Pentagalloyl glucose from Schinus terebinthifolia inhibits growth of carbapenem-resistant Acinetobacter baumannii

The rise of antibiotic resistance has necessitated a search for new antimicrobials with potent activity against multidrug-resistant gram-negative pathogens, such as carbapenem-resistant Acinetobacter baumannii (CRAB). In this study, a library of botanical extracts generated from plants used to treat infections in traditional medicine was screened for growth inhibition of CRAB. A crude extract of Schinus terebinthifolia leaves exhibited 80% inhibition at 256 µg/mL and underwent bioassay-guided fractionation, leading to the isolation of pentagalloyl glucose (PGG), a bioactive gallotannin. PGG inhibited growth of both CRAB and susceptible A. baumannii (MIC 64–256 µg/mL), and also exhibited activity against Pseudomonas aeruginosa (MIC 16 µg/mL) and Staphylococcus aureus (MIC 64 µg/mL). A mammalian cytotoxicity assay with human keratinocytes (HaCaTs) yielded an IC50 for PGG of 256 µg/mL. Mechanistic experiments revealed iron chelation as a possible mode of action for PGG’s activity against CRAB. Passaging assays for resistance did not produce any resistant mutants over a period of 21 days. In conclusion, PGG exhibits antimicrobial activity against CRAB, but due to known pharmacological restrictions in delivery, translation as a therapeutic may be limited to topical applications such as wound rinses and dressings.

Antibacterial activity of pentagalloyl glucose. To determine its activity against a wide range of A.
baumannii strains, PGG was tested for growth inhibition of one susceptible strain of A. baumannii and 23 drugresistant strains (selected for a diversity of resistance profiles, including 19 CRAB strains) ( Table 1), yielding minimum inhibitory concentrations (MICs) ranging from 64 to > 256 µg/mL (68 to > 272 µM) and IC 50 s ranging from 8 to 64 µg/mL (8.5-68 µM). PGG produced largely similar dose-response curves against 21 of the 24 A. baumannii strains tested (Fig. 3A); much of the variation in MIC values in these 21 strains emerges from small differences in growth inhibition at high concentrations of PGG, since PGG dose-response curves against A. baumannii seem to level off around 85-95% inhibition and our definition of MIC is the lowest concentration that exhibits 90% inhibition. PGG was also tested against a panel of other ESKAPE pathogens: Enterococcus faecium, Staphylococcus aureus, Klebsiella pneumoniae, Pseudomonas aeruginosa, and Enterobacter cloacae. The most potent growth inhibition by PGG was observed in P. aeruginosa, A. baumannii, and S. aureus, with MICs of 16,64, and 64 µg/mL, respectively ( Table 1). The commercially-sourced PGG (Sigma-Aldrich) was found to display identical growth inhibition of A. baumannii when compared to PGG isolated from S. terebinthifolia (Fig. 3B). Overall, PGG exhibited activity against a wide range of pathogenic bacteria at concentrations ≥ 64 µg/mL. aureus 13 , PGG was tested for biofilm formation inhibition and biofilm eradication against A. baumannii AB5075 at concentration gradients of 0.5-64 (sub-MIC) and 2-256 µg/mL, respectively. PGG did not exhibit any antibiofilm activity at the concentrations tested, showing mild promotion of biofilm formation relative to vehicle (DMSO) control at 8-64 µg/mL (Supplementary Figs. S9, S10). However, there was no significant difference between PGG and control with regards to biofilm promotion or inhibition at any concentration tested and no dose-response trend was noted. These data support our conclusion that PGG has no effect on biofilm formation or maintenance at this concentration range.
Cytotoxicity of pentagalloyl glucose. Immortalized human keratinocyte cells (HaCaTs) were used in a lactate dehydrogenase (LDH) assay to assess cytotoxicity by PGG and its parents. Both PGG and extract 429 exhibited an IC 50 of 256 µg/mL (Fig. 4). Using the median IC 50 (8 µg/mL) for growth inhibition of A. baumannii tested in this study, the therapeutic of index of PGG is 32.

Media supplementation experiments.
To elucidate potential mechanisms of action of PGG, A. baumannii growth inhibition and time-kill assays were carried out with various supplements in combination with PGG: 0.02% oleic acid, 0.02% polysorbate 80, 1 mM iron (II) sulfate, and 1 mM iron (III) sulfate. In growth inhibition assays of PGG in a gradient of 256-2 µg/mL, oleic acid supplementation produced no change in activity, but polysorbate 80, iron (II) sulfate, and iron (III) sulfate attenuated growth inhibition (Table 2). In time-kill experiments, the CFU/mL curve of PGG alone indicated that PGG's activity at 256 µg/mL is bacteriostatic rather than bactericidal (Fig. 5). Furthermore, PGG combined with oleic acid, polysorbate 80, and iron (II) sulfate treatments had roughly tenfold, 100-fold, and 1,000-fold higher CFU/mL measurements, respectively, than PGG alone at the 24 h timepoint.   , and a control plate containing no PGG. After 24 h of incubation, there was a full lawn on the control plate, but no colonies appeared on any of the PGG plates. Next, evolution of resistance was tested by serial passaging A. baumannii AB5075 for 21 days in the presence of a PGG gradient, using culture from the 0.5 × MIC treatment for each subsequent passage. Serial passaging with a tetracycline gradient served as a control. The MIC of PGG remained stable throughout the 21 daily passages, but the MIC of tetracycline increased from 4 µg/mL (susceptible) to 64 µg/mL (resistant) (Fig. 6).

Discussion
Pentagalloyl glucose (PGG) is a relatively well-known hydrolysable tannin with reported antimicrobial, antiviral, anticancer, antioxidant and antidiabetic properties 17 , but this study is the first to our knowledge to investigate its activity against A. baumannii. The bacterial growth inhibition exhibited by PGG in our study is in accordance with PGG's previously reported broad-spectrum activity, with MICs in the 50-500 µg/mL range and IC 50 values in the 5-50 µg/mL range 12 .
Previous studies with PGG have attributed its antibacterial activity against S. aureus to its chelation of iron, using a colorimetric assay to quantify binding of free iron by PGG and performing bioassays with iron supplementation and 2,2′-dipyridyl, a known iron chelator, as controls 12,14 . Iron chelators have been widely studied as antibacterial agents; in the case of gram-negative bacteria such as A. baumannii, potentially therapeutic chelators need to bind iron with higher affinity than the bacterial siderophores but also be distinct enough from these   www.nature.com/scientificreports/ siderophores to not associate with the receptors involved in transporting siderophore-bound iron through the outer membrane 18 . In our experiments with A. baumannii, supplementation of the growth media with iron (II) sulfate or iron (III) sulfate completely attenuated growth inhibition by PGG at the concentrations tested, indicating that the previously described chelation mechanism applies in an Acinetobacter system. Restoration of A. baumannii growth by addition of iron was also related to the concentration of PGG; the higher the concentration of PGG, the less restoration of bacterial growth was observed after iron supplementation, indicating that PGG interacts with iron in a concentration-dependent manner (Supplementary Fig. S11). This concentration-dependent interaction is also evidenced by optical density readings of PGG in iron (II) and iron (III) sulfate supplemented media (Supplementary Fig. S12). This suggests that PGG's sequestration capabilities, and thus A. baumannii inhibition, can be overwhelmed by high iron availability in the environment. Potential applications of PGG as a chelation agent would need to take this into consideration.
We found that 14 A. baumannii strains tested had MICs > 256 µg/mL. Speculating on the strain-specific resistance mechanisms based on these MIC values alone is difficult, but the high MIC values seen amongst many of these resistant strains may be attributed to the ability of A. baumannii to grow under iron deficient conditions, such as those generated by the iron chelating compound PGG. A previous study by Eijkelkamp et al. found that under iron limiting conditions A. baumannii strain ATCC 17978 upregulates genes associated with iron acquisition such as the iron-chelating compound acinetobactin and other related siderophores 19 . The high MIC values (> 256 µg/mL) found for many of the A. baumannii strains tested, including the reference strain ATCC 17978 which lacks a resistance profile, may be due to these low iron growth adaptations. It is likely that these same adaptations are found in other PGG resistant A. baumannii strains.
Other studies have found that PGG binds to lipopolysaccharide (LPS), a component of the outer membrane of gram-negative bacteria and a causative agent of sepsis 20 . LPS was previously considered essential to gramnegative bacteria, and therefore a target for antibiotics, but some strains, including strains of A. baumannii, have been found that can survive without LPS 21 . The attenuation of PGG's growth inhibition seen in our fatty acid www.nature.com/scientificreports/ supplementation experiments may be related to PGG's binding with LPS, but iron supplementation had a larger impact on PGG's growth inhibition of A. baumannii. A previous study has shown that colistin-resistant LPSdeficient A. baumannii strains may have higher susceptibility to non-polymyxin drugs due to their decreased membrane integrity 22 ; however, this has not been observed in clinical isolates, and compensatory mutations likely occur with respect to colistin-resistance [23][24][25] . If PGG's mechanism of A. baumannii inhibition is indeed through binding of LPS, these considerations may prove useful for the development of drug delivery systems involving PGG. For example, PGG may be utilized prior to an antibiotic, binding to LPS, in order to first disrupt bacterial membrane integrity. PGG's lack of anti-biofilm activity against A. baumannii contrasts with its documented anti-biofilm activity against S. aureus 13,14 . However, this discrepancy can be attributed to differences in biofilm mechanisms between A. baumannii and S. aureus; removal of iron prevents biofilm formation in S. aureus 14 , but studies with A. baumannii have shown an indifferent response 26 or even increased biofilm formation when iron is limited 27 . Furthermore, previous experiments with deferasirox, a clinical chelator with high affinity to iron, found no significant anti-biofilm activity on A. baumannii, suggesting that A. baumannii is highly capable of sequestering iron in biofilm 28 . However, every iron chelator has independent activity, and some results have been strain dependent 29,30 . Also, limiting iron negatively impacts motility 31 , which has been determined to play a key role in virulence 32,33 . From these results, it appears that PGG is capable of chelating iron from the medium to inhibit planktonic growth, but a physical or mechanistic difference in the bacteria's iron withholding capacity interferes with PGG's iron chelation capacity in biofilm. These results also reinforce the concept that differential responses between bacterial species can prevent antibacterial activity in one species from being generalized to other species without specific testing.
Through time-kill assays with PGG, we found it to be bacteriostatic against A. baumannii. Testing PGG against a panel of 24 A. baumannii strains with a wide range of resistance profiles showed that PGG's activity is generally consistent between strains, and similar growth inhibition activity in a panel of ESKAPE pathogens demonstrated that PGG has activity against both gram-positive and gram-negative bacteria. Passaging A. baumannii 21 times in the presence of PGG did not stimulate any spontaneous resistance, in contrast to the same passaging with tetracycline-another bacteriostatic agent-which produced a resistant phenotype (16-fold increase in MIC) after 10 passages. Together, these experiments indicate that PGG has broad-spectrum antibacterial activity, likely inhibiting growth by chelating iron, and that resistance to this activity does not develop quickly in A. baumannii. www.nature.com/scientificreports/ In cytotoxicity assays with human keratinocytes, PGG exhibited an IC 50 of 256 µg/mL, yielding a therapeutic index of 32 for PGG's growth inhibition of A. baumannii. PGG may therefore be useful as a topical therapy for infections, and the anti-infective traditional use of S. terebinthifolia and other members of the Anacardiaceae containing PGG is supported 10,12 .
Further development of PGG as a potential drug requires more mechanistic and pharmacokinetic/pharmacodynamic study. Previous pharmacological studies of PGG, recently reviewed by Patnaik et al. 34 , have found low oral bioavailability 35 but potential for intravenous administration 20,36 with possibilities for nanoparticle-based 37 and microbubble-based delivery 34 .
PGG's lack of specificity, typical to tannins 38 , may limit or at least modulate its use as an antimicrobial, but its activity against CRAB, an urgent healthcare threat 2 , supports its further investigation for anti-infective therapy.

Methods
Plant extracts. The leaves of Schinus terebinthifolia Raddi (Anacardiaceae), which were selected from a primary screen of the Quave Natural Product Library (QNPL), were collected from private property in DeSoto County, Florida, in November 2017. Plant samples were identified and catalogued by Dr. Cassandra Quave at the Emory Herbarium (GEO, Atlanta, Georgia) where voucher specimens were deposited for reference (CQ-651) and are accessible for view online via the SERNEC portal 39 . Fresh plant samples were dried in a dehumidified cabinet. Dried plant material was ground using a Wiley Mill Plant Grinder and filtered through a 0.5 mm mesh sieve. Powdered plant samples were subjected to two rounds of maceration in 80% aqueous ethanol for 72 h, filtered, and then dried via a rotary evaporator. Crude plant extracts were partitioned using a successive liquidliquid partitioning scheme. Extraction solvents used were: hexanes, ethyl acetate, n-butanol, and water and were labelled B, C, D, and E according to solvent, respectively. Dried crude plant extracts were stored in -20 °C until further use.
Isolation of bioactive compounds. Schinus terebinthifolia extract 429 underwent bioassay-guided fractionation as previously described 40  Mass spectrometry data was processed with Freestyle 1.6 software (Thermo Scientific). The predicted formula and mass were taken for 429C-F8-PF11-SF4 and the PGG standard according to the peak compound signature as determined previously by analytical HPLC of 429C-F8-PF11-SF4 compared to a standard of PGG.
Bacterial growth conditions. Bacteria were maintained on tryptic soy agar (TSA) plates and grown in cation-adjusted Mueller Hinton broth (CAMHB) for experiments according to CLSI guidelines 41  Time-kill. Time-kill methods were adapted from NCCLS guidelines 50 . PGG and meropenem were added at MIC concentrations (256 and 64 µg/mL, respectively) to A. baumannii AB5075 at 1 × 10 6 CFU/mL in CAMHB [alone or supplemented with 0.02% oleic acid or polysorbate 80 or 1 mM iron (II) sulphate]. At 0, 3, 6, and 24 h timepoints, 10 µL culture was removed from each of four replicates, serially diluted tenfold in sterile PBS, dispensed on TSA, and incubated to determine CFU/mL.

Biofilm inhibition and eradication. Methods for A. baumannii biofilm inhibition and eradication were
adapted from Tipton et al. 51 . Briefly, A. baumannii AB5075 cells were grown in Luria-Bertani broth (LB) for 24 h in the presence of treatment in the case of biofilm inhibition and for 24 h pre-treatment and 24 h post-treatment in the case of biofilm eradication. After incubation, biofilms were fixed with ethanol, stained with crystal violet, gently rinsed with water, and eluted with 33% acetic acid. The absorbance of each well was measured at 595 nm.
Mammalian cytotoxicity. Toxicity to immortalized human keratinocytes (HaCaTs) was assessed using an LDH cytotoxicity assay kit (G-Biosciences, St. Louis, MO, USA) as described previously 52 . Briefly, cells were standardized to 4 × 10 4 and incubated in 96-well plates for 48 h to allow for seeding, after which media was replaced and treatments were added in serial dilution from 256 to 32 µg/mL. Plates were subsequently incubated for 24 more hours and processed according to the manufacturer's protocol for chemical induced cytotoxicity.
Restoration assays. The effect of iron supplementation on PGG activity was determined for A. baumannii AB5075 by modifying Lin et al. 's iron restoration assays on S. aureus 14 . For the first half of the assay, PGG was mixed with solutions of cation-adjusted Mueller-Hinton Agar (CAMHA) and TSA at concentrations of 0.5 × MIC, 1 × MIC, and 2 × MIC (128, 256, and 512 µg/mL, respectively). A control condition with no PGG treatment was included. Each condition was tested in triplicate, along with media blank wells. Overnight culture of A. baumannii AB5075 was standardized to 5 × 10 5 CFU/mL in CAMHB media at OD 600 . AB5075 working culture was spread onto each treatment plate, statically incubated at 35 °C for 24 h, and observed for presence of bacterial growth. At the 24-h mark, 1 mM of iron (II) sulfate solution or 1 mM of iron (III) sulfate solution was spread onto the PGG-treated wells. The plates were visually observed for presence of bacterial growth at 24 and 48-h timepoints.

Development of resistance.
Spontaneous development of resistance was tested using a method adapted from Ling et al. 53 . A. baumannii AB5075 culture was spread on TSA plates containing 0, 128, 256, and 512 µg/ mL PGG and incubated for 24 h, after which colonies on each plate were counted.
Resistance over the course of serial passaging was tested using a method adapted from Maisuria et al. 54 . Briefly, a series of 21 microbroth dilution experiments for growth inhibition was carried with PGG or tetracycline against A. baumannii AB5075, using bacterial culture from the sub-MIC (0.5MIC) wells of each gradient to make the experimental culture for each subsequent experiment.
Disclaimer. Material has been reviewed by the Walter Reed Army Institute of Research (WRAIR). There is no objection to its presentation and/or publication. The opinions and assertions contained herein are the private ones of the authors and are not to be construed as official or reflecting the views of the Department of Defense, the Uniformed Services University of the Health Sciences, the Department of the Army, or any other agency of the U.S. Government.