Optimization of 4-Substituted Benzenesulfonamide Scaffold To Reverse Acinetobacter baumannii Serum-Adaptive Efflux Associated Antibiotic Tolerance

Abstract

Acinetobacter baumannii is a nosocomial pathogen of urgent concern for public health due to rising rates of multidrug and pandrug resistance. In the context of environmental cues such as growth in human serum, A. baumannii is known to display adaptive efflux, in which a multitude of efflux-associated genes are upregulated, resulting in efflux-mediated drug tolerance in strains that are otherwise susceptible to antibiotic therapy. Previously, we identified a sulfonamide-containing scaffold molecule (ABEPI1) that reversed serum-associated antibiotic tolerance in A. baumannii. Herein, we present structure-activity relationship studies on 29 newly synthesized analogues. These molecules were characterized for their ability to potentiate multiple antibiotics in serum, reduce serum-associated ethidium bromide efflux and depolarize bacterial cell membranes. In addition, they were assessed for toxicity to mammalian cells. Collectively, these molecules may represent promising potential adjuvants for use in combination with new and existing antibiotics to treat A. baumannii bacterial infections.

Graphical Abstract

No hiding here anymore: Acinetobacter baumannii cultured in lysogeny broth is susceptible to minocycline, but in human serum the bacteria upregulates efflux pumps and is no longer susceptible to the drug. This manuscript describes the optimization of analogues that restore minocycline activity in A. baumannii grown in human serum.

Introduction

Acinetobacter baumannii is a Gram-negative opportunistic human pathogen and a leading cause of nosocomial infections worldwide, which commonly manifest as ventilator-associated pneuomonia and central line-associated blood infections^[1] and result in mortality rates in excess of 60% in certain patient populations.^[2] The organism’s ability to persist in the environment for long periods and acquire resistance determinants has propelled A. baumannii’s clinical significance, making it one of the foremost organisms threatening current antibiotic options. Due to its propensity to acquire multidrug resistance (MDR), extensive drug resistance (XDR) and pandrug resistance, A. baumannii is one of the ESKAPE (Enterococcus faecium, Staphylococcus aureus, Klebsiella pneumoniae, Acinetobacter baumannii, Pseudomonas aeruginosa, and Enterobacter spp.) pathogens and was recently identified as an urgent public health threat by the Centers for Disease Control and Prevention (CDC), surpassing methicillin-resistant S. aureus (MRSA) and other members of the ESKAPE pathogens.^[3] Globally, 45% of all A. baumannii isolates are found to be MDR with rates as high as 70% in some regions.^[4]

Further complicating effective A. baumannii therapeutic strategies, the organism also has the ability to tolerate antibiotics in the context of the host. For example, it is hypothesized that otherwise clinically defined antibiotic susceptible strains of A. baumannii resist otherwise appropriate antibiotic treatment within the host. In other words, A. baumannii strains that are susceptible to antibiotics under non-biologically relevant laboratory conditions appear to develop resistance to physiologically achievable levels of the same antibiotics when the bacterium is in a host environment. Given that a commonality between the two most common manifestations of A. baumannii infections (pneumonia and septicemia) is the dissemination of the bacteria through the circulatory system to visceral organs, studies have focused on defining antibiotic tolerance mechanisms that are associated with growth in human serum. In that regard, we have previously reported that growth of A. baumannii in human serum results in the dramatic upregulation of genes associated with drug resistance, including 22 putative drug efflux pumps, and leads to efflux-mediated antibiotic tolerance to front-line treatment options.^[5] This phenomenon, termed adaptive efflux-mediated resistance, has also been shown to occur in P. aeruginosa and is associated with clinical antibiotic failure in that organism.

Considering the clinical importance of adaptive efflux-mediated resistance, mitigating the mechanisms that govern these processes are hypothesized to lead to improved clinical outcomes when used in conjunction with conventional antibiotics. Accordingly, we have previously characterized novel putative adaptive efflux-mediated resistance inhibitors Acinetobacter baumannii efflux-pump inhibitor 1 (ABEPI1) and ABEPI2 for potentiation of antibiotics in serum grown antibiotic-tolerant cells (Figure 1).^[6] The antibiotic minocycline, and its derivative tigecycline, are crucial drugs for the treatment of drug-resistant A. baumannii. When A. baumannii was cultured in serum in the presence of 0.5x MIC of minocycline, ABEPI1 potentiated minocycline activity and displayed a minimum effective concentration (MEC) of 2 μg/mL with at least a 2-log reduction in CFUs/mL while ABEPI2 displayed an MEC of 32 μg/mL. These molecules demonstrated broad spectrum potentiation of antibiotics in serum grown A. baumannii and P. aeruginosa.^[6]

Figure 1.

Structures of previously reported adaptive efflux inhibitors ABEPI1 (1) and ABEPI2.

Herein, hit optimization was carried out on the sulfonamide containing scaffold ABEPI1 that focused first on modification of the chemically unstable imine linker followed by exploring substitutions to the 4-Cl on the pendant phenyl. During this process, 29 new analogues of ABEPI1 were synthesized, and the corresponding structure-activity relationship studies are presented. We have characterized these molecules as serum-stable potentiators of multiple antibiotics including the glycylcycline tigecycline against A. baumannii cells displaying serum-associated antibiotic tolerance, while not having any stand-alone antimicrobial activity. Furthermore, these compounds reduced ethidium bromide efflux in serum grown cells and only potentiated antibiotics in serum, suggesting the putative target as being one related to adaptive efflux. In a twofold effort to evaluate mode of action as well as to evaluate potential toxicity, compounds were assessed for their ability to decouple bacterial cell membrane potential. Additionally, the analogues were evaluated for toxicity in mammalian cell culture. The data presented suggest these molecules represent promising candidates for use in combination with existing, as well as future antibiotics, for the treatment of A. baumannii infections.

Results and Discussion

Chemistry

All analogues of ABEPI1 were synthesized according to the protocol presented in Scheme 1. Analogues were synthesized beginning with the reactants 1a–1e. These starting materials varied at substituent X (phenol or aniline) and position (para or meta). Reactants 1a–1e were coupled through S_N2 chemistry at the phenol or aniline with various substituted benzyl bromides or chlorides (2) by using potassium carbonate as a base and dimethylformamide as the solvent at 80°C. The reactions were then quenched by pouring onto ice-water, if a white precipitate formed it was then filtered out to provide desired products. If the solution was milky white the aqueous phase was extracted with ethyl acetate then concentrated. These methods provide 3–31 in moderate yields for testing. Any molecules that did not meet a purity criteria of >95% purity by HPLC were purified by flash column chromatography.

Scheme 1.

Reagents and conditions. a) 1a–e (1.0 equiv), 2 (1.2 equiv), potassium carbonate (1.2 equiv), DMF, 80°C, overnight; yields 10–82%.

Structure–activity relationships for potentiation of minocycline

For the purpose of informing SAR and to evaluate the efficacy of the new analogues on reversing serum-associated tolerance, A. baumannii 98-37-09 was cultured in 100% human serum with 0.5x MIC of minocycline, a tetracycline derivative used to treat A. baumannii bacteremia, and increasing concentrations of compound. Antimicrobial effects were quantified by plating for colony forming units (CFU), as described in the Experimental Section. Given that our definition of MEC for this assay is the minimum concentration of analogue that results in at least 2- log reduction of bacteria in the presence of minocycline there are two aspects of this assay that informed the design of molecules: 1) the MEC measurement and 2) the log reduction of CFU/mL. As some molecules may display potent MECs they may not result in more than 2-log decrease of CFU/mL. Thus, we sought to design molecules that combined potent MEC values with greater reduction of CFU/mL. All MEC and CFU/mL reduction data for analogues are presented in Table 1, Figure 2, and Table S1 in the Supporting Information.

Table 1.

Minimum effective concentration (MEC) values of analogues for potentiation of minocycline activity against A. baumannii in human serum in presence of 0.5x MIC of minocycline.

Cpd	Y	Z	R	MEC[a]	Avg growth inhibition[b]	Cpd	Y	Z	R	MEC[a]	Avg growth inhibition[b]
APEBI1	NH2	4-N	4-Cl	2	2.3±4	17	NH2	4-O	4-Br	0.5	3±3
3	NH2	4-NH	4-Cl	0.5	4±2	18	NH2	4-O	3-Br	1	4±3
4	NH2	4-O	4-Cl	2[c]	6±0	19	NH2	4-O	2-Br	>8	1±5
5	OH	4-O	4-Cl	2	4±2	20	NH2	4-O	4-I	2	3±3
6	NH2	4-O	3-Cl	2	2±4	21	NH2	4-O	3-I	1	6.5±0
7	NH2	4-O	2-Cl	>8	1±5	22	NH2	4-O	4-CF3	4	6.5±0
8	NH2	3-O	4-Cl	2	6.5±0	23	NH2	4-O	3-CF3	>8	6.5±0
9	NH2	4-O	4-H	1	3±3	24	NH2	4-O	2-CF3	2	2±4
10	NH2	4-O	4-OMe	1	3±4	25	NH2	4-O	4-CN	0.5	2±4
11	NH2	4-O	3-OMe	8	6.5±0	26	NH2	4-O	3-CN	0.5	4±2
12	CH3	4-O	4-OMe	2	2±4	27	NH2	4-O	4-NO2	4	2±4
13	NH2	4-NH	4-OMe	2	4±2	28	NH2	4-O	3-NO2	4	2±6
14	NH2	4-O	4-F	1	7±0	29	NH2	4-O	4-C(CH3)3	2	4±4
15	NH2	4-O	3-F	2	3±3	30	NH2	4-O	2-CH3	4	3±3
16	NH2	4-O	2-F	2	7.3±0	31	NH2	4-O	Naphthyl[d]	2	4±2

^[a]MEC is defined at the minimum concentration of molecule required to result in at least a 2-log decrease of CFU/mL for A. baumannii in human serum. All values presented as μg/mL. All values have an error of ± 1-dilution.

^[b]The average growth inhibition represents the average log₁₀ CFU/mL decrease of four experiments ± SD where cells were treated with each analogue in combination with minocycline.

^[c]The MEC value for analogue 4 was an average of two synthetic lots.

^[d]This analogue has a naphthylene attached at the 2-position in place of the phenyl pendant group.

Figure 2.

Potentiation of minocycline activity against A. baumannii in human serum in terms of log-reduction (CFU/mL) for selected analogues. DMSO represents vehicle alone (gray bar) and ABEP1 is the parent compound that is known to potentiate minocycline. Molecules tested at 1x (black bars) and 2x (white bars) MEC. Values are an average of four experiments, performed in duplicate ± SD.

As described above, the original hit ABEPI1 was first discovered from a screening effort to identify novel chemical scaffolds that potentiate minocycline activity against A. baumannii during growth in human serum. ABEPI1 was one of a cluster of 13 hits within the same scaffold class from this screen and also one of the most potent in terms of both MEC. Reordered dry powder of the compound was tested and provided an MEC value of 2 μg/mL and resulted in a 2.3-log reduction of CFU/mL A. baumannii in culture. However, all members of the scaffold possessed a secondary imine linker, which under acidic aqueous conditions is susceptible to nucleophilic attack and subsequent hydrolysis leading to the corresponding amine and carbonyl; at neutral pH this process can still occur but does so at a reduced rate. Thus, optimization focused first on substitution of the imine linker by replacement with either an amine (3) or ether (4) linker. The linker was shown to be amenable to modification with the amine providing increased potency in terms of MEC values (0.5 μg/mL) with a modest increase in CFU reduction compared to ABEPI1. The ether linkage in 4 provided equipotent MEC to ABEPI1 but was more effective in reduction of A. baumannii CFU load, displaying a 6-log reduction in CFU load (Table 1). The necessity of the sulfonamide was probed by replacing it with a sulfonic acid substituent (5). This derivative displayed no change in MEC compared to 4 but did have reduced efficacy on average bacterial growth going from 6-log reduction for 4 to a 4-log reduction for 5. The sulfone derivative (12) displayed minimal efficacy in average growth inhibition (2-log reduction). Linker substitution in relation to the sulfonamide appears to have no effect on molecule efficacy as analogue 8, with the linker at the 3-position in relation to the sulfonamide, was as effective as para-substitution in 4. Thus, further SAR was explored while maintaining both the sulfonamide and the para-ether linker and focused on modifications to the pendant phenyl group.

The next set of analogues explored the effect of pendant phenyl modifications. Moving the chlorine from the 4-position (4) to 3-position (6) produced no change in MEC (2 μg/mL) but reduced the efficacy in growth inhibition (6-log vs 2-log reduction, respectively) while the 2-Cl (7) resulted in abrogation of the potentiating activity of the scaffold with >8 μg/mL MEC. Removing the chlorine to yield an unsubstituted phenyl derivative (9) also reduced the effect on bacterial load compared to 4. Chlorine was preferred over methoxy substituents as analogues 10 and 11 displayed diminished efficacy for one or both metrics.

Alternative halogens provided a boost in efficacy against A. baumannii. Fluorine substitution in place of the chlorine delivered the best combination of MEC potency with reduction of bacterial growth. Derivatives (14, 15, and 16) were essentially equipotent to the chlorine counterparts in the MEC assay (1–2 μg/mL). However, the 4-F (14) and 2-F (16) displayed significantly improved efficacy in CFU reduction compared to the chlorine counterparts with both displaying a 7-log reduction of CFU/mL of A. baumannii in the presence of minocycline. Interestingly, the 3-F analogue (15) was less effective in reducing bacterial load compared to the ortho and para derivatives. Bromine modifications (17, 18, and 19) followed the same trend as the chlorine set as the 4-Br (17, MEC=0.5 μg/mL) was slightly more potent than 3-Br (18, 1 μg/mL) while the 2-Br (19) lost activity completely. The para- and meta-Br analogues displayed moderate reductions of CFUs while the ortho derivative displayed no activity at the concentrations tested. Finally, iodine substitution was essentially equipotent to the other halogens at the para (20) and meta (21) positions with respect to MEC. However, the substitution of iodine displayed moderate effects on CFUs with 3- to 4-log reduction. The ability of the halogen series to reduce bacterial load generally correlated with size, especially at the para position, as F>Cl> Br=I. The 4-F and 2-F derivatives were the best of the halogen set with similar MEC values of 1–2 μg/mL and robust efficacy in potentiation of minocycline by reducing CFU/mL by 7-log units.

To round out the series numerous other substituents were explored including trifluoromethyl, cyano, nitro, methyl, tert-butyl and naphthyl analogues. These remaining analogues provided a range of MECs from 0.5–4 μg/L. The 4-CF₃ analogue (22) was most effective in bacterial load reduction (6.5-log decrease), however, none of the remaining exhibited greater than 4-log reduction of CFU/mL.

Antimicrobial activity of analogues in human serum

The original hit ABEPI1 was shown to potentiate minocycline activity without exhibiting antimicrobial activity against A. baumannii. To ensure the analogues from this series were solely involved in potentiation of the antimicrobial activity of minocycline rather than contributing antimicrobial effects of their own we tested all analogues in in vitro antimicrobial assays against A. baumannii in human serum. Almost all compounds tested lacked any detectible antimicrobial activity of their own when tested at 1x and 2x MEC. The lone exception being the sulfone containing analogue 12 which began to display reduced CFU/mL at both 1x and 2x MEC concentrations (Table S1), reducing A. baumannii CFU/mL by 1-log unit at a concentration of 2 μg/mL. At the same concentration of 12 in the presence of minocycline there was only a 2-log decrease of CFU/mL (Figure 2 and Table S1). Moreover, 12 displayed more potent antimicrobial activity by itself at 2x MEC (5-log reduction on CFU/mL at 2x MEC; 4 μg/mL) than at the same concentration in combination with minocycline. Thus, the reduction of A. baumannii may be attributed to the weak antimicrobial activity of 12. This data would indicate converting the sulfonamide to a sulfone does appear to alter the potentiating activity of the scaffold. However, this observation is limited to the sulfone as converting the sulfonamide to sulfonic acid (analogue 5) did not exhibit any antimicrobial activity in the absence of minocycline at any concentration (Table S1).

Combination treatment of prioritized analogues with minocycline in serum

Analogues 10, 14, 16 and 17 were prioritized for dose response testing to further distinguish the antimicrobial potentiating effect of minocycline due to their low MEC values and ability to potentiate minocycline. To do so, A. baumannii was cultured in lysogeny broth (LB) or 100% human serum with increasing concentrations of minocycline (Figure 3). Putative efflux-pump inhibitors were dosed at their respective MECs and antimicrobial effects were evaluated by plating for CFUs. PAβN was added at 50 μg/mL as a positive control for efflux-pump inhibition. As expected, bacterial cell growth was completely inhibited in LB at 1 μg/mL minocycline, whereas at an identical minocycline concentration in serum, cells grew to an approximate 10⁵ CFU/mL, thereby demonstrating the serum-associated antibiotic tolerance phenotype. The addition of PAβN reduced the MIC in serum to 1 μg/mL, suggesting that serum-induced tolerance is efflux pump mediated. Putative EPI analogues 10, 14, and 17 all reduced the MIC of minocycline to 1 μg/mL in serum, at concentrations up to 100-fold lower than PAβN. Analogue 16 did not reduce the MIC, but did result in a 3-log decrease in CFUs at 2 μg/mL. Furthermore, the addition of these molecules (including PAβN) to cells grown in LB had no effect on the MIC, supporting the concept of serum-induced efflux (Figure S1). Taken together, this data indicates that these compounds are able to reverse serum-induced antibiotic tolerance, presumably through the inhibition of efflux pumps.

Figure 3.

Putative EPI analogues reverse serum-associated tolerance to minocycline. A. baumannii 98-37-09 was grown in LB (blue) or 100% human serum (red) in the presence of increasing concentrations of minocycline with and without efflux-pump inhibitors. PAβN (green) was supplemented to 50 μg/mL, EPIs 10 (pink) and 14 (orange) were supplemented to 1 μg/mL, EPI 16 (black) was supplemented to 2 μg/mL and EPI 17 (brown) was supplemented to 0.5 μg/mL. Bacterial growth was enumerated by plating for CFUs after 48 h at 37°C. Values are an average of two experiments performed in duplicate ± SD.

Ethidium bromide accumulation

To determine whether the ability of each compound to potentiate minocycline correlates with its ability to inhibit serum-induced efflux properties, standard ethidium bromide efflux assays were performed in the presence and absence of each putative efflux inhibitor, as described.^[6,7] The assay is based on measuring the intrinsic fluorescent properties of ethidium bromide upon intercalation into cellular nucleic acids; efflux active cells display relatively lower fluorescence due to ethidium bromide being expelled and thus less intercalation. Conversely, inhibition of efflux consequently results in more ethidium bromide intracellular accumulation and higher fluorescence.

The original molecule ABEPI1 was shown to inhibit serum-induced efflux activity in A. baumannii at 2 μg/mL.^[6] To directly compare the new analogues to the parent molecule, A. baumannii strain 98-37-09 cells were treated with 2 μg/mL of each putative efflux-pump inhibitor (Figure 4 and Table S2). Figure 4 provides representative ethidium bromide fluorescence assay results in the presence and absence of prioritized compounds 10, 14, 16, and 17. Cells grown in serum (Figure 4, DMSO) express higher levels of efflux pumps and, in turn, exhibit lower fluorescence due to enhanced removal of ethidium bromide from the cell. The fluorescence from cells grown in serum was normalized to 100% and test samples were compared in relation to this value. Cells grown in LB had reduced efflux pump expression and, therefore, 18% greater fluorescence compared to the serum control sample. The well characterized efflux-pump inhibitor, PAβN, was used as a positive control and had the most obvious effect on efflux resulting in approximately 17% more accumulation than the DMSO control (Figure 4 and Table S2). However, the concentration of PAβN required to generate this signal (40 μg/mL) was 20 times higher than for the other compounds. ABEPI1 at a dose of 2 μg/mL (Figure 4, blue line) resulted in approximately 5.5% more fluorescence due to ethidium bromide accumulation than DMSO treated cells (Figure 4 and Table S2). Prioritized analogues all increased ethidium bromide accumulation ranging from 5–8% more than the DMSO control when dosed at 2 μg/mL, presumably by inhibition of putative efflux pumps (Figure 4). Table S2 summarizes the effects of all analogues on ethidium bromide accumulation of A. baumannii in serum. Of all the analogues, 13 had the strongest effect on efflux inhibition, resulting in approximately 11% increased ethidium bromide accumulation. Interestingly, this analogue only produced a 4-log decrease in CFU/mL in the presence of A. baumannii.

Figure 4.

Putative efflux-pump inhibitors reduce ethidium bromide efflux in serum-grown A. baumannii. Cells were grown in serum and putative efflux inhibitors ABEPI1 (cyan), analogue 10 (pink), analogue 14 (orange), analogue 16 (black) and analogue 17 (brown) were added to 2 μg/mL. DMSO (red) represents vehicle alone and full ethidium bromide efflux potential. Known efflux inhibitor PAβN (green) was added to 40 μg/mL. Cells grown in LB (blue) show little to no ethidium bromide efflux. Values are an average of three experiments performed in duplicate ± SD.

Membrane depolarization assay

Multiple modes of action have been reported for known efflux-pump inhibitors, including the downregulation of efflux pump genes, inhibition of functional efflux pump assembly, direct blockage of the pump to prevent substrate binding, and collapse of the energy mechanism required for pump function.^[8] While membrane active agents represent powerful antimicrobials, discovery efforts have historically avoided molecules that target the membrane’s energetics due to their propensity to cause toxicity in mammals.^[9] Accordingly, as a twofold effort to begin to evaluate mechanism of action of these molecules, as well as assess potential toxicity issues, membrane depolarization assays were performed.

Membrane potential experiments were conducted using the voltage-sensitive dye, 3,3’-dipropylthiadicarbocyanine iodide DiSC₃(5).^[10] The combination of the cationic and sufficiently hydrophobic properties of DiSC₃(5) allow it to act as a potentiometric probe and accumulate in polarized cells, where-by accumulation under normal conditions results in quenching of the dye.^[10] However, changes in proton motive force (PMF) cause DiSC₃(5) to either be expelled from the cell and fluoresce, or accumulate more so than normal and self-quench. In bacteria, the PMF is made up of two components: the transmembrane proton gradient (ΔpH) and the electric potential (ΔΨ). Bacteria exercise a finely tuned control over each of these components in order to maintain a constant PMF, such that if one component dissipates, the other one can be adjusted to counteract and compensate the other.^[11] Thus, we used agents that are known effectors of each component to be employed as controls; CCCP is a well-studied efflux inhibitor and is known to dissipate ΔpH and valinomycin is a known disrupter of ΔΨ. As shown in Figure 5, increasing CCCP concentrations resulted in a decrease in overall fluorescence (plotted as PMF) due to the compensatory effect of ΔΨ as a result of ΔpH dissipation; DiSC₃(5) accumulation in the cell is largely driven by ΔΨ. Conversely, increasing the concentration of valinomycin resulted in higher PMF (higher fluorescence) due to the dissipation of ΔΨ and DiSC₃(5) with the most noticeable effect at 128 μg/mL (data point not shown in plot, additional plot showing this extended dose – response with valinomycin in Figure S2).

Figure 5.

Putative EPIs do not modulate PMF. Valinomycin (blue) and CCCP (red) were used as controls to disrupt ΔΨ and ΔpH, respectively. DMSO (0 μg/mL) represents cells treated with vehicle only and was normalized to 100% PMF. ABEPI1 (cyan) and analogues were analyzed at concentrations up to 32 μg/mL. Values were plotted relative to 100% PMF. Values are an average of two experiments performed in duplicate. Dotted line indicates 20% change in PMF.

Putative efflux inhibitors were analyzed for their ability to potentiate bacterial cell membranes at concentrations up to 32 μg/mL, which was approximately 16X the MEC value. Values from DMSO only (0 μg/mL) were normalized to be 100% PMF and all other fluorescence values were plotted relative to this. Compounds were considered membrane potentiators if they resulted in a 20% change in PMF, which was based upon the control data and previous studies.^[9,10] While ABEPI1 dissipated the PMF (specifically ΔΨ) at high concentrations, analogues 10, 14, 16 and 17 did not (Figure 5), thus suggesting that disruption of membrane energetics is not the mechanism by which they potentiate minocycline. Tabular data for the remaining analogues show that thirteen of the compounds exceed the 20% threshold (Table S3). However, the majority of these only affected membrane potential at concentrations of 4x the MEC or greater.

Cytotoxicity versus human cell lines

For the series to have any translational impact the molecules must not exhibit cellular toxicity against human cell lines. For an initial assessment of toxicity, we tested all analogues for toxicity against human liver epithelial (HEP G2) cells in a 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT) assay, following the manufacturer’s guidelines (see the Experimental Section). Figure 6 represents cell viability data from analogues 10, 14, 16 and 17. Molecules were assayed at 2, 4, and 8 μg/mL. Mitomycin C (125 μg/mL) was used a positive control and resulted in 34.1% cell viability (Table S4). 70% human cell viability was used as the cut-off criterion according to ISO guidelines, as described in the Experimental Section (Figure 6, dashed line). None of the top candidate analogues demonstrated toxic effects to mammalian cells at any of the concentrations examined (Figure 6). In general, the scaffold displayed relatively low toxicity (>80% survival compared to DMSO treated control) against the HEP G2 cells (Figure 6 and Table S4). Exceptions include the iodine containing analogues 20 and 21, the 3-CF₃ analogue 23, and the naphthyl containing 31. Fortunately, none of these analogues were noteworthy with respect to potentiation of minocycline in human serum and they were further deprioritized due to the potential for human cell toxicity. Although these results are promising, additional evaluation of the scaffold in more exhaustive toxicity assays will help to further de-risk the series.

Figure 6.

Effects of putative efflux-pump inhibitors on cell viability. ABEPI1 (cyan), and putative EPIs 10 (pink), 14 (orange), 16 (black) and 17 (brown) were added to HEPG2 cells in increasing concentrations. The dashed line represents the lower limit of cell viability, below which compounds are considered to have toxic effects. Values are an average from one experiment, performed in triplicate ± SD.

Combination treatment of putative EPIs with tigecycline in serum

Tigecycline was the first of a new class of modified tetracyclines known as glycylcyclines, which include the structural addition of a 9-tert-butyl-glycylamido side chain to the minocycline backbone.^[12] The antibiotic overcomes the two major determinants of tetracycline resistance: ribosomal protection and active efflux.^[13] Thus, tigecycline possesses expanded activity against multi-drug resistant Gram-positive and Gram-negative organisms. While tigecycline was once considered a last-resort antibiotic for the treatment of A. baumannii infections, rising resistance rates via upregulation of efflux pumps has limited its use.^[14] Antibiotic susceptibility assays confirmed that growth in human serum resulted in tolerance to tigecycline (Figure 7, LB (blue) compared to serum (red)). As putative analogues restored susceptibility to minocycline, we evaluated their ability to sensitize serum-grown A. baumannii to tigecycline at their respective MECs (Figure 7). PAβN, at concentrations up to 100- fold higher than the analogues (50 μg/mL), only modestly potentiated tigecycline at 2 μg/mL. Compounds 10, 16 and 17 resulted in a 6-log decrease in bacterial growth at 1 μg/mL tigecycline, whereas analogue 14 resulted in an approximate 2- log decrease at the same concentration. At 0.5 μg/mL tigecycline, all four analogues reduced bacterial growth by approximately 5-logs. Analogues were also evaluated in LB-grown cells in the presence of tigecycline and had no effect, suggesting that these analogues inhibit processes that occur through adaptive efflux (Figure S3).

Figure 7.

Putative EPI analogues reverse serum-associated tolerance to tigecycline. A. baumannii 98-37-09 was grown in LB (blue) or 100% human serum (red) in the presence of increasing concentrations of tigecycline with and without efflux-pump inhibitors. PAβN (green) was supplemented to 50 μg/mL, EPIs 10 (pink) and 14 (orange) were supplemented to 1 μg/mL, EPI 16 (black) was supplemented to 2 μg/mL and EPI 17 (brown) was supplemented to 0.5 μg/mL. Wells were plated for CFUs after 48 h at 37°C. Values are an average of two experiments performed in duplicate ± SD.

Discussion

A. baumannii has emerged as a virulent pathogen in regard to its ability to colonize and thrive in its environment, in part due to its extensive repertoire of resistance factors and antibiotic resistance determinants.^[1,15] A. baumannii harbors representatives from each of the five bacterial drug efflux families, which include the major facilitator superfamily (MFS) pumps, the multidrug and toxic compound extrusion (MATE) transporters, the small multidrug resistance (SMR) family transporters, the resistance nodulation division (RND) family pumps, and the ATP binding cassette (ABC) transporters. Collectively, these efflux pumps aid in resistance to several antibiotic classes including fluoroquinolones, carbapenems, aminoglycosides and tetracyclines.^[15b,16] Additionally, the organism harbors many putative uncharacterized efflux pumps; novel efflux pumps are still being identified, while others have yet to be identified or characterized as clinically relevant,^[17] and their regulatory networks are poorly understood.

Although in vitro antimicrobial antibiograms provide an invaluable tool for guiding clinicians with putatively effective antibiotics toward an infecting bacterial species and strain, it is becoming increasingly apparent that conventional in vitro antimicrobial testing conditions do not recapitulate or directly predict the performance of antibiotics in the context of certain host settings.^[18] Indeed, clinically defined antimicrobial susceptible A. baumannii strains frequently do not respond to in vitro defined antibiotic therapies due to bacterial tolerance mechanisms that allow the invading pathogen to transiently express mechanisms to adapt, and hence overcome, physiologically achievable antibiotic concentrations.^[18d,19] In that regard, studies have revealed that the organism is capable of expressing antibiotic efflux systems in response to growth in physiological salt conditions and human serum resulting in efflux-associated tolerance to antibiotics.^{[5,6, 20}] This phenomenon has been termed adaptive efflux-mediated resistance and has been hypothesized to, in part, contribute to the therapeutic failure of antibiotics toward otherwise clinically defined susceptible A. baumannii strains.^[21] Given the biological importance of adaptive efflux-mediated antibiotic tolerance in the context of A. baumannii drug resistance, small molecule inhibitors of serum-associated efflux may represent promising therapeutics to be used in combination with both new and existing antibiotic agents. Here we describe molecules that potentiate the antibacterial activity of minocycline in serum-grown A. baumannii.

More directly, A. baumannii is known to strongly upregulate at least 22 known or putative drug efflux proteins and display adaptive efflux-mediated resistance to a multitude of antibiotics, including minocycline, when cultured in human serum.^[5,6] A high throughput screen for A. baumannii efflux-pump inhibitors (ABEPI) has previously identified the sulfonamide class of molecules reported in this article, ABEPI1, that has been shown to restore minocycline activity against A. baumannii in serum by presumably inhibiting the organism’s adaptive drug efflux properties. The goals of the current work were to expand our characterization of the structure activity relationship of ABEPI1 and identify more potent chemical series analogues for downstream development.

SAR optimization focused on four regions for modification (Figure 8). As illustrated by one of the best-performing analogues in 14 the main takeaways from the series are that the sulfonamide containing an ether linkage was the most effective at reducing A. baumannii CFU load among the nearest-neighbor analogues. Analogues that merged these attributes with halogens on the pendant phenyl, specifically in the 4-position, provided the best combination of both low MEC value and significant reduction of bacterial load of A. baumannii when dosed alongside minocycline in human serum. Additionally, with the exception of 12, none of the analogues tested exhibited antimicrobial activity against A. baumannii on their own in either LB or human serum establishing their ability to potentiate antibiotics.

Figure 8.

Observed SAR trends for sulfonamide putative efflux-pump inhibitors.

The potentiation of antibiotics can be accomplished by two predominant means, efflux inhibition or membrane disruption, the latter of which is less therapeutically desirable due to potential toxic effects on host cells. Accordingly, each of the analogues developed here were directly evaluated in conventional efflux inhibition and membrane depolarization assays using ethidium bromide and DiSC₃(5) assays, respectively. As expected, and similar to the parent ABEPI1, the derivatives evaluated here largely displayed a decrease in efflux potential, which tracked well with their minimum effective concentration in combination with the model antibiotic minocycline. Admittedly, the analogues displayed a modest accumulation of ethidium bromide, in the range of 4–11% when tested at 2 μg/mL, whereas the known efflux-pump inhibitor PAβN provided greater ethidium bromide accumulation up to 17% when tested at 40 μg/mL. When added at comparable levels, PAβN had similar effects on ethidium bromide accumulation, however the working range of PAβN where potentiation of minocycline observed is between 25–50 μg/mL. The inhibition of ethidium bromide efflux, ostensibly, might be low compared to the potentiating activity of analogues 10, 14, 16, and 17 due to a difference in efflux pumps that remove ethidium bromide versus minocycline; that is, the sulfonamide-containing analogues may be more efficacious at putative efflux pumps that are minocycline-specific as opposed to the pumps used for ethidium bromide extrusion. Conversely, very few compounds demonstrated a membrane depolarization phenotype, suggesting that they are unlikely to be modulating antibiotic potentiation by merely disrupting A. baumannii membrane energetics. In keeping with this, the compounds also displayed very little or relatively minor host cell cytotoxicity up to 4x MEC, suggesting the series is worthwhile advancing for therapeutic interrogation. However, we do not know the mechanism by which the series seemingly potentiates antibiotics via efflux inhibition and studies are currently underway to distinguish whether they are direct inhibitors of A. baumannii efflux biochemical activity, inhibitors of efflux pump expression, or functioning in an unrecognized manner.

In addition to potentiation of minocycline in serum, the compounds analyzed were also able to reverse serum-associated tolerance to tigecycline (Figure 7), a front-line antibiotic once used to treat MDR A. baumannii infections that has had increasing resistance mounted against it by A. baumannii, due to the overexpression of the RND efflux system AdeB.^[22] Potentiation of tigecycline was observed in serum-grown cells only, and not in cells grown in standard laboratory media (Figure S3). These observations further support the notion that serum-associated tolerance is likely due, at least in part, to adaptive efflux and that the analogues characterized here seem to specifically target these processes. Whether these mechanisms play a part in clinical antibiotic failure and further, whether or not these analogues will have efficacy in vivo, remains to be examined. Studies are currently underway to define the additional classes of antibiotics that may be potentiated by the putative efflux inhibitors described here and their effects on an expansive panel of contemporary clinical isolates.

In summary, we report the results of our study to improve the scaffold of the originally reported ABEPI1. The chemical stability of the scaffold was improved and made more drug-like during the optimization campaign with the removal of the imine linker. The series also possesses good physicochemical properties for translation; for example, compound 14 has a molecular weight of 281.0 g/mol and a log P of 2.54. When analyzed in the cheminformatics program FAF-Drugs3 there are no red flags for compound liability or toxicity.^[23] The lack of toxicity was confirmed against HEP G2 cells up to 4x MEC, and furthermore as compounds that do not depolarize bacterial cell membranes, which can be an indicator of toxicity. Although these molecules represent a good start on the way to a viable lead, we believe that we can further improve MEC potency and bacterial reduction. The SAR to date is still somewhat limited, thus future studies will focus on expanding the SAR on the pendant phenyl. We will also investigate the effects of additional modification to the sulfonamide containing phenyl ring, including sulfonamide isosteres. Finally, linker alternatives will also be explored by varying size and flexibility. Nonetheless, the sulfonamide scaffold presented represents a promising new agent in the fight against drug resistance.

Conclusion

A. baumannii is known to dramatically upregulate known or putative drug efflux proteins and display adaptive efflux-mediated resistance to a multitude of antibiotics, including minocycline, when cultured in human serum.^[5,6] ABEPI1 was shown to restore minocycline activity against A. baumannii in serum by presumably inhibiting the organism’s adaptive drug efflux properties. SAR optimization has led to the development of a more potent chemical series for potentiation of minocycline in human serum, which combine low MEC with the ability to significantly reduce bacterial burden. These compounds were also found to reduce efflux potential by ethidium bromide accumulation assays in a similar manner to the hit. While the exact mechanism for efflux perturbation is still unknown, we have ruled out disruption of PMF. Additional studies are required to determine whether these compounds are acting via direct blockage of the pump, by interfering with genetic expression of the pumps, or by an alternate mechanism. Moreover, the prioritized analogues were able to potentiate tigecycline, a frontline antibiotic that has seen dwindling use in the treatment of A. baumannii infections due to rising resistance. Collectively, these molecules may represent a promising collection of adjuvants to be used in combination with existing antibiotics for the treatment of A. baumannii infections.

Experimental Section

Chemistry

General:

¹H and ¹³C NMR spectra were recorded on Bruker DRX500 spectrometer (operating at 500 and 126 MHz) in DMSO-d₆ with or without the internal standard of TMS at 0.05% v/v. The chemical shifts (δ) are reported as parts per million (ppm). The purity of all final compounds was >95% purity as assessed by HPLC according to current American Chemical Society guidelines for publication. Final compounds were analyzed on an Agilent 1200 series chromatograph. The chromatographic method utilized as Thermo Scientific Hypersil GOLD C-18 or silica column. UV detection wavelength=220/254 nm; flow-rate=1.0 mL/min; solvent=acetonitrile/water for reverse phase and ethyl acetate/hexane for normal phase. Both organic and aqueous mobile phases contain 0.1% v/v formic acid. The mass spectrometer used is an Advion CMS-L Compact Mass Spectrometer with an APCI source. Samples are submitted for analysis using the atmospheric solids analysis probe (ASAP). Compounds were prepared according to scheme 1 and protocol is detailed below for compound 4. All other compound synthesis and characterization data is provided in the supporting information.

General procedure for synthesis of analogues

4-((4-Chlorobenzyl)oxy)benzenesulfonamide (4). To a vial was added the 1a (0.20 g, 1.2 mmol, 1.0 equiv) and DMF (3 mL) followed by addition of potassium carbonate (0.098 g, 1.4 mmol, 1.2 equiv) and 1-(bromomethyl)-4-chlorobenzene (0.29 g, 1.4 mmol, 1.2 equiv) at room temperature. The reaction stirred at 80°C and was monitored by TLC. When the reaction was complete is was poured into ice water and a white precipitate was formed and filtered under vacuum filtration. The white solid was purified by flash chromatography (0–60% ethyl acetate/hexanes) to afford the final product 4 (0.092 g, 0.53 mmol, 46%). ¹H NMR (500 MHz, [D₆]DMSO): δ=7.75 (dd, J=8.6, 1.6 Hz, 2H), 7.51–7.46 (m, 4H), 7.23 (s, 2H), 7.15 (dd, J=8.6, 1.6 Hz, 2H), 5.19 (s, 2H). ¹³C NMR (125 MHz, [D₆]DMSO): δ=160.5, 136.5, 135.5, 132.6, 129.6, 128.6, 127.7, 114.9, 68.7. APCI-MS: m/z (%) 296.0 (90) [M H] . HPLC retention time: 12.105 min. HPLC purity=100%.

Bacterial strains and growth conditions:

A. baumannii 98–37–09 is a clinical isolate from the Center of Disease Control and Prevention and has been previously described.^[5] A. baumannii was grown in LB or 100% human serum (Corning Life Sciences) at 37°C. Where indicated, LB or serum was supplemented with the indicated concentrations of minocycline (TCI Chemicals, Portland, OR) or tigecycline (BioTang Inc., Lexington, MA).

Antibiotic susceptibility assays:

A. baumannii 98–37–09 was grown overnight in LB at 37°C, diluted into fresh medium (1:100) and grown to optical density at 600 nm (OD₆₀₀) 0.4–0.5 at 37°C with aeration. A total of 1×10⁵ colony forming units (CFU) were transferred to each well of a 96-well round bottom plate containing 100 μL of LB or 100% human serum supplemented with increasing concentrations of the indicated antibiotic and incubated at 37°C for 48 h. To quantify the antimicrobial effects of each antibiotic, each well was serially diluted in 0.8% sodium chloride (NaCl) and plated on LB agar to enumerate CFU/mL. To determine the minimum effective concentration (MEC) of each compound, antibiotic susceptibility assays were conducted in LB or 100% human serum with 0.5x minimum inhibitory concentration (MIC) of minocycline (MIC is equal to 1 μg/mL in LB and 2 μg/mL in human serum) and increasing concentrations of compound.

Ethidium bromide efflux assay:

Bacterial ethidium bromide efflux assays were performed as described.^{[6,7, 16b, c]} An overnight culture of A. baumannii 98–37–09 was grown in LB at 37°C with aeration. The culture was diluted 1:100 in fresh LB or 100% human serum and grown to mid-exponential phase (OD₆₀₀ 0.4–0.5). The cell cultures were centrifuged for 20 min at 900g and 4°C, and pellets were washed three times with 20 mM sodium phosphate buffer. Pellets were resuspended to OD₆₀₀ of 0.2 in 20 mM sodium phosphate buffer. 1×10⁶ CFU were transferred to each well of a 96-well white plate (Falcon, Corning Life Sciences). 10 μg/mL ethidium bromide was added to each well and ethidium bromide fluorescence was monitored (λ_ex=530 nm; λ_em=600 nm) every 5 min for 60 min on a SPECTRAmax5 fluorimeter (Molecular Devices). To determine the effect on ethidium bromide efflux, the cells were treated with 2 μg/mL compound or 40 μg/mL phenylalanine arginine β-naphthylamide (PAβN; MP Biomedicals, Irvine, CA). To calculate percent inhibition of ethidium bromide accumulation, cells grown in serum alone were considered to represent 0% ethidium bromide accumulation. Accordingly, fluorescence values in the presence of compound were compared in terms of ethidium bromide accumulation and accumulation percentages for each compound were generated.

Membrane polarization assays:

To determine whether putative efflux-pump inhibitors depolarize bacterial cell membranes, membrane polarization assays were performed. An overnight culture of A. baumannii 98–37–09 was grown in LB at 37°C with aeration. The culture was diluted 1:100 in fresh LB and cells were grown to mid-exponential phase (OD₆₀₀=0.4). Cells were centrifuged for 20 min at 900g and 4°C, and cell pellets were washed three times with 5 mM HEPES (pH 7.3), 5 mM glucose. Cell pellets were resuspended in 5 mM HEPES (pH 7.3), 5 mM glucose and 100 mM potassium chloride (KCl) to OD₆₀₀ 0.2. 3,3’-Dipropylthiadicarbocyanine iodide (DiSC₃(5); 2 μM; Thermofisher) was added to the cells, which were then incubated for 15 min in the dark at room temperature. Approximately 1×10⁶ cells were added to each well of a 96-well black plate (Falcon, Corning Life Sciences). Carbonyl cyanide 3-chlorophenylhydrazone (CCCP; Sigma–Aldrich) and valinomycin (Millipore-Sigma) were used as positive controls to disrupt the pH and electrochemical gradients of bacterial membrane potential, respectively. Putative efflux inhibitors were added in increasing concentrations from 0.25 μg/mL to 32 μg/mL. PAβN was used for comparison from 1 μg/mL to 128 μg/mL. Fluorescence (λ_ex=622 nm; λ_em=670 nm) was measured immediately using a SPECTRAmax5 fluorimeter. Corresponding fluorescence values from cells that were untreated were set to 100% proton motive force (PMF). Fluorescence values corresponding to cells treated with compound were compared to the untreated cells and percent PMF values were generated. Compounds were considered to have effects on membrane energetics if they resulted in a 20% change in percent PMF. We chose 20% as a cutoff based on previous studies.^[9,10]

Cytotoxicity assays.

Human liver epithelial cells (HEP G2) were cultured in Dulbecco’s modified Eagle medium (DMEM; Fisher Scientific) supplemented with 10% heat inactivated fetal bovine serum (FBS; Corning Life Sciences) and 1% penicillin/streptomycin (Thermofisher Scientific). Cells were incubated at 37°C with 5% CO₂ in Nunc (Roskilde, Denmark) tissue culture flasks. Cells were grown in monolayers until exceeding 70% confluency, removed with 0.25% trypsin (Fisher Scientific), resuspended in fresh medium and used to seed approximately 2.5×10⁵ cells/mL into each well of a 96 well tissue culture microplate (Nunc) containing 200 μL of fresh medium. Cytotoxicity testing was conducted according to the guidelines of the International Organization for Standardization (ISO) 10993–5:2009. After 24 h of incubation, the cell medium was removed, and adherent cells were washed with 1x phosphate buffered saline (PBS). Cell media was supplemented with 5% per volume of compound at final concentrations between 2 μg/mL to 8 μg/mL and incubated 20–24 h at 37°C with 5% CO₂. After incubation, 20 μL (3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT) reagent (MTT cell proliferation kit; ATCC, Manassas, VA) was added to each well. Cells were incubated at 37°C with 5% CO₂ for 2–4 h before adding 100 μL detergent reagent (MTT cell proliferation kit; ATCC) and incubating for an additional 2 h at 37°C, in the dark. Using a SPECTRAmax5 microplate reader, absorbance was read at 570 nm. Cells were treated with 125 μg/mL mitomycin C (Fisher Scientific) to serve as a positive control and 1% DMSO as a negative control. Following ISO guidelines, compounds that resulted in cell viability below 70% were considered cytotoxic. All compounds were tested in triplicate and cell viability was expressed as a percent of the negative control.