Modifications to a Biphenolic Anti-Bacterial Compound: Activity Against ESKAPE Pathogens

Abstract

Forty-four analogs of honokiol, a compound with known antibacterial activity, especially with respect to oral bacteria, were synthesized to explore the structure-activity relationships against the ESKAPE pathogens. Compounds with high therapeutic indices (hemolysis₂₀/MIC) were identified. In particular, ester-linked compounds were found to be active that would be less than environmentally durable than biaryl ether antibacterials such as the broadly used triclosan. MRSA mutants could be generated against some, but not all, of the highly active compounds. Based on gene sequencing results, membrane permeability, intracellular sodium, and intracellular pH assays revealed overlapping mechanisms of action.

Graphical Abstract

A potent biphenolic inhibitor of oral bacteria has been modified for therapeutic use against ESKAPE pathogens. Mechanistic study via fluorescence-based assays and the generation of resistant bacterial strains revealed multiple, overlapping mechanisms for compounds that are structurally similar.

Introduction

Previous mismanagement and overuse of antibiotics has resulted in increased levels of antibacterial resistance. The World Health Organization has called for the development of new classes of antibiotics to treat multidrug-resistant infections caused by bacteria with growing resistance to current antibiotics, known collectively as “ESKAPE” pathogens. Of these, Enterococcus faecium, Staphylococcus aureus, and Enterobacter spp. are gram positive while Klebsiella pneumoniae, Acinetobacter baumannii, and Pseudomonas aeruginosa are gram negative.^[1]

Honokiol, a bisphenolic natural product isolated from the bark of magnolia trees, has been used for centuries in traditional eastern medicine as an anti-inflammatory, antidepressant, antiviral, and antibacterial agent.^[2–4] Previous studies of honokiol analogs revealed that biphenolic compound 4C was more effective against oral bacteria than honokiol itself.^[5] An additional SAR study revealed that two phenols bearing tert-butyl groups and linked by and ethylene unit provide the best activity against oral bacteria.^[6] While membrane permeability and cell lysis were similar between 4C and the known antibacterial cetylpyridinium chloride, membrane depolarization outcomes differed. Further, 4C possessed lower toxicity as judged by hemolysis and a higher therapeutic index. Incorporation of primary amines retained the superior therapeutic index and high activity against MRSA while providing modest activity against select gram-negative bacteria.^[7–9] In this report, the effects of novel amide, ester, and amine linkers are examined along with phenol alkylation and different positional variants to assess the activity against both gram-positive and gram-negative bacteria in the ESKAPE pathogen series.

Results and Discussion

Synthesis of Analogs.

Synthesis of analogs was accomplished according to Scheme 1. The zero-carbon linker analog 5A and was generated from biphenol 12 via Friedel-Crafts ortho-alkylation.^[10] Two-carbon analog 1B was generated by palladium-catalyzed hydrogenation of alkene 13.^[11] Compound 5B was synthesized via Friedel-Crafts ortho-alkylation of biphenol 14. Three-carbon linker compounds were generated via aldol reaction of 15 with 16A or 16B.^[12] Triethylsilane reduction of the enone in the aldol adduct to the alkane was followed by phenol deprotection to form 17A and 17B. The three-carbon analogs 2, 3B, 5C, and 6D were obtained by Friedel-Crafts ortho-alkylation of the corresponding biphenols 17A, 17B, 18A, and 18B. The four-carbon linker analog was obtained from Grignard reaction of 19 with 20.^[13] Hydrogenolysis of the resultant alcohol followed by phenol deprotection provided 21. Subsequent Friedel-Crafts ortho-alkylation afforded 5D.

Scheme 1.

Synthesis of New Honokiol Analogs.^[a]

^[a](a) t-BuOH, H₂SO₄, CH₂Cl₂, rt, 18 h, 5–24%; (b) Zn, TiCl₄, THF, 70 °C, 6 h, 95%; (c) H₂, Pd/C, EtOH, rt, 18 h, 7–100%; (d) NaOH, EtOH, rt, 2 h, 52–100%; (e) TFA, Et₃SiH, rt, 5 h, 61–99%; (f) BBr₃, CH₂Cl₂, rt, 18 h, 18–85%; (g) (3-methoxyphenyl)magnesium bromide, THF, rt, 3 h, 68%; (h) DMAP, DCC, CH₂Cl₂, rt, 18 h, 8–90%; (i) EDAC, acetone, 70 °C, 18 h, 7–23%; (j) LiAlH₄, THF, rt, 18 h, 59–81%; (k) PBr₃, DCM, rt, 3 h, 70%; (l) 23A, K₂CO₃, DMF, rt, 18 h, 48%; (m) RBr, K₂CO₃, acetone, 60 °C, 18 h, 15−41%; (n) TFA, H₂O, 3–8%; (o) PhMgBr, THF, rt, 18 h, 59%.

Ester-linked analogs 7A-F were generated by Steglich esterification of a carboxylic acid (22A-C) and phenol (23A-B). Amide-linked analogs 8A and 8B were synthesized via DCC coupling of 22A or 22C with aniline 24. For the ether-linked analogs, LiAlH₄ reduction of carboxylic acids 22A and 22C provided the corresponding benzylic alcohols.^[14] Subsequent bromination generated benzyl bromide 25 which was used to alkylate resorcinol 23A providing analog 9A. Synthesis of an amine-linked analog was achieved by the reduction of amide 8B to 9C and subsequent phenol deprotection to form 9B.

Phenol-protected analogs were explored by monoalkylation of symmetric biphenol 4C with a variety of alkyl bromides, yielding compounds 10B, 10D, 10F, 26A, 26B, and 26C.^[5] Analogs 26A-C were subject to Boc deprotection producing primary amines 10H-J. Analogs to probe the sterics of the phenol alkyl groups were obtained in two ways. McMurry coupling of 27 followed by alkene hydrogenation and phenol deprotection yielded 11E.^[15] Grignard reaction of 28 with phenyl magnesium bromide followed by hydrogenolysis led to the formation of 11F.^[6]

Antibacterial Activity.

Minimum inhibitory concentration (MIC) assays were performed to determine the antibacterial activity of these compounds against the gram-positive bacteria Methicillin-resistant S. aureus, E. faecalis, and E. faecium (see Figure 2). Antibacterial efficacy was dependent on the steric bulk of the group ortho to the phenol. Analogs with increasingly large groups in this position displayed more effective MICs, with the bulky tert-butyl analog 4C being the most effective. While benzyl substituted 11F was moderately effective against bacteria, phenyl substituted 11E exhibited no activity, suggesting that aromatic groups are not optimal for antibacterial activity.

Figure 2.

Chemical structures of synthesized analogs with average inhibitory activity against E. faecalis and MRSA. All compounds assayed were ≥90% pure as judged by UPLC unless otherwise noted (see SI).

Overall, any alkylation of the phenols decreased antibacterial activity. Furthermore, the level of activity diminished with increasing size the alkyl group protecting the phenol. While methylated compound 10A maintained modest activity compared to its biphenol parent, there was a sharp reduction in antibacterial activity for analogs with larger alkyl groups. Compounds 10C-10E are a notable exception to this trend, while they are of similar sizes, there is an increase in activity with increasing unsaturation. Propargyl 10E is more effective than both allyl 10D and propyl 10C

Varying the position of the phenol had moderate effects on the efficacy of analogs. Amongst the positional and linker length analogs, 1A was the least effective, but it still inhibited gram-positive bacteria at ≥16 μM. The flexibility in the linker region may allow compounds to reorient themselves, diminishing differences between analogs with differing substitution patterns.

Analogs with 3,3’- and 3,4’-biphenols with linkers between two and four carbons (4B-4D, 5B-5D) were most effective against gram-positive bacteria, while analogs lacking this substitution pattern and having linker lengths outside of this range were 2–8 fold less potent. While the vast majority of analogs with heteroatom linkers displayed reduced antibacterial activity, 3,3’-ester-linked analog 7A displayed activity similar to that of lead compound 4C, suggesting that this compound is oriented in a beneficial way. The rigidity of ester and amide linkers presumably prohibits reorientation of the phenyl units.

While potent analogs were found against gram-positive bacteria, the majority of compounds were ineffective against gram-negative bacteria. Broad-spectrum antibacterial activity was achieved with compounds 10H-J (see Table 1) in line with previous work showing that inclusion of a primary amine allows a gram-positive antibiotic to accumulate in gram-negative bacterial cells.^[16] While these analogs were 8–32 fold less effective than their biphenolic parent 4C, they did maintain activity against gram-negative bacteria A. baumannii. The analog with the shortest primary amine tether 10H, was the most effective of these compounds and was also active against E. coli. No compounds were found to inhibit the growth of P. aeruginosa

Table 1.

Activity against gram-negative bacteria.^[a]

Compound	10H	10I	10J
n=	2	4	6
A. baumannii	11.5	49.7	26.8
E. coli	23.3	-	>100
P. aeruginosa	>100	-	>100

^[a]MICs for A. baumannii, E. Coli, and P. aeruginosa are reported in μg/mL.

Toxicity Studies.

Hemolysis assays were also conducted to determine the toxicity of these compounds against mammalian red blood cells. To compare toxicity with bacterial inhibition, hemolysis data is reported as an HD₂₀ value, the concentration at which 20% of mammalian red blood cells hemolyze. The ratio of the HD₂₀ value to a bacterial MIC can be used to calculate the therapeutic index (TI) of a compound. A high TI corresponds to a compound that selectively targets bacteria without damaging eukaryotic cells.

Of the seven analogs that achieved average MICs ≤ 4 μM against gram-positive bacteria, six had TIs between 8 and 16 (Table 2). The remaining analog, ester-linked 7A, displayed a hemolysis value of 250 μM, 8-fold higher than the carbon-linked analogs with comparable antibacterial activity. While phenol-protected analogs exhibited reductions in antibacterial activity, their hemolysis values were also larger than their biphenolic counterparts. Methoxy analog 7E exhibits TIs against gram positive bacterial similar to 7A, while 10A and 10E have TIs 2–4 fold higher than 4C. Lead broad spectrum analog 10H has a modest TI of 8 for gram-positive bacteria that is even lower for gram-negative bacteria.

Table 2.

Hemolysis Values and Therapeutic Indices for the most effective gram-positive antibacterials and most effective broad spectrum antibacterial.^[a]

Compound	HD20	E. faecium TI	E. faecalis TI	MRSA TI	A. baumannii TI	E. coli TI
4B	9.69 μg/mL	-	16	8	-	-
4C	10.1 μg/mL	8	16	8	-	-
4D	10.6 μg/mL	8	16	8	-	-
5B	10.1 μg/mL	-	16	8	-	-
5C	10.6 μg/mL	8	16	8	-	-
5D	11.0 μg/mL	8	16	8	-	-
7A	85.6 μg/mL	63	63	63	-	-
7E	357 μg/mL	63	125	63	-	-
10A	85.1 μg/mL	-	63	31	-	-
10E	91.1 μg/mL	31	63	16	-	-
10H	46.2 μg/mL	8	8	8	4	2

^[a]Hemolysis values were calculated with defibrinated sheep blood and reported as Lysis₂₀ values which refer to 20% cell lysis compared to the Triton X control.

Mechanism of Action Studies.

Cell permeabilization was explored using SYTOX Green Nucleic acid stain with E. faecalis (Figure 3). All compounds tested showed various degrees of membrane permeation with the exception of the amides. The biphenols cause cell membrane permeabilization faster than counterparts where one of the phenols was methylated indicating that both hydroxyl groups play important roles (Figure 3b). Increasing the size of the phenol O-alkyl groups further hinder cell permeation, but the unsaturation of the O-allyl (10D) and O-propargyl (10E) analogs restored the permeation rate (Figure 2a). When the O-alkyl group contained a terminal amine(10H-J), permeation was reduced, with the shortest alkyl tether (10H) providing minimal permeation (Figure 3a). The biphenol with an ethylene linker (4C) permeated the membrane faster than the isosteric ester 7A (Figure 3b). Methylation of one phenol in both of these analogs reduced both the rate and overall amount of permeabilization (Figure 3b). In particular, the methylated ester variant 7E showing a larger drop in activity relative to the methylated ethylene-linked variant 10A. DiBAC₄(3) was used to examine changes to the cell membrane potential.^[17] An increase in fluorescence suggests that ethylene-linked compounds 4C and 10A and ester analogs 7A and 7E act via a mechanism that depolarizes the cell membrane (Figure 4).

Figure 3.

Cellular membrane permeability measured with SYTOX Green for 1 h with E. Faecalis. (a) Alkoxy groups diminish degree and rate of permeabilization at 50 μM. (b) Biphenolic ethylene outcompetes ester and methoxylated analogs at 25 μM.

Figure 4.

Cellular membrane depolarization measured with DiBAC4(3) for 1 h with E. Faecalis. Compounds 4C, 7A, 7E, and 10A all disrupt the membrane potential.

Bacterial Mutant Studies.

To further elucidate the mechanism of action, bacterial mutants were evolved through serial passage. However, after 25 passages, no resistance was generated against biphenols 4C and 7A. On the other hand, wild type MRSA quickly developed resistance against analog 7E resulting in three mutants. Data from MIC assays for one of these, MRSA_7E, revealed resistance against most monoalkoxylated analogs while remaining susceptible to biphenolic compounds (Table 3).

Table 3.

MIC Values for Wild Type MRSA and Mutant MRSA_7E.^[a]

Compound	E. faecalis WT	MRSAWT	MRSA7E
04C	0.65	1.31	0.65
07A	1.37	1.37	2.74
07B	5.48	5.48	2.74
07C	1.37	2.74	2.74
07D	1.37	2.74	2.74
07E	2.85	5.70	>90
07F	1.43	22.5	>90
10A	1.36	2.72	>90
10B	22.3	5.67	>90
10C	5.90	11.4	>90
10D	2.93	11.4	>90
10E	1.46	5.83	>90
10H	5.91	5.91	11.5
10I	12.3	12.3	12.3
10J	26.8	13.2	13.2

^[a]MICs for E. faecalis_WT, MRSA_WT, and MRSA_7E are reported in μg/mL.

Whole genome sequencing of the MRSA_7E mutant revealed three single nucleotide polymorphisms (SNPs) compared to the ATCC reference strain.^[18] One of these SNPs corresponded to intragenic regions with no predicted effects. A second SNP generates a stop codon early in the sequence corresponding to putative multidrug resistance protein emrY. This protein is thought to encode efflux pumps and cells without emrY are more sensitive to stressors including hydrogen peroxide, mitomycin C, and ultraviolet radiation.^[19]

The final SNP in MRSA_7E alters the sodium-binding pocket of the amino-acid carrier protein alsT, a sodium-alanine symporter that is conserved amongst different bacterial strains. The size and shape of the binding pocket of this ten-pass transmembrane protein determines what amino acids may enter the cell.^[20] The SNP in MRSA_7E corresponds to a serine to tyrosine mutation in a highly conserved transmembrane region of alsT (TM6). This amino acid coordinates to sodium ions in the binding pocket of TM6. The mutation maintains the hydroxy group involved in binding, but decreases the size of the pocket, perhaps preventing monoalkoxylated analogs from binding.

Intracellular Ion Assays.

To confirm the results of whole genome sequencing, changes in intracellular sodium concentration were detected in E. faecalis using the ratiometric indicator SBFI/AM (see Figure 5a). Analogs 7E, 10A, 10B and 10E created concentration dependent changes to intracellular sodium concentrations consistent with the mutation, while their biphenolic parents 7A and 4C had no effect.

Figure 5.

Intracellular ion concentractions measured with E. faecalis. (a) Methoxylated analogs change intracellular sodium concentration at 150 μM. (b) Biphenolic ethylene analog diminishes intracellular pH at 31 μM.

Cell membrane potential is affected by a number of ions, including H⁺. Changes in intracellular pH were detected in E. faecalis using the ratiometric indicator dye BCECF/AM. As seen in Figure 5b, carbon-linked biphenol 4C created concentration dependent changes to cytosolic pH, while alkoxy analogs 10A and 7E had no effect consistent with the activity observed above with the mutant MRSA_7E. Ester-linked biphenol 7A had no effect on cellular pH. This result further supports the theory that the ester-linked and carbon-linked biphenols act via different mechanisms.

Conclusion

Biphenolic honokiol analogs were successful at inhibiting gram-positive ESKAPE pathogens, Staphylococcus epidermidis, Enterococcus faecalis, and methicillin-resistant Staphylococcus aureus. Analogs with 3,3’- and 3,4’-biphenols with linkers between two and four carbons were most effective. Ester analog 7A showed comparable activity vs ethylene analog 4C, but with less toxicity thereby increasing the therapeutic index for gram-positive bacteria 3-fold.

Analogs 10H-J, featuring primary amine moieties, inhibited A. baumannii, achieving broad spectrum activity. Amongst these derivatives, 10H also inhibited E. coli, but no synthesized compounds displayed activity against P. aeruginosa. Most of the analogs permeabilize the cell membrane, with phenol protection inhibiting the rate of permeabilization.

In examining the mechanism further all compounds outside the pink box in Figure 7 causes depolarization of the cell membrane. Notably, not all the depolarizing compounds appear to involve identical mechanisms. A subset disrupted intracellular pH (orange box, Figure 7) and a different subset disrupted sodium transport (green box, Figure 7). The results from mutation studies with monoalkylated 7E are consistent with the sodium transport disruption as one mutation centers on alsT, a sodium-alanine symporter that is conserved amongst different bacterial strains. However, the more active compounds with two phenolic OH groups (4C, 7A) did not give rise to resistant mutants upon serial passage. Given the simplicity and easy access of 4C and 7A combined with limited ability to bacterial to evolve resistance, they are candidates for topical antibacterials or to treat surfaces. Preliminary assays reveal low toxicity to mammalian cells and the compounds are unlikely to be as environmentally durable as antibacterial triclosan, especially 7A because the ester can undergo biodegradation via hydrolysis.^[21]

Figure 7.

Overlapping activity profiles of the analogs.

Figure 1.

Structures of antibacterials previously investigated by the Kozlowski group.

Figure 6.

Intracellular ion concentrations measured with E. faecalis. (a) Methoxylated analogs change intracellular sodium concentration at 150 μM. (b) Biphenolic ethylene analog diminishes intracellular pH at 31 μM