursors (B) against the Indicated Bacteriaa

A	S. aureus	E. coli K12	S. enterica
MnBpyCtz	2	>32	>32
ReBpyCtz	0.5	>32	>32
MnBiqCtz	8	>32	>32
MnPyBzimCtz	10	>32	>32
MnBpydinonCtz	>32	>32	>32
MnPyNHCCtz	0.5	>4	Ndb
MnbisNHCCtz	0.5	>4	Ndb
MnBpyBr	>32	>32	>32
ReBpyBr	>32	>32	>32
MnPyNHCI	>4	>4	Ndb
MnbisNHCBr	>4	>4	Ndb
Ctz	4	>32	>32

MICs (μg/mL) were determined as described in the Materials and Methods section and for at least three biological samples.

Nd—not determined.

To further discriminate the conjugate CORMs’ activity, we evaluated the growth of S. aureus, E. coli, and S. enterica when treated with CORMs (Figure and Supporting Information Table S2).

Figure 3

S. aureus cells are sensitive to conjugated CORMs. Cells of S. aureus MRSA, E. coli MG1655, and S. enterica SL1344 were treated with the indicated CORMs (35 μM for Gram-negative bacteria and 10 μM for S. aureus). The percentage of the growth rate was determined in relation to untreated cells collected at 2 h and 4 h. Data represent the average of three independent biological samples, with error bars representing the standard deviation. *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001.

Several CORM–Ctz conjugates reduced the growth of Gram-negative bacteria, with ReBpyCtz showing the highest activity against E. coli (a reduction of ∼63% of growth after 4 h) and S. enterica (a reduction of ∼37% growth after 4 h). For E. coli and S. enterica, a reproducible impact on growth was observed only for CORMs with the smaller ancillary rings, namely, MnBpyCtz, ReBpyCtz, MnPyNHCCtz, and MnbisNHCCtz. While cells exposed to MnBpyCtz after 4 h recovered and exhibited a growth behavior similar to that of nonexposed cells (93% growth relative to untreated cells), no growth recovery occurred for E. coli or S. enterica exposed to ReBpyCtz.

All CORM–Ctz conjugates were toxic to S. aureus, except MnBpydinonCtz, causing a significant growth decrease in the following order: MnPyBzimCtz < MnbisNHCCtz ≈ MnPyNHCCtz < MnBiqCtz ≈ MnBpyCtz ≈ ReBpyCtz (Supporting Information Table S2). Prolonged exposure of S. aureus (∼4 h) to the latter five conjugates reduced growth to values of only 2–4% of those of untreated cells (Supporting Information Table S2).

Altogether, these results revealed a set of new CORMs with antimicrobial properties that go beyond the effect of free clotrimazole. All complexes have a common structural fragment [M(CO)3(Ctz)]+, which is bound to several different bidentate N–N, N–C, and C–C ligands. Given the similarity of the electronic properties suggested by the relatively close νCO values and the chemical stability revealed in the blood tests (see below), it seems that the fine structural details of the ancillary ligands are key to controlling the activity of the CORMs against S. aureus. The Re and Mn isostructural analogues (MnBpyCtz and ReBpyCtz, respectively) present a flat and compact bipyridyl ligand attached to the [M(CO)3(Ctz)]+ fragment. When compared with cells not exposed to bacteria, cells had their growth decreased to ca. 8 and 2% at 2 and 4 h after the challenge, respectively. Increasing the bulk of the bipyridyl ligand at the distal side to the metal, that is, the 4 and 4′ positions of the ring, completely prevents activity, making MnBpydinonCtz innocuous. The presence of CH3 or CO2CH3 groups at the 4 and 4′ positions of bipyridyl also decreases the antimicrobial effect, as previously reported.[15] This is due to a steric effect since the electronic effects of nonyl and carboxyl substituents are opposed, yet the antimicrobial activity follows in the same direction.

The superior activity of the small, compact, and unencumbered Bpy derivatives led us to explore other [M(CO)3(L–L)(Ctz)]+ complexes with bidentate L–L ligands structurally similar to 2,2′-bipyridyl. Biq was an obvious choice, and in fact, complex MnBiqCtz significantly limited S. aureus growth, much like the effect of 2,2′-bipyridyl derivative MnBpyCtz. The larger occupation of the equatorial plane of conjugate MnBiqCtz by the extended π-ligand was no significant impediment to the CORM activity. On the contrary, a smaller yet asymmetrical occupation of the equatorial plane had an important effect, lowering the antibacterial activity of MnPyBzimCtz in comparison to that of MnBiqCtz and MnBpyCtz. The growth of S. aureus in the presence of MnPyBzimCtz did not increase significantly with time, remaining similar at 2 and 4 h postexposure. It also emerged that the bipyridine and biquinoline complexes are more active at short incubation times than all other compounds tested. Complexes MnPyNHCCtz and MnbisNHCCtz that feature NHC ligands are also very active, but their action is markedly slower than that of the Bpy and Biq derivatives. The fact that the bis-NHC ligand in MnbisNHCCtz is not flat, unlike the PyNHC ligand in MnPyNHCCtz, does not make any difference in the activity profile of both complexes.

The antibacterial potency of the clotrimazole-free [M(CO)3(L–L)Br] precursor complexes was also checked against the same bacteria. As shown in Supporting Information Table S2, the halide complexes [M(CO)3(L–L)Br] (M = Mn and Re), MnBpyBr and ReBpyBr, that are the precursors to MnBpyCtz and ReBpyCtz, respectively, exhibited MICs > 32 μg/mL for all bacteria and percentages of growth higher than 90%. This lack of bactericidal activity for the halide complexes was also reported by Sovari et al.(19) A similar behavior was observed for complexes [Mn(CO)3(L–L)X] (X = Br and I), MnPyNHCI and MnbisNHCBr, which are the precursors of MnPyNHCCtz and MnbisNHCCtz, respectively. The ligands of the most potent CORM–Ctz conjugates, namely, bipyridyl, and the NHC cationic precursors of the ligands in MnPyNHCCtz [(PyNHCMe)·HI] and MnbisNHCCtz [(bisNHCMe)·2(HBr)] did not affect bacterial growth. Bipyridyl, though, was moderately active against E. coli and S. enterica at 250 μM, inhibiting their growth to ca. 50–60% after 4 h of exposure, but there was no effect on S. aureus.

In summary, we conclude that antimicrobial activity of the conjugates is dependent on the presence of the clotrimazole ligand, which, in its free form, is nontoxic to E. coli and S. enterica and moderately active against S. aureus. The fact that compounds MnBpyCtz, ReBpyCtz, MnBiqCtz, MnPyNHCCtz, and MnbisNHCCtz are significantly more potent against S. aureus than clotrimazole alone demonstrates that some kinds of [M(CO)3(L–L)]+ fragments are able to substantially enhance Ctz’s antibacterial activity.

CORM–Clotrimazole Conjugates Are Not Cytotoxic at the MIC Concentration

When conjugating clotrimazole to a rhenium-based CORM, toxicity dosage to human hepatic (HepG2) and porcine kidney (LLC-PK1) cell lines decreases. Half-maximum inhibitory concentration (IC50) values for ReBpyBr and clotrimazole are 10–30 times higher than the IC50 values of ReBpyCtz (Table ).

Table 2

Half-Maximum Inhibitory Concentration (IC50) Expressed in Micromolara

	HepG2	LLC-PK1
	IC50 (μM)	IC50 (μM)
ReBpyCtz	1.4 ± 0.4	2.0 ± 0.8
ReBpyBr	31.1 ± 11.0	16.4 ± 8.3
Ctz	29.2 ± 21.9	11.6 ± 3.2

IC50 values of ReBpyCtz, ReBpyBr, and clotrimazole were determined for human hepatic (HepG2) and porcine kidney (LLC-PK1) cell lines.

CORM–Clotrimazole Conjugates Are Stable Complexes in Blood and Release CO into Bacteria

In vivo, antibiotic molecules must reach the target bacterial cells intact, resisting chemical attack and decomposition caused by blood components, namely, serum albumin and hemoglobin. Serum albumin can radically modify and hijack the coordination sphere of CORMs bearing labile ligands, as in the case of CORM-3 and other fac-{Ru(CO)3}d6-based CORMs.[6,23] Hemoglobin can rapidly scavenge labile CO ligands from many CORMs, forming carboxyhemoglobin (COHb) and accelerating their irreversible decomposition, as in the case of [Mo(CO)3(histidinate]Na (ALF186).[24] If CO is involved in the killing mechanism, as is expected from a CORM, such scavenging inhibits its activity.[2,25] For all these reasons, testing the stability of CORM conjugates is an essential step in their characterization and was performed for the more active complexes, namely, MnBpyCtz, ReBpyCtz, MnBiqCtz, MnPyBzimCtz, MnPyNHCCtz, and MnbisNHCCtz.

For this purpose, solutions of CORMs were added to sheep whole blood in Alsever’s solution, and the blood parameters were read in a blood oximeter at 15 min intervals for 1 h, as previously described.[24] The value of COHb for all complexes remained at the level of the control (Figure A). Also, the value of %Met-Hb remained unchanged throughout the experiment (Figure B). The COHb values did not change, showing that the CORM conjugates do not decompose in blood. Simultaneously, the %Met-Hb value remained unaltered (Figure B). These effects were confirmed in the control experiment carried out with the weak bactericidal complex [Mn(CO)5Br] (ALF021) that generates CO to increase %CO-Hb already at the time of mixing and until ca. 30 min postincubation, while the %Met-Hb decreases relative to the control (Figure C,D). The instability of [Mn(CO)5Br] in blood is readily understood considering the well-known lability of two CO ligands of [Mn(CO)5X] complexes to form kinetically stable fac-[Mn(CO)3]+ derivatives similar to those found in our CORM conjugates. The reducing effect that decreases %Met-Hb is unavoidable for CORMs bearing metals in low oxidation states, as in the case of Mn(I).[26]

Figure 4

Release of CO from conjugated CORMs into blood and bacteria. (A) Percentage of CO–hemoglobin formed in blood samples incubated, for 1 h, with MnBpyCtz, ReBpyCtz, MnbisNHCCtz, MnPyNHCCtz, MnBiqCtz, MnPyBzimCtz, and DMSO. (B) Percentage of Met–hemoglobin formed in blood samples incubated for 1 h with MnBpyCtz, ReBpyCtz, MnbisNHCCtz, MnPyNHCCtz, MnBiqCtz, MnPyBzimCtz, and DMSO. (C) Percentage of CO–hemoglobin bound in blood samples incubated, for 30 min, with ALF21. (D) Percentage of Met–hemoglobin bound in blood samples incubated, for 30 min, with ALF21. (E) Fluorescence microscopy images of S. aureus MRSA CORM–Ctz conjugates (10 μM, 15 min) and incubated with the fluorescent probe COP-1. Representative images of fluorescent cells treated with MnBpyCtz (a), ReBpyCtz (b), MnbisNHCCtz (c), MnPyNHCCtz (d), MnBiqCtz (e), and MnPyBzimCtz (f) (upper row) and their corresponding bright field images (g, h, i, j, k, and l, respectively) (lower images). The population values correspondent to fluorescent green cells represented on the images are 80% (a), 74% (b), 84% (c), 44% (c), 86% (d) and 85% (e), respectively. At least 300 cells, from at least three independent experiments, were analyzed for each condition.

To analyze CO release into bacterial cells, S. aureus was exposed to active CORM conjugates and COP-1, a probe that forms with CO a fluorescent carbonylation product.[27]Figure E shows that these complexes release CO into S. aureus, while cells treated with the clotrimazole-free precursors of MnBpyCtz and ReBpyCtz, namely, MnBpyBr and ReBpyBr, respectively, did not present significant COP-1 fluorescence (Supporting Information Figure S1).

Altogether, we concluded that all [M(CO)3(L–L)(Ctz))]+ conjugates are stable in blood under normoxic conditions, meaning that their CO load is not scavenged by the hemoglobin in circulation. Furthermore, these conjugated CORMs are capable of releasing CO inside bacterial cells.

MoA Study

Among the set of Ctz conjugates analyzed in this study, ReBpyCtz was selected as the prototype for further mechanism of action (MoA) studies due to its high antibacterial activity. In all MoA investigations, ReBpyCtz was evaluated alongside the two molecules that represent its two components, that is, Ctz and ReBpyBr, to unravel which precursor conveyed which characteristics to the resulting hybrid.

Spectrum of Bacterial Target Species

In order to evaluate their target spectrum, we first tested the three molecules against a broader range of bacterial species, including well-characterized model species and a panel of ESKAPE pathogens.[28] MIC values are presented in Table . ReBpyCtz inhibited the Gram-positive representatives of the panel (Bacillus subtilis, S. aureus, and Enterococcus faecium) with MICs of 0.25–2 μg/mL (0.27–2.2 μM) and the Gram-negative Enterobacteriaceae and nonfermenters with MICs of 8–32 μg/mL (8.7–34.7 μM). While Ctz on its own possessed moderate antibacterial activity (MIC 2–8 μg/mL; 5.8–23.2 μM) against the Gram-positive species only, the conjugated compound ReBpyCtz proved significantly more potent and also showed moderate activity against Gram-negative strains. Organometallic precursor ReBpyBr did not yield an MIC up to the highest concentration tested (32 μg/mL; 64 μM). In all following MoA studies, the concentration range was adjusted to their respective MIC levels to correlate the observed biological activities directly to cell death.

Bioreporter Profiling

To obtain an unbiased first impression of the kind of stress bacterial cells feel when exposed to ReBpyCtz, Ctz, and ReBpyBr, we used a bioreporter panel based on the sensitive reference strain B. subtilis 168, designed to specifically sense and signal interferences with most metabolic pathways commonly affected by antibiotics.[29,30]

The readout was based on the induction of the firefly luciferase cloned behind the promoters of genes yorB, helD, bmrC, ypuA, or liaI, which were previously shown to specifically signal the following kinds of damage: P (interference with the DNA synthesis or structure); P (interference with RNA synthesis); P (stalling of ribosomal translation); P (diverse kinds of stress affecting the cell wall or membrane); and P (cell envelope stress, particularly interference with cycling the membrane standing lipid carrier undecaprenyl-phosphate). While Ctz and ReBpyBr gave no signal in this assay, ReBpyCtz strongly induced P at a concentration of 0.5 μg/mL (2× MIC) (Figure A and Supporting Information Figure S2), suggesting an attack on the cytoplasmic membrane and, furthermore, potential interference with peptidoglycan synthesis.

Figure 5

Effect of ReBpyCtz on the cell envelope of growing B. subtilis and S. aureus cells. (A) Firefly luciferase bioreporter profiling demonstrates liaI promotor induction by ReBpyCtz. The liaI promotor responds to cell envelope damage and particularly to agents interfering with cycling of the membrane-standing undecaprenyl-P precursor. Luminescence values recorded at a single predetermined time point after ReBpyCtz addition (see the Material and Methods section) were normalized to the untreated control (100%, black, horizontal dashed line) and plotted against the concentration of ReBpyCtz. At the MIC, the liaI bioreporter is induced, while the signals for the noninduced bioreporters fall below the background level, indicating impaired biosynthetic capacity. The corresponding experiments for Ctz, ReBpyBr, and reference antibiotics are shown in Supporting Information Figure S2. Error bars represent the SD of two biological replicates (i.e., cultures grown and treated independently on different days) with two technical replicates per test day. (B) ReBpyCtz and Ctz disturb the topology of the cytoplasmic membrane. Membrane staining by FM 5-95 of S. aureus NCTC 8325 treated for 30 min with Ctz (8 μg/mL, 2× MIC), ReBpyBr (64 μg/mL), and ReBpyCtz (0.5 μg/mL, 2× MIC) compared to the negative control treated with DMSO. Scale bar, 1 μm. One experiment representative of three biological replicates is shown. (C) Time-resolved effect of Ctz and ReBpyCtz on the membrane potential of S. aureus NCTC8325 measured by DiOC2 (3) staining in relation to the respective MICs. Protonophore CCCP (5 μM, 0.4× MIC) was used as a positive control and DMSO as a negative control. The arrow indicates the time point of compound addition. Four biological replicates and SD are presented. (D) Membrane integrity is not severely affected. Exposure of S. aureus NCTC8325 to either Ctz (16 μg/mL, 4× MIC), ReBpyCtz (1 μg/mL, 4× MIC), or ReBpyBr (64 μg/mL) for 120 min, followed by staining with membrane-permeant Syto9 (green) and membrane impermeant PI (red). Nisin (100 μg/mL) served as a positive control and DMSO (1%) as a negative control. Merged overlay of three photographs of the same cells acquired in the brightfield, green, or red fluorescence channel. Scale bar, 2 μm. (E) ReBpyBr leads to membrane blebbing in B. subtilis 168, indicative of peptidoglycan weakening. Bright-field visualization of the formation of blebs after 30 min treatment with ReBpyBr (64 μg/mL) or vancomycin (Van, positive control, 2 μg/mL). No blebs were induced by Ctz (8 μg/mL, 4× MIC), ReBpyCtz (1 μg/mL, 4× MIC), or DMSO (1%, negative control). Scale bar, 2 μm. (F) UDP-MurNAc-pentapeptide accumulation in S. aureus ATCC 29213 after 30 min of treatment with DMSO (negative control), vancomycin (2 μg/mL, 8× MIC, positive control), Ctz (40 μg/mL, 10× MIC), ReBpyBr (64 μg/mL), or ReBpyCtz (0.5, 2.5, and 5 μg/mL, corresponding to 2×, 10×, and 20× MIC, respectively) shown as AUC values of EICs. An exemplary EIC (shown for the ReBpyCtz at 10× MIC sample) is presented as an inset, depicting the mass of UDP-MurNAc-pentapeptide (m/z = 1148.354) measured in the negative ionization mode. Further EICs and corresponding total ion chromatograms are presented in Supporting Information Figure S5. (G) Monitoring nascent peptidoglycan in growing B. subtilis cells. HADA labeling of B. subtilis 168 after 5 min exposure to Ctz (4 μg/mL, 2× MIC), ReBpyCtz (0.5 μg/mL, 2× MIC or 1 μg/mL, 4× MIC), or CCCP (5 μM) compared to DMSO (0.5%), the negative control. Fluorescence channel (bottom) and overlay with the phase contrast (top). Microscopic settings were selected to yield a good signal in the DMSO control and then kept constant for all micrographs. Scale bar, 5 μm. (H) Quantification of (G). The violin plots represent the relative septal HADA intensities quantified from ≥100 cells per experiment from three independent biological replicates; ****P < 0.0001.

Effects on the Topology, Functionality, and Integrity of the Cytoplasmic Membrane

Next, we monitored the distribution of the membrane dye FM 5-95 after treatment with Ctz, ReBpyBr, and ReBpyCtz. FM 5-95 is a lipophilic dye, which stained the entire cytoplasmic membrane of a dimethyl sulfoxide (DMSO)-treated control aliquot of S. aureus NCTC8325 with homogeneous red fluorescence. The addition of ReBpyCtz and Ctz close to their respective MICs triggered a spotty phenotype, demonstrating dye rearrangement and aggregation and suggesting a general disturbance of the membrane topology (Figure B).

Exploring the membrane defect further, we investigated the effect of ReBpyCtz on the membrane potential of S. aureus NCTC8325 using fluorescence spectroscopy with dye 3,3′-diethyloxacarbocyanine iodide [DiOC2(3)]. When entering the cell, the green fluorescence of DiOC2(3) shifts toward a red emission, due to self-association of the dye molecules. The degree of self-aggregation, and thus the extent of the green-to-red shift, depends on the membrane potential. ReBpyCtz reduced the membrane potential of S. aureus significantly and reproducibly, although comparably high compound concentrations were required (Figure C and Supporting Information Figure S3). A slight effect emerged at 4× MIC and increased in a concentration-dependent manner, but only at 64× MIC did the impact of ReBpyCtz match the strength of the effect that protonophore carbonyl cyanide m-chlorophenylhydrazone (CCCP) already yielded at 0.4× MIC (Supporting Information Figure S3). Consequently, it is highly unlikely that membrane depolarization is the cause of ReBpyCtz-mediated bacterial death. Ctz did also reduce the membrane potential, while ReBpyBr showed no effect.

We also investigated the integrity of the cytoplasmic membrane. S. aureus NCTC8325 was simultaneously exposed to two fluorescent dyes, cell-permeant SYTO9 and membrane non-permeant, positively charged propidium iodide (PI). The latter can cross the bacterial cytoplasmic membrane only if pores or other gross membrane inhomogeneities emerge, and when this occurs, PI inserts into the DNA, thereby emitting red fluorescence. None of the three agents (at 4× MIC for Ctz and ReBpyCtz and at 64 μg/mL for ReBpyBr) allowed for the passage of PI through the membrane, excluding extensive impairment of the membrane barrier function in all three cases (Figure D and Supporting Information Figure S4).

Impact on Peptidoglycan Biosynthesis

As the induction of the liaI bioreporter strain had signaled potential impairment of peptidoglycan synthesis, we also investigated the integrity of the peptidoglycan sacculus. Antibiotics weakening the peptidoglycan sacculus, such as the transglycosylation inhibitor vancomycin, are known to cause membrane extrusions through holes in the peptidoglycan meshwork, when cells are subjected to a fixation procedure with acetic acid/methanol.[31] Unexpectedly, we observed the same phenotype clearly and repeatedly for ReBpyBr at 64 μg/mL, but neither for ReBpyCtz nor for Ctz (Figure E). This result indicates that although the ReBpyBr precursor was not potent enough to yield an MIC, it did possess a bioactivity related to peptidoglycan metabolism.

Inhibitors of membrane-associated stages of peptidoglycan synthesis commonly induce the accumulation of the ultimate soluble peptidoglycan precursor UDP-N-acetylmuramic acid-pentapeptide (UDP-MurNAc-PP) in the bacterial cytoplasm. Therefore, we next analyzed the cytoplasmic level of this metabolite in treated cells. Indeed, ReBpyCtz led to a strong induction of UDP-MurNAc-PP, and the accumulation was already significant at 2× MIC (Figure F), suggesting that inhibition of peptidoglycan synthesis contributes to the antibacterial effect of ReBpyCtz. The strong accumulation of UDP-MurNAc-PP was a new characteristic of the ReBpyCtz conjugate, which was not shown by either ReBpyBr or Ctz. The fact that UDP-MurNAc-PP could accumulate supports the experiments described above. Precursor biosynthesis requires energy, in line with the observed lack of major membrane depolarization at low multiples of the MIC. It also precludes the presence of membrane pores through which the peptidoglycan precursor UDP-MurNAc-PP, ions, and further cytoplasmic content would leak out.

A disturbance of the peptidoglycan synthesis process was also suggested by a pulse-labeling experiment with fluorescent 7-hydroxycoumarincarbonylamino-d-alanine (HADA). HADA allows us to monitor sites of nascent peptidoglycan synthesis and murein remodeling activities in living bacterial cells by its incorporation into the fifth position of the stem peptide by D,D-transpeptidases in vegetative B. subtilis cells.[32] In untreated B. subtilis cells, the label was most clearly visible at the septa, in accordance with the high peptidoglycan synthetic activity known for the septal region and the corresponding abundance of D-Ala–D-Ala moieties.[33] Exposure of B. subtilis cells to ReBpyCtz and Ctz clearly affected HADA incorporation. Comparing the effects of both agents side by side at corresponding multiples of the MIC, it emerged that Ctz inhibited HADA incorporation more rapidly and strongly than ReBpyCtz (Figure G,H). The immediate phenotype induced by Ctz was reminiscent of the one induced by the uncoupler CCCP. ReBpyBr did not inhibit HADA incorporation even after longer CORM exposure (Supporting Information Figure S6). This observation implied a divergent molecular impact of ReBpyCtz and Ctz on cell physiology and the peptidoglycan synthesis process, a hypothesis that we investigated further in in vitro peptidoglycan synthesis assays (Figure ).

Figure 6

Inhibition of in vitro peptidoglycan synthesis by ReBpyCtz and interaction with peptidoglycan precursors. (A) Schematic of lipid II synthesis in S. aureus NCTC8325. UDP-MurNAc-PP and lipid carrier C55P are used as substrates for lipid I synthesis by MraY. MurG catalyzes the addition of GlcNAc from UDP-GlcNAc, yielding lipid II. (B) Influence of ReBpyCtz, ReBpyBr, and Ctz on lipid II synthesis using MraY- and MurG-containing membrane preparations of Micrococcus luteus. At a 10-fold molar excess over C55P, ReBpyCtz and Ctz slightly reduced the amount of lipid II formed, while ReBpyBr did not. NC, negative control, reaction mixture without membranes; DMSO (1%). Error bars represent the SD of two replicates. (C) Influence of ReBpyCtz on lipid II synthesis from purified lipid I (2 nmol) and UDP-GlcNAc by purified MurG. Test compounds were applied in DMSO (1%) at a 10-fold molar excess over lipid I. TLC curves showing the extracted lipids at the end of the reaction. Vancomycin was used as an inhibition control; its binding to lipid I retains the complex in the aqueous phase and prevents it from being extracted. ReBpyCtz sightly inhibited the conversion of lipid I to lipid II. (D) Quantification of the bands visible on the TLC plate in (C). ReBpyCtz inhibits the MurG reaction. Error bars represent the SD of two replicates. (E) LiaI lux bioreporter strain based on the Photorhabdus luminescens luciferase system, yielding a continuous fluorescence signal. The liaI promotor in B. subtilis 168 is induced upon treatment with ReBpyCtz and Ctz around their respective MIC levels, albeit to different extents. No induction was observed for ReBpyBr. Vancomycin (2 μg/mL, 2× MIC) served as a positive control. (F,G) Antagonization of the bioreporter induction by peptidoglycan precursor addition. In the case of ReBpyCtz (F), purified lipid II (triangles) antagonized the most effectively, and in the case of clotrimazole (G), C55PP antagonized the best. C55P was not significantly effective. The depicted experiment is representative of two independent biological replicates.

The fact that ReBpyCtz had triggered a prominent accumulation of the ultimate cytoplasmic peptidoglycan precursor UDP-MurNAc-PP suggested a blockade at a later stage of peptidoglycan biosynthesis. Therefore, we focused our follow-up experiments on the membrane-associated stages of this pathway (Figure A) and, in particular, on the reactions leading to the membrane-standing precursors undecaprenyl-pyrophosphate-MurNAc-PP (lipid I) and undecaprenyl-pyrophosphate-MurNAc-PP-N-acetyl glucosamine (lipid II). In a first step, using a crude membrane preparation of Micrococcus luteus (Figure B), we studied the formation of lipid II from soluble precursors UDP-MurNAc-PP and UDP-GlcNAc and membrane carrier C55P in a coupled reaction series that involved the MraY and MurG enzymes contained in the membrane. Here, ReBpyCtz showed a slight but reproducible trend of inhibition of lipid II formation. Next, we split the metabolic cascade into the individual enzyme reactions using purified, recombinant enzymes and purified substrates. The MurG-mediated conversion of lipid I to lipid II was clearly inhibited by ReBpyCtz (Figure C,D), while the MraY-mediated formation of lipid I was not affected (Supporting Information Figure S7).

Inhibition of the MurG reaction can occur by direct interference with the catalytic function of the enzyme or by steric hindrance through sequestration of the peptidoglycan precursor serving as a substrate. To test if ReBpyCtz can bind to peptidoglycan precursors, we used another bioreporter strain. The luciferase of Photorhabdus luminescens yields a continuous light signal, and fused to the liaI promotor, it is a sensitive sensor of an interference of an agent with the lipid II cycle.[34]ReBpyCtz strongly induced this bioreporter strain at 1× MIC (Figure E and Supporting Information Figure S8). Capitalizing on this stable signal, we tested if the external addition of certain peptidoglycan precursors to the reporter strain might be capable of antagonizing the ReBpyCtz-triggered signal. Indeed, when purified lipid II was added to the supernatant of the bioreporter assay mixture, it prevented ReBpyCtz from eliciting the liaI stress response in the bioreporter strain (Figure F). The explanation is that lipid II supplemented to the medium forms a complex with ReBpyCtz, thereby preventing the CORM conjugate from binding to the native lipid II in the B. subtilis cell membrane. ReBpyCtz clearly displayed its potential to bind lipid II and also to a lesser extent C55PP. In contrast, Ctz showed the highest affinity for C55PP (Figure G), suggesting that ReBpyCtz and Ctz have a different molecular binding mode.

Discussion

Most of the so-far reported CORMs derived from the fac-{M(CO)3}d6 fragment (M = Mn and Re) only reveal biological activity, and in particular antibacterial activity, when irradiated to promote the release of CO.[10,35−37] In contrast, complex MnBpyCtz is active without irradiation and stands out for its high activity against Gram-positive bacteria. Particularly interesting is the fact that its bactericidal activity exceeds that of its clotrimazole ligand.[16]

As these characteristics posed several questions, including the contribution of CO to the enhanced antibacterial activity of MnBpyCtz when compared to free Ctz, we synthesized compounds with various modifications relative to the structure of MnBpyCtz and studied their antibacterial potency. The complexes have the general formulation [M(CO)3(L–L)(Ctz)][OTf], where M = Mn and Re and L–L are bidentate N–N, N–C, and C–C ligands. All the N–N ligands are di-imines based on N-heterocyclic rings. The C ligands are imidazolidenes or NHC. All complexes are positively charged, have the same triflate counterion, and cover a continuum of electronic density at the metal that places the νsymCO vibrations between 2038 cm–1 (for MnBpyCtz) and 2013 cm–1 for MnbisNHCCtz.

The overall data indicate that MnBpyCtz and its analogue ReBpyCtz are active against Gram-positive bacteria, and the introduction of bulky substituents on the Bpy ligand cancels or strongly impairs the bactericidal activity. The substitution of Bpy with biquinoline or other small bicyclic ligands with N(sp2) or C(sp2) binding atoms such as py-imidazolidene or bis-imidazolidene also leads to highly active clotrimazole conjugates. The fact that the pyridine-benzimidazole ligand in MnPyBzimCtz causes an intermediate effect suggests that there is a net stereochemical component driving the activity of this family of complexes. Electronic factors do not seem to play any role as the activity of MnBpyCtz, with the lower electronic density (highest νsymCO), is close to that of the very electron rich bis–carbene complex MnbisNHCCtz (lowest νsymCO).

The presence of the clotrimazole ligand is necessary, though not sufficient, to achieve antibacterial activity of the conjugates. Indeed, while MnBpyCtz, ReBpyCtz, MnBiqCtz, MnPyNHCCtz, and MnbisNHCCtz are much more active than clotrimazole alone, conjugate MnPyBzimCtz is less active than free Ctz, and MnBpydinonCtz lacks activity. Since all of these conjugates share a common [M(CO)3(Ctz)]+ fragment and neither the metal nor their ligand-controlled electronic properties play a role, we are compelled to assume that the structural properties of the bidentate L–L ligands are the main determinant of activity. However, the structural properties of the precursors do not per se confer activity as none of the conjugate precursors with the formula [M(CO)3(L–L)Br] revealed any bactericidal activity, regardless of the nature of the L–L ligand. This result suggests that the lack of a positive charge in these organometallic precursors hinders their interaction with the bacterial cells, rendering them inactive. Indeed, all Ctz conjugates bear a positive charge that by itself is not a sufficient condition for potent antibacterial activity, as that of conjugate MnBpydinonCtz demonstrates. The positive charge effect is well-recognized to be strongly implicated in the mechanism of many antibacterial agents including [Re(CO)3(L3]+ derivatives.[13,38,39] Although electronic properties do not appear to be determinants of the antibacterial activity as such, they can still play a role in determining the chemical stability of the conjugates and hence in the release of CO. Since we observed that none of the conjugates decomposed in whole blood, that is, in the presence of plasma proteins and erythrocytes, under normoxic conditions, we must acknowledge their significant stability in biological media.

In contrast to the literature examples already mentioned above,[10,36,37] light irradiation is not necessary to trigger a bactericidal effect in clotrimazole conjugates, and CO is released into S. aureus cells treated with MnBpyCtz, ReBpyCtz, MnBiqCtz, MnPyBzimCtz, MnPyNHCCtz, and MnbisNHC. The stability toward CO release demonstrated by these conjugates in normoxic blood and the absence of the need for photochemical decomposition support the notion that the conjugates enter the bacterial cells structurally intact and only decompose to release CO once attacked by the cell’s chemical machinery.

The contribution of CO to the bactericidal activity of CORMs is well-established through binding and impairment of essential bacterial heme proteins, such as those of the respiratory chain.[40,41] Moreover, evidence has accumulated that CO inhibits not only heme proteins but also iron proteins.[42−44] Previous studies showed that the cationic photoCORM [Mn(CO)3(κ3-tpa)]Br is not internalized, and its killing effect is driven by the CO released upon irradiation and diffusion into the cells. Meanwhile, the Mn scaffold of the photoCORM remaining outside of the cell generates highly toxic HO radicals through the reaction with the H2O2 that leaks from the cell.[11] In contrast, our present study demonstrates that bactericidal Ctz–CORM conjugates enter the bacterial membrane.

The clotrimazole ligand bestows a new quality to the CORMs, enabling their membrane insertion and potentially even passage into the cytoplasm. Nevertheless, the conjugated Ctz–CORM ReBpyCtz is more than the sum of its parts. Although precursor ReBpyBr has no antibacterial activity and Ctz shows only moderated MICs, the potency of ReBpyCtz is 1 order of magnitude higher than that of Ctz, while its spectrum of bacterial target species is expanded toward Gram-negative bacteria. In general, the activity of ReBpyCtz is higher against Gram-positive bacteria than against Gram-negative bacteria, in accordance with the high molecular weight of the compound (920 g/mol) and the resulting difficulties in crossing the Gram-negative outer membrane. When ReBpyCtz is added to S. aureus cells, the topology of the cytoplasmic membrane is immediately and severely disturbed. The rapid aggregation of the FM 5-95 dye in large clusters is probably just a direct illustration of what may be happening to various other membrane components that can no longer take their normal positions and are prevented from their normal interactions. The membrane potential dissipates, although severe depolarization only takes place at concentrations well above the MIC, indicating that membrane depolarization is not the primary cause of cell death. There is no indication of prominent pores in the cytoplasmic membrane of S. aureus, as the cell’s intake of the PI dye was still precluded after 2 h of treatment at 4× MIC.

Apart from a general impact on the membrane topology and energy functions, ReBpyCtz demonstrated inhibitory effects on peptidoglycan biosynthesis in several independent assays. The liaI promotor was strongly induced, known to prominently signal disturbance by compounds that interfere with cycling of the undecaprenyl precursor in peptidoglycan biosynthesis (“lipid II cycle”).[34] Accordingly, the ultimate cytoplasmic peptidoglycan precursor UDP-MurNAc-pentapeptide accumulated in the cytoplasm of treated cells. Dissecting the peptidoglycan synthesis pathway via in vitro assays with purified enzymes and substrates revealed that ReBpyCtz inhibited the MurG-mediated conversion of lipid I to lipid II but not the preceding MraY-mediated formation of lipid I, indicating a marked preference for the MurG reaction. Effective antagonization of the liaI stress response via the addition of external lipid II indicates that ReBpyCtz has the potential to bind to lipid-bound peptidoglycan precursors, and the binding affinity was higher for lipid II than for C55PP. The observed inhibition of the MurG reaction in vitro implies that lipid I is also recognized and bound, thus offering a rationale for the inhibition of the reaction by substrate complexation rather than by direct enzyme inactivation. A further indication for lipid I binding is the result that the reaction product lipid I accumulated in the MraY assay as a result of shifting the reversible enzyme reaction to the product side.

Contemplating the main cause of the bacterial death, our data show that B. subtilis and S. aureus cells signal a disturbance of peptidoglycan synthesis already close to the MIC (0.5–2× MIC), as evidenced by the bioreporter studies and UDP-MurNAc-PP accumulation, respectively, while prominent membrane potential depolarization requires higher concentrations. At 4× MIC, the membrane potential of ReBpyCtz is only transiently reduced, and 16× MIC is required for a permanent effect. It is also noteworthy that none of the other bioreporters, covering a broad range of classical antibiotic target areas, gave a signal. Therefore, it seems likely that inhibition of peptidoglycan synthesis contributes to the killing of bacterial cells by ReBpyCtz. However, the strong aggregation of dye molecules in the membrane that starts within minutes of compound addition is not a typical feature of peptidoglycan synthesis inhibitors and suggests additional damage based on an overall disturbance of the membrane topology. In addition, the observed CO release has to be taken into account, which probably also contributes to cell death in a multifarious manner.

Interestingly, Metzler-Nolte and co-workers observed similar pleiotropic antibacterial effects while studying derivatives of the [Re(CO)3(dpa-R)]+ type, where the R substituents contain short PNA chains that can be terminated by organometallic moieties such as ferrocene, ruthenocene, or cymanthrene or simply by organic isosteres. Several of these complexes showed antibacterial activity against S. aureus in the same range of low micromolar MIC values, as did ReBpyCtz. Also, in their case, the dpa-R ligand alone, that is, without the {Re(CO)3}+ fragment, was innocuous, just like other [Re(CO)3(dpa-R′)]+ complexes, where R′ did not contain the PNA chains. Moreover, the PNA chain-containing fragment was also devoid of any bactericidal activity. Consistent with our results, only the combined molecular constructions formed by the same fragment fac-{Re(CO)3}d6 and linked to particular organic components exhibited bactericidal activity.

The antibacterial potential of ReBpyCtz is clearly greater than that of either precursor ReBpyBr or Ctz alone, and it is tempting to speculate which component might have contributed which feature to the resulting hybrid. The membrane is the established target site of clotrimazole, and as an antifungal drug, it is used against lanosterole-14-demethylase in fungal ergosterol biosynthesis.[45] With its imidazole portion, Ctz binds to iron in the protoporphyrin ring at the active site of the enzyme, and with its remaining part, it interacts with the apo-enzyme. Activities against further membrane-standing enzymes and ion channels have also been reported in mammalian cells.[45] Apart from directly interfering with membrane enzymes, clotrimazole was shown to bind to lipids, and particularly, the addition of unsaturated phospholipids effectively antagonized its antimycotic activity against C. albicans.[46] The effects that we observed in our current study for bacterial membranes are consistent with these previous results. We here report undecaprenyl-pyrophosphate and to a lesser degree lipid II as additional lipid species that can be bound by Ctz. While Ctz demonstrated the highest affinity for undecaprenyl-pyrophosphate among the peptidoglycan precursors tested, ReBpyCtz showed binding preference for lipid II, which contains a larger head group. The second component, ReBpyBr, may bestow additional interaction sites to the hybrid, enhancing the affinity for the carbohydrate portion of lipid I and II and allowing for stronger peptidoglycan synthesis inhibition.

In conclusion, our results clearly demonstrate the enhanced antibacterial activity of ReBpyCtz compared to that of free clotrimazole and shed light on the MoA of this kind of conjugated CORMs. It involves a sequence of events initiated by fast and efficient membrane insertion that is most probably followed by a pleiotropic mechanism of bacterial growth inhibition. The cytoplasmic membrane is disorganized, peptidoglycan synthesis is blocked, CO is released, and the membrane potential breaks down. The overall outcome is the killing of the pathogens at a much lower concentration than that with clotrimazole alone. Thus, the promoter effect of these [M(CO)3(L–L)]+ (M = Mn and Re) fragments can be potentially exploited to increase the activity of other antibiotics, contributing to overcome the developed antibiotic resistance shown by several human pathogens.

Materials and Methods

Synthesis of CORMs

All manipulations were carried out under a dry nitrogen atmosphere using standard Schlenk techniques. Solvents for reactions and column chromatography were degassed and distilled from the indicated drying agents (in brackets): tetrahydrofuran, abbrv. THF (Na), diethyl ether (Na), methanol (Na), dichloromethane (CaH2), hexane (CaH2), and ethyl acetate (Na2SO4). Acetone was dried over 4 Å molecular sieves under a nitrogen atmosphere, and water was degassed. Mn(CO)5Br and Re(CO)5Br were acquired from Strem Chemicals (Europe). 2,2′-Bipyridyl (Bpy), 2,2′-biquinoline (Biq), 2-pyridyl-benzimidazole, pyridyl-imidazole-2-ylidene (PyNHCMe), 4,4′-dinonyl-2,2′-bipyridyl (Bpydinon), and clotrimazole (Ctz) were acquired from Sigma-Aldrich, and Ag(CF3SO3) was acquired from Alfa Aesar. Literature methods were used to prepare the following complexes: Mn(CO)3Br(Bpy),[47] Re(CO)5Br,[48] Re(CO)3Br(Bpy),[49]fac-[Mn(CO)3(Biq)Br],[21]fac-[Mn(CO)3(2-pyridyl-benzimidazo-le)Br],[50] Mn(PyNHCMe)(CO)3Br,[51] and Mn(BisNHCMe)(CO)3Br, in which BisNHCMe is bis-imidazole-2-ylidene.[22] The purity of the compounds used in the biological tests was established as ≥95% via C, H, N, and S elemental analyses performed by the Elementary Analysis service of Instituto de Tecnologia Química e Biológica António Xavier, Universidade Nova de Lisboa, Oeiras, Portugal, using a LECO TruSpec CHN Micro elemental analyzer.

Complex MnBpyCtz was prepared with triflate instead of PF6– as the counterion using a slightly modified previously published protocol,[16] and the FTIR, NMR, and elemental analyses were in agreement with the earlier data.

NMR spectra were recorded on a Bruker AVANCE III 400 spectrometer. Chemical shifts are quoted in parts per million (ppm) from the residual protic solvent signal (acetone-d6: 1H 2.05 ppm; CD2Cl2: 1H 5.32 ppm; and DMSO-d6: 1H 2.50 ppm and 13C 39.52 ppm), and coupling constants J are given in Hertz.

FTIR spectra (KBr pellets or ATR) were recorded in an Unicam Mattson 7000 spectrophotometer. Intensities of the stretching vibrations are marked as very strong (vs), strong (s), medium (m), and weak (w). HRMS–electrospray ionization MS (ESI-MS) mass spectra were recorded using a Micromass Quatro LC instrument; nitrogen was employed as a drying and nebulizing gas. The synthesis of the several compounds was pefrormed as follows.

1-fac-[Mn(CO)3(Bpy)Ctz][OTf]: to an orange solution of Mn(CO)3(Bpy)Br (150 mg; 0.4 mmol) in acetone (12 mL) was added Ag(CF3SO3) (123 mg; 0.48 mmol), and the solution was left stirring at room temperature for 3 h in the dark. The yellow solution was filtered to discard AgBr, followed by the addition of Ctz (166 mg; 0.48 mmol), and the suspension was left stirring overnight at room temperature. The solution was then evaporated to dryness, and the residue was washed with diethyl ether. A yellow powder was isolated by filtration after recrystallization from CH2Cl2/diethyl ether (1:8) at −20 °C for 48 h and dried (250 mg, 79%). Anal. Calcd (%) for C36H25ClF3MnN4O6S, (789.1): C, 54.8; H, 3.19; N, 7.10; S, 4.06. Found: C, 54.8; H, 3.25; N, 6.89; S, 4.26. FTIR (KBr; νCO; cm–1): 1941 (vs), 2038 (vs). 1H NMR (400 MHz, DMSO-d6, 25 °C): δ (ppm) 9.22 (d, J = 5.0 Hz, 2H), 8.59 (d, J = 8.0 Hz, 2H), 8.26 (td, J = 8.0, 1.3 Hz, 2H), 7.73–7.69 (m, 2H), 7.46 (td, J = 7.9, 1.4 Hz, 1H), 7.40–7.31 (m, 8H), 6.98 (dt, J = 12.3, 1.4 Hz, 2H), 6.72 (d, J = 6.9 Hz, 4H), 6.68 (dd, J = 7.9, 1.4 Hz, 1H), 6.53 (s, 1H).

2-fac-[Re(CO)3(Bpy)Ctz][OTf]: to a greenish solution of Re(CO)3(Bpy)Br (202 mg; 0.4 mmol) in acetone (30 mL), Ag(CF3SO3) (123 mg; 0.48 mmol) was added, and the solution was left stirring at room temperature for 3 h in the dark. The solution was filtered to discard AgBr, and after the addition of Ctz (166 mg; 0.48 mmol), the suspension was left stirring overnight at room temperature. The solution was then evaporated to dryness, and the solid was washed with diethyl ether. A yellow powder was isolated by filtration after recrystallization from CH2Cl2/diethyl ether (1:8) at −20 °C for 5 days and dried (239 mg; 65%). Anal. Calcd (%) for C36H25ClF3ReN4O6S, (920.3): C, 47.0; H, 2.74; N, 6.09; S, 3.48. Found: C, 46.8; H, 2.72; N, 5.95; S, 3.56. FTIR (KBr; νCO; cm–1): 1914 (vs), 1936 (vs), 2028 (vs). 1H NMR (400 MHz, DMSO-d6, 25 °C): δ (ppm) 9.12 (d, J = 5.2 Hz, 2H), 8.68 (d, J = 8.1 Hz, 2H), 8.34 (td, J = 8.1, 1.2 Hz, 2H), 7.75–7.72 (m, 2H), 7.47 (td, J = 7.7, 1.2 Hz, 1H), 7.40–7.31 (m, 8H), 7.03 (d, J = 19.1 Hz, 2H), 6.82 (s, 1H), 6.7–6.71 (m, 5H).

3-fac-[Mn(CO)3(Biq)Ctz][OTf]: fac-[Mn(CO)3(Biq)Br] (124 mg; 0.28 mmol), Ag(CF3SO3) (77.3 mg; 0.3 mmol), and Ctz (97.3 mg; 0.28 mmol) were dissolved in acetone (30 mL). The mixture was warmed to 50 °C, stirred overnight, and then dried. The oily residue was mixed in toluene (25 mL) and filtered through a Celite plug to remove AgBr. After the evaporation of toluene, the residue was dissolved in MeOH (10 mL). The solution was covered with a layer of diethyl ether (40 mL) and kept in the dark at 5 °C. Red crystals that were recovered by filtration were washed with ether and dried under high vacuum. Anal. Calcd (%) for: C44H29N4ClMnSF3O6, (889.2): C, 59.43; H, 3.29; N, 6.3; S, 3.6. Found: C, 59.24; H, 3.19; N, 5.91; S, 3.99. FTIR (νCO; cm–1): 2031 (s), 1943 (vs). HRMS ESI-MS (positive mode): 739.1304 [M]+; 277.0776 [Ph2(C6H4Cl)C]+. 1H NMR (400 MHz, acetone-d6, 25 °C): δ (ppm) 8.97 (dd, J = 8.8 Hz, 4H), 8.83 (d, J = 8.6 Hz, 2H), 8.33 (d, J = 8.2 Hz, 2H), 8.26 (t, J = 8.2 Hz, 2H), 8.00 (t, J = 7.4 Hz, 2H), 7.52–7.24 (m, 10H), 6.84 (s, 1H), 6.78 (d, J = 7.6 Hz, 4H), 6.26 (s, 2H).

4-fac-[Mn(CO)3(2-pyridyl-benzimidazole)(Ctz)][OTf]: fac-[Mn(CO)3(2-pyridyl-benzimidazole)Br] (151 mg; 0.36 mmol), Ag(CF3SO3) (93.4 mg; 0.36 mmol), and Ctz (126 mg, 0.37 mmol) were dissolved in acetone (30 mL). The mixture was warmed to 50 °C, stirred overnight, and then left to dry. The oily residue was mixed in toluene (25 mL) and filtered through a Celite plug to remove AgBr. After the evaporation of toluene, the resulting yellow-brown foam was ground in hexane at very low temperature (thawing from liquid nitrogen solidification), producing a yellow powder that was separated by decantation and chromatographed in a silica gel column. After elution with dichloromethane, the product was treated with acetone and dried in vacuum to produce a yellow-greenish foam, which was dried under high vacuum. Anal. Calcd (%) for C38H26N5ClMnSF3O6, (828.1): C, 55.12; N, 8.46; H, 3.16; S, 3.87. Found: C, 55.31; N, 8.20; H, 3.43; S, 3.95. FTIR (νCO; cm–1): 2024 (s), 1906 (vs). HRMS ESI-MS (positive mode): 678.1090 [M]+; 277.0776 [Ph2(C6H4Cl)C]+. 1H NMR (400 MHz, acetone-d6, 25 °C): δ (ppm) 9.27 (d, J = 5.3 Hz, 1H), 8.23 (d, J = 6.8 Hz, 1H), 7.91 (t, J = 4.4 Hz, 1H), 7.68 (t, J = 6.0 Hz, 1H), 7.48–7.18 (m, 15H), 6.93 (s, 1H), 6.85 (s, 1H), 6.74 (t, J = 6.9 Hz, 4H), 6.66 (s, 1H), 6.59 (d, J = 7.9 Hz, 1H).

5-fac-[Mn(CO)3(4,4′-dinonyl-2,2′-bipyridyl)(Ctz)][OTf]: a solution of Mn(CO)5Br (275 mg, 1 mmol) magnetically stirred in acetone (40 mL) was treated with Ag(CF3SO3) (257 mg, 1 mmol) for 10 min at room temperature in the dark and then refluxed for 1 h to form [Mn(CO)3(OCMe2)3][CF3SO3].[52] The precipitate of AgBr was filtered off, and 4-4′-dinonyl-2-2′-bipyridyl (407 mg; 1 mmol) was added, followed by Ctz (345 mg; 1 mmol). The mixture was allowed to react for 1 h at reflux and then concentrated to give an oily residue, which was washed with hexane and dried. The oily residue was chromatographed on an alumina column under nitrogen and eluted with ethyl acetate/CH2Cl2. The yellow fraction collected was left to dry, which originated a yellow foam that was dried in vacuum. Anal. Calcd (%) for C54H61N4ClF3O6MnS, (1041.5): C, 62.27; H, 5.90; N, 5.38; S, 3.08. Found: C, 61.90; H, 6.15; N, 5.26; S, 3.10. FTIR (νCO; cm–1): 2031 (s), 1929 (vs). 1H NMR (400 MHz, CD2Cl2, 25 °C): δ (ppm) 8.92 (d, J = 5.6 Hz, 2H), 8.00 (s, 2H), 7.41–7.25 (m, 12H), 6.82 (d, J = 7.5 Hz, 4H), 6.72 (s, 1H), 6.67 (d, J = 6.4 Hz, 2H), 2.81 (t, J = 7.4 Hz, 4H), 1.69 (m, J = 6.3 Hz, 4H), 1.38 (s, 4H), 1.29 (s, 20H), 0.88 (t, J = 6.7 Hz, 6H).

6-fac-[Mn(N-methyl-N′-2-pyridylimidazol-2-ylidene)(CO)3Ctz][OTf]: silver triflate (135 mg, 0.5 mmol) was added to a solution of Mn(PyNHCMe)(CO)3Br (200 mg; 0.47 mmol) in degassed acetone (15 mL), and the mixture was stirred in the dark for 3 h. The solution was filtered under a nitrogen atmosphere, and Ctz (181 mg; 0.53 mmol) was added to the filtrate. After stirring the mixture in the dark for 20 h, the solvent was removed, and the residue was recrystallized from CH2Cl2/diethyl ether. The resulting precipitate was collected and washed with ether to afford the product as a pale solid. Yield: 104 mg, 23%. Anal. Calcd (%) for C35H26MnN5O6ClF3S, (792.1): C, 53.07; H, 3.31; N, 8.84; S, 4.05. Found: C, 52.70; H, 3.19; N, 9.12; S, 3.99. FTIR (KBr, νCO; cm–1): 2030 (s), 1941 (s), 1919 (s). HRMS ESI-MS-(positive mode): 642.1098 [M]+. 1H NMR (400 MHz, DMSO-d6, 25 °C): δ (ppm) 8.41 (s, 1H), 8.24 (t, J = 7.7 Hz, 1H), 8.08 (d, J = 8.4 Hz, 1H), 7.72 (s, 1H), 7.50–7.42 (m, 3H), 7.38–7.33 (m, 7H), 6.98 (s, 1H), 6.89 (s, 1H), 6.79 (t, J = 6.1 Hz, 4H), 6.72 (d, J = 7.9 Hz, 1H), 6.58 (s, 1H), 4.02 (s, 3H). 13C NMR (DMSO-d6, 25 °C): δ (ppm) 223.20 (CO), 217.95 (CO), 214.93 (CO), 197.88 (Mn-Ccarbene), 153.58, 152.10, 142.51, 140.53, 139.44, 139.17, 138.02, 134.32, 132.32, 131.35, 131.05, 130.82, 129.22, 129.10, 128.62, 128.44, 127.72, 127.45, 124.11, 123.62, 117.67, 112.53, 75.64 (NCH2N), 37.81 (NCH3).

7-fac-[Mn(methylene bis-N-methylimidazole-2-ylidene)(CO)3(Ctz)][OTf]: silver triflate (91 mg; 0.35 mmol) was added to a solution of Mn(BisNHCMe)(CO)3Br (117 mg; 0.29 mmol) in degassed acetone (10 mL), and the mixture was stirred in the dark for 3 h. The solution was filtered under a nitrogen atmosphere, and Ctz (122 mg; 0.35 mmol) was added to the filtrate. After stirring in the dark for 20 h, the solvent was removed, and the residue was recrystallized via diffusion with CH2Cl2/diethyl ether. The resulting precipitate was collected and dried under vacuum to obtain the product as a pale solid. Anal. Calcd (%) for C35H29MnN6O6ClF3S.0.2CH2Cl2, (809.1): C, 51.18; H, 3.59; N, 10.17; S, 3.88. Found: C, 50.69; H, 3.41; N, 10.25; S, 4.41. FTIR (KBr, νCO; cm–1): 2013 (s), 1927 (s), 1899 (s). HRMS ESI-MS-(positive mode): 659.1362 [M]+. 1H NMR (400 MHz, DMSO-d6, 25 °C): δ (ppm) 7.58 (d, J = 1.5 Hz, 2H), 7.53–7.49 (m, 2H), 7.44 (d, J = 1.5 Hz, 3H), 7.41–7.36 (m, 6H), 7.11 (s, 1H), 6.95–6.89 (m, 5H), 6.09–5.91 (m, 2H), 5,76 (s, 1H), 3.89 (s, 1H), 3.48 (s, 6H) 13C NMR (DMSO-d6, 25 °C): δ (ppm) 220.74 (CO), 218.35 (CO), 188.73 (Mn-Ccarbene), 141.79, 139.23, 138.29, 133.99, 132.40, 131.00, 130.53, 128.91, 128.42, 127.75, 124.29, 123.66, 122.13, 75.60 (NCH2N), 36.96 (NCH3). Note: it was not possible to completely remove crystallization CH2Cl2 and improve the purity of the crystalline compound.

Bacterial Strains

The bacterial strains used in this work were E. coli K12 MG1655; S. enterica serovar Typhimurium SL1344; S. aureus USA300 JE2 (MRSA) (NARSA, Chantilly, VA, United States); S. aureus NCTC8325; S. aureus ATCC29213; B. subtilis 1S34; B. subtilis 168 strain TMB1617; E. faecium BM4147-1; Klebsiella pneumoniae ATCC12657; Enterobacter aerogenes ATCC13048; P. aeruginosa ATCC27853; and Acinetobacter baumannii 09987.

Growth Rate Determination

Overnight cultures were used to inoculate fresh media to an optical density at 600 nm (OD600nm) of 0.1. When cells reached an OD600nm of 0.3, they were treated with the compounds or with 1% (v/v) of the respective solvent. The growth rates were determined at up to 24 h for at least three biological samples. Unpaired t-test was performed for each condition using GraphPad Prism version 7.0 for iOS, GraphPad Software, La Jolla California USA.

MIC Determination

The MIC values represented in Table were determined using the broth microdilution method in cation-adjusted Mueller–Hinton broth according to the guidelines of the Clinical and Laboratory Standards Institute (CLSI), as previously described.[53] Briefly, a 2-fold serial dilution of the compound (starting from 32 μg/mL) was prepared in polystyrene microtiter plates and seeded with a final bacterial inoculum of 5×105 colony-forming units (cfu) per mL. After overnight incubation at 37 °C, while shaking at 90 rpm, in ambient air, the lowest compound concentration preventing visible bacterial growth was considered the MIC. For the values presented in Table S2, the microtiter plates were inoculated with an overnight culture of E. coli, S. aureus, and S. enterica with approximately 107 cfu/mL and incubated for 18 h at 37 °C while shaking at 90 rpm. MIC values were determined in triplicate for at least three independent experiments.

Mammalian Cell Cytotoxicity Assay

The porcine kidney epithelial cell line (LLC-PK1) was grown in Dulbecco’s modified Eagle’s medium (DMEM) supplemented with 10% (v/v) heat-inactivated foetal bovine serum (FBS) and a 1% (v/v) penicillin–streptomycin mixture. The human hepatoma cell line (HepG2) was cultured in Eagle’s minimum essential medium supplemented as described above. Cells were distributed into 96-well plates at a density of 10,000 cells/well. The plates were incubated for 24 h for cell seeding. On the second day, the compounds were administered, and three independent biological samples were tested in triplicate. After 24 h, the culture medium was discarded, and 100 μL of fresh medium containing 0.5 mg/mL MTT, preheated at 37 °C, was added to each well. The cells were then incubated in the dark for 1 h at 37 °C, in a 5% CO2 atmosphere. After the removal of the medium, the formazan crystals were solubilized with 100 μL of DMSO. The absorbance was measured spectrophotometrically at 550 nm in a Multiskan GO microplate spectrophotometer, Thermo Scientific. The IC50 value was calculated using software Prism 5 (GraphPad) by fitting the best dose–response curve to the normalized data set.

In Vitro CO Release in Blood and Bacteria

CORM solutions were prepared in DMSO, and 50 μL of the solution swas added to 1 mL of sheep whole blood in Alsever’s solution (TCS Biosciences Ltd, UK) and incubated at 37 °C. A final concentration of 0.3 mg mL–1 of CORM was used to mimic an in vivo dosage of approximately 25 mg kg–1 for an average 20 g mouse. To follow the increase of the CO–hemoglobin (COHb) levels, samples were analyzed over time in an Avoximeter 4000 from A-vox Instruments Inc. The released CO was calculated based on the amount of the compound initially added, the total amount of hemoglobin used, and the %COHb formed. Control spectra of untreated blood samples were recorded. The experiment was performed with two different blood batches.

Determination of the CO released from CORMs into bacteria was carried out using carbon monoxide probe-1, COP-1, as previously described.[54] Briefly, bacterial cells grown aerobically in LB to an OD600 of ∼0.3 were incubated with the selected CORMs for 15 min at 37 °C and 150 rpm. Bacteria were incubated with 1 μM COP-1 for 15 min, after which an aliquot was obtained, washed with phosphate-buffered saline (PBS), and analyzed in a Leica DM6000B fluorescence microscope, with a phase contrast Uplan F1 100× objective and a CCD Ixon camera (Andor Technologies). All samples were excited at 488 nm with filter set for fluorescein isothiocyanate, and the emission was collected using a META detector at 525 nm. The microscope images were analyzed using ImageJ software version 2.3.0/1.53f. Images were captured from at least three biological samples, and in total, at least 300 cells were analyzed.

Bioreporter Assays

For target pathway detection, induction of the firefly luciferase linked to the promoters of yorB, helD, yheI, ypuA, or liaI in the genetic background of B. subtilis 1S34 was determined as previously described.[55] Reporter signals were recorded in either Belitzky minimal medium (BMM) (yheI strain) or LB medium (all other strains). Serial 2-fold dilutions of the test compounds (starting at 4× MIC for Ctz and ReBpyCtz and at 64 μg/mL for ReBpyBr) in 60 μL of LB or BMM were prepared in white 96-well flat-bottom polystyrene microtiter plates, and plates were inoculated with 60 μL of a bacterial suspension adjusted to an OD600nm of 0.02. Plates were incubated at 37 °C for a predetermined time depending on the induction kinetics of the reporter strain: 1 h for the liaI and ypuA reporters, 1.5 h for the helD strain, 3.5 h for the yorB strain, and 4 h for the yheI strain. Then, 60 μL of citrate buffer (0.1 M, pH 5) containing 2 mM luciferin (Serva) was added, and flash luminescence was recorded using a microtiter plate reader (TECAN infinite M200).

For the bioreporter quenching assay with purified peptidoglycan precursors, B. subtilis 168 strain TMB1617 was employed, in which the liaI promotor was fused to the bacterial luciferase from P. luminescens (Plia-lux),[56] and the assay was conducted as described previously.[57] The bioreporter was grown to an OD600nm of 0.5 at 30 °C in cation-adjusted Mueller–Hinton broth containing 5 μg/mL chloramphenicol and added to microtiter plates containing 2-fold dilution series of ReBpyCtz, ReBpyBr, or Ctz. Luminescence was measured every 15 min for 10 h at 30 °C using a Spark 10M microplate reader (Tecan). Purified cell wall precursors lipid II, C55PP, or C55P were tested for their ability to antagonize the stress response. The peptidoglycan precursors were added in a 10-fold molar excess over the test compounds and preincubated with the compounds for 15 min, before the cells were added, and luminescence was measured as described above.

Membrane Labeling in Live Cell Microscopy

FM 5-95 labeling after antibiotic exposure was conducted as described.[58] Briefly, S. aureus NCTC8325 was grown in LB medium at 37 °C and 200 rpm to an OD600 of 0.3. Culture aliquots were treated with either Ctz (8 μg/mL, 2× MIC), ReBpyBr (64 μg/mL), or ReBpyCtz (0.5 μg/mL, 2× MIC), using DMSO (1%) as a negative control, and incubated at 37 °C for 30 min. Samples were labeled with 20 μg/mL N-(3-trimethylammoniumpropyl)-4-(6-(4-(diethylamino)phenyl)hexatrienyl) pyridinium dibromide (FM 5-95, Molecular Probes) and visualized via fluorescence microscopy (λEx 506 nm and a window for emission from 650 to 700 nm) on microscope slides covered with a thin film of 1% agarose using a Nikon Eclipse Ti-E microscope. Images were acquired on an Orca Flash 4.0 LT camera (Hamamatsu) using software NIS Elements Advanced Research (Nikon) and further analyzed and processed using software NIS Elements.

Microscopic Cell Wall Integrity Assay and HADA Incorporation Microscopy

B. subtilis 168 was grown in cation-adjusted Mueller–Hinton medium at 37 °C and 200 rpm to an OD600 of 0.35. Aliquots (100 μL) of the cells were treated with Ctz (8 μg/mL, 4× MIC), ReBpyCtz (1 μg/mL, 4× MIC), and ReBpyBr (64 μg/mL). Vancomycin (2 μg/mL, 8× MIC) and DMSO (1%) served as positive and negative controls, respectively. After 30 min incubation, 25 μL of cells was transferred to a fresh tube containing 100 μL of a 1:3 (v/v) mixture of acetic acid and methanol. Samples were imaged using brightfield microscopy on microscope slides covered with a thin film of 1% agarose using a Zeiss Axio Observer Z1 automated microscope. Images were acquired using an Orca Flash 4.0 V2 camera (Hamamatsu) and an alpha Plan-Apochromat 100×/1.46 Oil Ph3 objective (Zeiss). Images were processed using the Zen software package (Zeiss).

To monitor the effect of CORMs on peptidoglycan synthesis microscopically, we stained B. subtilis cells with the fluorescent d-amino-acid HCC-amino-d-alanine (HADA) probe,[59] which was custom-synthesized by EMC Microcollections (Tübingen). Cells were grown in LB overnight at 37 °C and 180 rpm, diluted 1:100 in fresh medium, and grown at 37 °C to an OD600 of 0.3–0.4. Aliquots of 500 μL in 24-well plates were treated for 5 min at 37 °C and 500 rpm with Ctz (4 μg/mL, 2× MIC), ReBpyCtz (0.5 μg/mL, 2× MIC and 1 μg/mL, 4× MIC), ReBpyBr (64 μg/mL), or carbonyl cyanide 3-chlorophenylhydrazone (CCCP, 5 μM). DMSO (0.5%) served as a negative control. Then, samples were centrifuged (5000g, 5 min), most of the supernatant (450 μL) was removed, and cells were labeled in the remaining 50 μL with a final concentration of 1 mM HADA and mixed briefly by vortexing and incubation for 3.5 min at 37 °C. The reaction was stopped by adding 1 mL of ice-cold PBS, and unbound dye was removed via three washing steps with ice-cold PBS (16,000 rpm, 3 min, 4 °C). Subsequently, the cells were placed on microscopy slides covered with a thin layer of agarose (1.2% in PBS) and visualized via brightfield and fluorescence microscopy in a Zeiss Axio Observer Z1 LSM800 at a λex of 353 nm and a λem of 465 nm. Images were acquired using an Orca Flash 4.0 V2 camera (Hamamatsu) and an α Plan-Apo 100×/1.46 Oil Ph3 objective (Zeiss). To quantitatively assess d-amino-acid incorporation, all samples were stained and imaged under the same conditions. Image processing was performed in FIJI,[60] and the relative fluorescence units of the HADA signals in the septal regions were quantified using MicrobeJ.[61] Cells were identified in the phase contrast channel with a width limit from 0.5 to 1.3 μm and a length above 2 μm. Lysed cells, if present, were excluded from the analysis. The septal HADA labeling intensity was then quantified within individual cells (N > 100), and the data were further analyzed and plotted using GraphPad Prism software. The significance was determined via one-way ANOVA.

Membrane Potential Measurement and Membrane Pore Formation

The membrane potential of S. aureus NCTC8325 was determined as described previously.[62] Briefly, cells grown in LB were resuspended in PBS, loaded with 3,3′-diethyloxacarbocyanine iodide [DiOC2(3), Molecular Probes, Fisher Scientific], and transferred to a black 96-well flat-bottom polystyrene microtiter plate. Baseline measurements were obtained in a microplate reader (TECAN Spark) for 2 min [λex = 485 nm; and two emission wavelengths λem = 530 nm (green) and λem = 630 nm (red)], after which Ctz, ReBpyBr, and ReBpyCtz were added in concentration series, and the measurement was continued as above for a total of 15 min. CCCP (5 μM) was used as a positive control and DMSO (1%) as a negative control.

Pore formation was monitored as described[62] using the Live/Dead BacLight bacterial viability kit (Molecular Probes). S. aureus NCTC8325 was grown in LB and treated for 120 min with either Ctz (16 μg/mL, 4× MIC), ReBpyBr (64 μg/mL), ReBpyCtz (1 μg/mL, 4× MIC), a crude preparation of the pore-forming antibiotic nisin (100 μg/mL, Sigma-Aldrich, positive control), or DMSO (1%, negative control). Then, cells were labeled with a mixture of syto9 and PI for 15 min and visualized via brightfield and fluorescence microscopy in a Zeiss Axio Observer Z1 automated microscope at λex 483 nm/λem 500 nm emission (syto9) and λex 305 nm/λem 617 nm emission (PI) on agarose-covered microscope slides. Images were acquired using an Orca Flash 4.0 V2 camera (Hamamatsu) and an alpha Plan-Apochromat 100×/1.46 Oil Ph3 objective (Zeiss) and processed using the Zen software package (Zeiss).

Quantification of UDP-MurNAc-PP

Metabolite extraction was performed as described.[63] Briefly, S. aureus ATCC29213 was grown in LB medium at 37 °C and 200 rpm to an OD600 of 0.5; then, 130 μg/mL of chloramphenicol was added; and incubation was continued for further 15 min. Aliquots of 5 mL were treated with ReBpyCtz, ReBpyBr, Ctz, or vancomycin as the positive control for 30 min at the indicated concentrations. Cells were cooled on ice, washed with Millipore water, pelleted, and frozen. For analysis, pellets were resuspended in 500 μL of Millipore water and disrupted in a Precellys Evolution homogenizer at 3 × 20 s, 6500 rpm with 30 s intervals using 0.1 mm glass beads. After centrifugation at 14,000 rpm, 4 °C for 20 min, 400 μL of the supernatant was transferred to a fresh tube containing 1.6 mL of acetone [high-performance liquid chromatography (HPLC) grade]. Samples were centrifuged as above for 15 min, and the supernatant was transferred into a fresh tube and dried in a vacuum centrifuge. Dried samples were resuspended in 50 μL of Millipore water, and 5 μL aliquots were analyzed via LC–MS using a Gemini C18 column (150 × 4.6 mm, 110 Å, 5 mm; Phenomenex) and an UltiMate 3000 RS (Dionex) coupled to a micro-TOF II mass spectrometer (Bruker Daltronics) operated in the negative or positive ionization mode, as indicated. Samples were separated using a 30-min gradient of 0–40% acetonitrile at a flow rate of 0.2 mL/min, as described previously.[64] Extracted ion chromatograms (EICs) were used to calculate the area under the curve (AUC) using Prism 8.4 (GraphPad).

Purification of Cell Wall Precursors

Peptidoglycan precursors lipid I and lipid II were synthesized in a large scale as previously described.[65] Concentrations of purified precursors were determined based on their phosphate content, which was measured via a phosphate test according to Rouser et al.(66) Uridine diphosphate-N-acetyl glucosamine (UDP-MurNAc-PP) used for precursor synthesis was purified in a crude preparation according to the method of Kohlrausch and Höltje.[67] Undecaprenyl phosphate (C55P) and undecaprenyl pyrophosphate were purchased from Larodan Fine Chemicals AB (Malmö, Sweden).

In Vitro Lipid II Synthesis with Isolated Bacterial Membranes

Membranes of M. luteus were used to determine the inhibitory activity of the compounds for lipid II synthesis by a crude membrane fraction as described previously.[68,69] 5 nmol C55P was used in a total volume of 50 μL containing 0.6% triton X-100, 10 mM MgCl2, 1 mM uridine diphosphate-N-acetyl glucosamine (UDP-GlcNAc), and 100 mM Tris–HCl, pH 7.5. ReBpyCtz, Ctz, or ReBpyBr was added to lipid carrier C55P in a molar ratio of 10:1 and incubated for 15 min at room temperature. Crude preparations of Staphylococcus simulans 22 containing UDP-MurNAc-PP, as well as membrane preparations of M. luteus, were added to start the reaction. Samples were incubated at 30 °C for 2 h, extracted with an equal volume of n-butanol:pyridine acetate, pH 4.2 (2:1, v/v), and centrifuged for 3 min at 13,500 rpm. The upper phase was applied to thin-layer chromatography (TLC) silica plates and developed in a solvent containing chloroform/methanol/water/ammonia (88:48:10:1, v/v/v/v). Lipid bands were visualized via phosphomolybdic acid staining.[65]

In Vitro Peptidoglycan Biosynthesis with Purified Precursors and Recombinant Purified Enzymes

Synthesis of lipid I and lipid II catalyzed by MraY and MurG, respectively, were performed using purified recombinant enzymes as described previously.[57,70] Briefly, for lipid I synthesis by purified MraY, samples were prepared in a total volume of 50 μL containing 5 nmol substrate C55P, 60 mM MgCl2, 1.25 mM CaCl2, 0.6% triton X-100, and 100 mM Tris–HCl, pH 7.5. Reactions were initiated by the addition of a crude extract containing UDP-MurNAc-PP and the purified MraY enzyme. After incubation at 30 °C for 2 h, reaction mixtures were extracted as mentioned above and analyzed via TLc. MurG-catalyzed lipid II synthesis was conducted using 2 nmol purified lipid I in a final volume of 30 μL that contained 25 nmol UDP-GlcNAc, 0.8% triton X-100, 5.7 mM MgCl2, and 200 mM Tris–HCl, pH 7.5. Samples were incubated at 30 °C for 1 h, extracted, and analyzed via TLC, as described above. In all in vitro reactions, ReBpyCtz, Ctz, and ReBpyBr were added to the purified cell wall precursors in a molar ratio of 10:1 and incubated for 15 min at room temperature, and reactions were started by adding the purified proteins.