Dimensionally Enhanced Antibacterial Library Screening

Abstract

The emergence and spread of antimicrobial resistance is a major public health threat, and there is an urgent need to develop new strategies to address the issue. In this study, the possibility of enhancing a whole cell based antibacterial library screen by increasing the dimensionality of the screening effort is explored using methicillin-resistant Staphylococcus aureus (MRSA) as the target organism. One dimension involved generating and screening a human liver microsome metabolized FDA approved drug library. Comparative screening of the un-metabolized (UM) and pre-metabolized (PM) libraries allows identification of intrinsically active agents from the UM library screen and of agents with active metabolites from the PM library screen. To further enhance this screening effort, it was combined with a −/+ resistant to antibiotic screen (−/+ cefoxitin; Cef). This allows the identification of agents that can act synergistically with the resistant to antibiotic. This approach revealed five compounds with substantially improved activity after metabolism and four compounds with substantial synergistic activity with cefoxitin. Capecitabine in particular only had significant antibacterial activity after metabolism. Its metabolites were isolated, identified, and characterized for spectrum of activity along with several other anticancer drugs with anti-MRSA activity. Floxuridine, gemcitabine, novobiocin, and rifaximin were identified as having substantial synergy with cefoxitin from the −/+Cef screens. Checkerboard assays verified synergy for these agents. Floxuridine demonstrated a particularly high degree of synergy with cefoxitin (FIC = 0.14). This study demonstrates how a dimensionally enhanced comparative screening effort can identify new antibacterial agents and strategies for countering antibacterial agent resistance.

Graphical Abstract

Pathogenic bacteria are becoming increasingly drug resistant, with some now virtually untreatable.^1,2 There has concurrently been a lack of new antibacterial agents identified over the last 30 years to counter this threat.^3–6 Methicillin-resistant Staphylococcus aureus (MRSA) is one of the ESKAPE (Enterococcus faecium, Staphylococcus aureus, Klebsiella penumoniae, Acinetobacter baumannii, Pseudomonas aeruginosa, and Enterobacter species) pathogens^7,8 and is a common cause of both nosocomial and community-acquired infections.^9,10 It is characterized by resistance to methicillin, oxacillin, most other commonly used β-lactam antibiotics, and to many other antibiotic classes and agents.¹¹

Chemical compound library screening is a foundational approach for the discovery of new bioactive agents, including for antibacterial activity. However, large whole cell and targeted library screening efforts for new antibacterial agents have given overall disappointing results.^12,13 An alternative to large library screens is smaller scale efforts such as FDA approved drug library screening, which has become a popular strategy for “drug repurposing”.¹⁴ This strategy can reveal novel new activities of FDA approved drugs, which provides a potentially greatly shortened path to clinical application.

The use of human liver microsomes for drug metabolism studies has also become commonplace.¹⁵ Nearly all drugs and other xenobiotics are transformed into at least one metabolite, and generally more. Such metabolites frequently have distinct biological activities.^16,17 It seemed likely that in situ microsomally generated metabolites of an FDA approved drug library could identify FDA approved drug metabolites as novel antibacterial agents, the demonstration of which was the initial major objective of this study. This requires a comparative screening approach, in which the un-metabolized (UM) and pre-metabolized (PM) libraries are screened in parallel.

Another strategy to counter antimicrobial resistance is to identify agents that can act synergistically with another antibiotic or restore the activity of a resistant to antibiotic. To further enhance the proposed UM vs PM library screen, it was combined in this study with a −/+ resistant to antibiotic (cefoxitin; Cef) screen. This provided a 2 × 2 experimental design (UM/PM vs −/+Cef), which provided both a degree of redundancy (replication) in this screening effort and two extra dimensions of information for use in filtering and assessing the screening results. This study demonstrates the ability of such a multidimensional screening approach to identify library compounds with active metabolites, as well as agents with substantial synergistic activity.

RESULTS AND DISCUSSION

Library Pre-Metabolism and Optimization of the Library Screening Workflow.

Several commercial sources of human liver microsomes had substantial levels of contaminating bacteria, which needed to be eliminated for the proposed study. An effective solution was developed in which, after metabolism of the FDA approved drug library, metabolized library samples were dried and compounds extracted from reaction mixtures with DMSO. These extracted PM library samples were dried and reconstituted in DMSO to provide a sterile PM working library at the same nominal concentration as the diluted UM working library (isomolar mixture of the parent drug and its metabolites). Initial anti-MRSA screens were performed using optical density (OD₆₀₀) to measure wells for growth/no growth, but this gave poor results in 384-well plates. However, a resazurin-based bacterial viability assay¹⁸ gave very good separation of known actives from known inactives, with Z′-factors¹⁹ in the range of 0.6–0.7 (Supplementary Figure 1).

Two-Dimensional FDA Library Screening to Increase Informational Value.

It seems reasonable to repeat a library screen at least once to ensure reproducibility. However, identical replication has low information content over the initial screen. A strategy of performing replication under somewhat different but informative conditions seemed optimal, which led to the two-dimensional (UM/PM vs −/+Cef) screening strategy described here (Figure 1). Initial library screens were performed at relatively high library compound concentrations of 200 μM and in the absence or presence of 8 μg mL⁻¹ cefoxitin (to which this strain of MRSA is resistant). The rationale for using high library compound concentrations in the initial screen (200 μM) was (1) to not miss any prospective actives, (2) to identify agents with possible active trace metabolites, and (3) to identify agents with weak intrinsic activity but potentially interesting synergistic or antagonistic interactions with Cef (the resistant to antibiotic). In view of the results summarized below, an initial screen at 50 or 25 μM would provide comparable results in future efforts with this approach (i.e., no high interest compounds had minimum MICs (minimum inhibitory concentrations) above 25 μM).

Figure 1.

UM/PM vs −/+Cef library screening workflow. (a) The PM library is generated by in vitro microsomal metabolism and working UM and PM libraries prepared at 1 mM. (b) UM and PM libraries are screened for antibacterial activity in the presence and absence of cefoxitin (8 μg mL⁻¹) to generate a merged hit list. (c) MICs are determined for each hit in the merged hit list under all four screening scenarios (UM/PM vs −/+Cef).

Analysis of Hits.

Following library screening, a pooled hit list was made (i.e., any compound that gave a hit (was active) in any of the four UM/PM vs −/+Cef screens was added to the list) for follow-up minimum inhibitory concentration (MIC) determinations. MICs for all of the compounds in this pooled hit list were then determined under all four screening conditions (UM−Cef, UM+Cef, PM−Cef, and PM+Cef) to give a final table of MICs. The results from these MIC determinations, sorted by minimum MIC and for minimum MICs of ≤100 μM, are summarized in Supplementary Table 1. A complete list of inactive agents is provided in Supplementary Table 2.

The Supplementary Table 1 list is expectedly dominated by known antibacterial agents. A number of “non-antibacterials” in the UM−Cef library screen were identified as having anti-MRSA activity, most of which have previously been identified in FDA approved drug library antibacterial activity screens.²⁰ These previously identified anti-MRSA “non-antibacterials” include floxuridine,^21–23 gemcitabine,^24,25 closantel,^26–28 carmofur,^23,29 5-fluorouracil,^23,30,31 doxifluridine,²³ and daunorubicin^23,32 (Supplementary Table 1). The UM−Cef library screen also identified epirubicin (MIC 12.5 μM) and idarubicin (MIC 12.5 μM) as having anti-MRSA activity (Supplementary Table 1).

Two comparisons are possible using the MIC data in Supplementary Table 1; one highlights those compounds with significant increases in activity after metabolism (Table 1), and the other highlights those compounds showing the greatest differences in activity between the −Cef and +Cef library screens (Table 2). These comparisons were made using an average log₂ (AL2) function.

Table 1.

FDA Library Anti-MRSA Hit MICs Sorted by Greatest Average Increase in Activity after Metabolism (AL2 ≥ 1.5)

	UM MICs (μM)		PM MICs (μM)
compound	−Cef	+Cef	−Cef	+Cef	Min_MIC	AL2(UM/PM)a
capecitabine	NAb	NAb	12.5	25	12.5	≥4.5
balofloxacin	0.78	0.78	9.8 × 10−2	9.8 × 10−2	9.8 × 10−2	3.0
marbofloxacin	6.25	6.25	0.78	0.78	0.78	3.0
carmofur	12.5	50	6.25	6.25	6.25	2.0
gatifloxacin	0.78	0.39	9.8 × 10−2	0.39	9.8 × 10−2	1.5

^aSee the following equation: $\text{Avg}\left({\text{log}}_2\left(\frac{{\text{UM}}_{-\text{Cef}}}{{\text{PM}}_{-\text{Cef}}}\right),{\text{log}}_2\left(\frac{{\text{UM}}_{+\text{Cef}}}{{\text{PM}}_{+\text{Cef}}}\right)\right)$

^bNA - no activity (MIC ≥ 200 μM, counted as 400 μM).

Table 2.

FDA Library Anti-MRSA MICs Sorted by Greatest Difference in Activity between −/+ Cefoxitin (Abs(AL2) ≥ 2)^a

	UM MICs (μM)		PM MICs (μM)
compound	−Cef	+Cef	−Cef	+Cef	Min_MIC	AL2(−Cef/+Cef)b
floxuridine	3.1	3.1 × 10−3	3.1	6.25	3.1 × 10−3	4.5
gemcitabine	2.4 × 10−2	1.5 × 10−3	0.78	4.9 × 10−2	1.2 × 10−2	3.5
novobiocin	0.20	1.5 × 10−3	9.8 × 10−2	0.20	1.5 × 10−3	3.0
rifaximin	0.39	1.2 × 10−2	6.25	3.1	1.2 × 10−2	3.0
dicloxacillin	3.1	6.25	0.78	6.25	0.78	−2.0
ceftiofur	6.25	100	100	100	6.25	−2.0
tebipenem pivoxil	25	100	12.5	100	12.5	−2.5
cloxacillin	3.1	25	3.1	25	3.1	−3.0
nafcillin	1.6	50	6.25	50	1.6	−4.0

^aThe upper section shows those agents with substantially enhanced activity in the presence of 8 μg mL⁻¹ cefoxitin. The lower section shows those compounds with substantially reduced activity in the presence of cefoxitin.

^bSee the following equation: $\text{Avg}\left({\text{log}}_2\left(\frac{{\text{UM}}_{-\text{Cef}}}{{\text{UM}}_{+\text{Cef}}}\right),{\text{log}}_2\left(\frac{{\text{PM}}_{-\text{Cef}}}{{\text{PM}}_{+\text{Cef}}}\right)\right)$

For the UM vs PM comparison:

$$\text{AL}{2}_{(\text{UM}/\text{PM})}=\text{Avg}\left({\text{log}}_2\left(\frac{{\text{UM}}_{-\text{Cef}}}{{\text{PM}}_{-\text{Cef}}}\right),{\text{log}}_2\left(\frac{{\text{UM}}_{+\text{Cef}}}{{\text{PM}}_{+\text{Cef}}}\right)\right)$$ (1)

For the −/+Cef comparison:

$$\text{AL}{2}_{(-\text{Cef}/+\text{Cef})}=\text{Avg}\left({\text{log}}_2\left(\frac{{\text{UM}}_{-\text{Cef}}}{{\text{UM}}_{+\text{Cef}}}\right),{\text{log}}_2\left(\frac{{\text{PM}}_{-\text{Cef}}}{{\text{PM}}_{+\text{Cef}}}\right)\right)$$ (2)

Identification of Novel Antibacterial Drug Metabolites.

This analysis for UM vs PM MICs (Table 1) revealed a number of compounds with increased activity after metabolism. (A number of active agents, not surprisingly, lost substantial activity after metabolism.) These include several fluoroquinolones, in particular balofloxacin, as well as several other antibiotics and several anticancer drugs. Capecitabine demonstrated a dramatic increase in activity after metabolism (Table 1), indicating at least one active metabolite. A scaled up microsomal metabolism reaction of capecitabine was performed and the reaction mixture fractionated by semipreparative HPLC. Active fractions were identified by anti-MRSA activity and active components identified by LC-MS/MS analysis and comparison with commercially available standards. This revealed two anti-MRSA active microsomal metabolites of capecitabine: 5′-deoxy-5-fluorocytidine (DFCR) (MIC = 3.1 μM) and doxifluridine (5′-deoxy-5-fluorouracil; DFUR) (MIC = 1.6 μM) (Supplementary Figure 2). DFUR is further converted to 5-fluorouracil (5-FU), the active anticancer agent, in tissues through the action of pyrimidine nucleoside phosphorylase.³³ The anti-MRSA activity of DFUR and 5-FU was also found in this (Supplementary Table 1) and in prior FDA library screening studies.^23,30,31 The discovery of the anti-MRSA activity of DFCR, which is not in the FDA approved drug library, demonstrates the potential of this approach to identify novel antibacterial metabolites of library agents. Active metabolite identification efforts of the other PM active compounds in Table 1 are underway.

Spectrum of Activity Determination.

A number of other 5-fluorouracil related anticancer agents were observed in this and other studies to be active against MRSA, as cited above. To assess the potential of this group of agents as anti-MRSA and antibacterial agents, their spectrum of activity was assessed against several MRSA strains, one VRE (faecium) strain and one E. coli strain (Table 3). Two control antibiotics (vancomycin and doxycycline) were also included in this study.

Table 3.

Spectrum of Activity of Capecitabine Metabolites and Related Compounds (MIC, μM)

compounda	MRSA (F-182)b	MRSA (N315)	MRSA (HI022)	MRSA (MN8)	MRSA (TCH70)	MRSA (RN1)	MRSA (COL)	VRE (faecium, clinical)	E. coli (Seattle 1946)
capecitabine	NAc	NAc	NAc	NAc	NAc	NAc	NAc	NAc	NAc
DFCR	3.1	NAc	NAc	6.3	25	12.5	3.1	NAc	NAc
DFUR	1.6	6.3	3.1	6.3	6.3	1.6	3.1	NAc	NAc
floxuridine	3.1	3.1	1.6	6.3	1.6	1.6	1.6	50	NAc
5-fluorocytidine	3.1	3.1	3.1	6.3	3.1	1.6	3.1	50	NAc
5-fluorouracil	3.1	12.5	6.3	12.5	12.5	3.1	3.1	12.5	NAc
5-fluorouridine	3.1	6.3	6.3	12.5	12.5	3.1	6.3	12.5	NAc
carmofur	12.5	25	6.3	12.5	12.5	6.3	6.3	6.3	NAc
gemcitabine	2.4 × 10−2	2.4 × 10−2	9.8 × 10−2	4.9 × 10−2	9.8 × 10−2	4.9 × 10−2	4.9 × 10−2	6.1 × 10−3	NAc
vancomycind	0.39	0.20	0.39	0.39	0.39	0.20	0.39	NAc	NAc
doxycyclined	0.39	0.39	0.39	0.39	0.39	0.39	50	3.1	1.6

^aSee the structures shown in the graphic above.

^bATCC 43300 strain used for library screening. Other vendor IDs given in the text.

^cNA - not active at 50 μM, the highest concentration used in these MIC determinations.

^dControl antibiotic.

None of these compounds demonstrated activity against E. coli. Capecitabine was inactive against all of the tested strains. DFCR was active against 5/7 MRSA strains but inactive against the VRE strain. DFUR was active against all of the MRSA strains but inactive against the VRE strain. 5-Fluorocytidine and floxuridine were active against all of the MRSA strains and weakly active against the VRE strain. 5-Fluorouracil, 5-fluorouridine, gemcitabine, and carmofur (another prodrug form of 5-fluorouracil with intrinsic antibacterial activity) were active against all of the MRSA and VRE strains. Gemcitabine in particular appears to have the potential for further study against G+ organisms.

Synergistic Activity of Library Hits with Cefoxitin.

Several compounds showed apparent synergy with cefoxitin (Table 2). In particular, floxuridine demonstrated an apparent strong synergistic activity with cefoxitin (AL2_(UM/PM) = 4.5, a 23-fold average change in the MIC of floxuridine −/+Cef). Gemcitabine, novobiocin, and rifaximin also demonstrated significant apparent synergy. To confirm and further assess this synergy, checkerboard assays³⁴ were performed with cefoxitin (Figure 2). Among these four compounds, floxuridine showed the highest synergy with an FIC³⁵ of 0.14. This value is comparable to the FIC of 0.15 for ampicillin/sulbactam synergy.³⁶ Gemcitabine, novobiocin, and rifaximin also exhibited appreciable synergistic activity with cefoxitin, with FICs of 0.38 each. The biochemical bases of these synergistic interactions are presently unknown but could provide a useful approach to countering β-lactam resistance in MRSA or other organisms. This is currently under further investigation. A number of antagonistic interactions were also revealed (Table 2, bottom). Of particular note was the observation that several β-lactams demonstrated significant anti-MRSA activity in the absence of cefoxitin, which was lost in the presence of cefoxitin (nafcillin, cloxacillin, dicloxacillin, and ceftiofur). This observation suggests that these four β-lactam antibiotics are unable to induce β-lactam resistance. Cefoxitin however is able to induce resistance, which then provides β-lactam resistance to these sensitive-to-β-lactam antibiotics as well. Further study of these antagonistic interactions is underway.

Figure 2.

Checkerboard assay results for combinations of cefoxitin with floxuridine, gemcitabine, novobiocin, and rifaximin. Isobolograms for combinations of cefoxitin (y-axes) with (x-axes): (a) floxuridine, (b) gemcitabine, (c) novobiocin, (d) rafiximin. FIC (min) values are given in the isobolograms. The dashed line in the isobolograms is for the expected no interaction (additive MICs) curve. Cef = cefoxitin, Flox = floxuridine, Gemc = gemcitabine, Novo = novobiocin, Rifa = rifaximin.

Conclusions.

This study initially aimed to evaluate the potential of in situ metabolism of an FDA approved drug library to enhance a drug repurposing screening effort by allowing active metabolites to be identified. To further enhance the ability of this approach to reveal novel and interesting activities, it was combined with a −/+ resistant to antibiotic (cefoxitin) screen. This provided a two-dimensional screening approach (2 × 2 experimental design). The results from this approach exceeded our expectations. Comparison of UM vs PM screens (Table 1) revealed a number of agents with potential active metabolites. Capecitabine was one such agent, and its anti-MRSA metabolites were fractionated to identify DFCR and DFUR. Several fluoroquinolones demonstrated substantially enhanced activity after metabolism. Fractionation of several of these metabolized fluoroquinolones reveals very low levels (<1%) of one or more very potent metabolites, which we are working on identifying. Comparison of −Cef vs +Cef screens identified a number of potentially synergistic agents with cefoxitin (Table 2), which was confirmed for four of these by checkerboard analysis (Figure 2). Floxuridine in particular demonstrated a high degree of synergy with cefoxitin. The mechanism of synergy of these agents with cefoxitin is unknown and is currently being investigated. Such synergistic activities could provide a means of restoring a resistant to antibiotic’s activity, or with a sensitive to antibiotic to enhance its activity and reduce the emergence of resistance. This study illustrates the utility of a comparative library screening approach to identify new antibacterial agents and approaches.

MATERIALS AND METHODS

General.

The FDA approved drug library of 978 compounds was from Selleckchem. Human liver microsomes were from Sekisui XenoTech LLC (catalog # H0630, pooled human liver microsomes from 50 donors of mixed sex, age, and race) and stored and used according to the manufacturer’s instructions. Cation adjusted Mueller-Hinton broth (CAMH) and Leucosnostoc mesenteroides glucose-6-phosphate dehydrogenase (G6PDH) were from Sigma-Aldrich. Glucose-6-phosphate (G6P) and β-nicotinamide adenine dinucleotide phosphate (NADP⁺) were from Alfa Aesar (a division of ThermoFischer Scientific). Multiwell polypropylene cluster tube plates (96 wells) were from Corning (catalog # 4411), and 96-well U-bottom polypropylene storage plates were from Becton Dickinson (catalog # 351190). Sterile 384-well microtiter plates were from Corning (catalog # 3680), and sterile 96-well microtiter plates were from MidSci (catalog # TP92096). Other reagents were from standard sources and were reagent grade or better. Bacterial strains were obtained from American Type Culture Collection (ATCC) and the Biodefense and Emerging Infections Research Resources Repository (BEI). The bacterial strain used for library screening was methicillin-resistant Staphylococcus aureus (MRSA) strain F-182 (ATCC 43300). Other bacterial strains used in this study were MRSA strain N315 (BEI NR-45898), HI022 (BEI NR-30550), MN8 (BEI NR-45918), TCH70 (BEI HM-139), RN1 (BEI NR-45904), and COL (BEI NR-45906); a clinical strain of vancomycin-resistant Enterococcus faecium (VRE, VanA type); and Escherichia coli strain Seattle 1946 (ATCC 25922).

Library Replication and Addition of Metabolism and Antibacterial Control Compounds.

The FDA approved drug library was delivered in columns 1–11 in 96 deep well plates, 12 plates total, with each sample well containing 100 μL of a 10 mM solution of a compound dissolved in DMSO. Aliquots (20 μL) of library samples were transferred to 96-well cluster tube plates using a liquid handling workstation (Biomek 3000). Antibiotic controls (20 μL of 10 mM stock solutions of vancomycin, fosfomycin, ampicillin, doxycycline, and chloramphenicol) were added to column 12 of odd numbered plates. Microsomal (CYP) substrate controls (20 μL of 10 mM stock solutions of phenacetin, tolbutamide, dextromethorphan, coumarin, chlorzoxazone, and diclofenac) were added to column 12 of even numbered plates. DMSO, which can interfere with microsomal drug metabolism reactions, was removed by freezing the plates at −80 °C and drying the replicated library plates under a strong vacuum (<50 μmHg) in a Genevac Quatro centrifugal concentrator.

In Vitro Microsomal Metabolism to Provide the PM Library.

To the dried plated library samples was added 20 μL of acetonitrile/water (20/80%, v/v) to redissolve samples. Plates were incubated for 2 h at 35 °C, followed by the addition of 980 μL of freshly prepared (on ice) microsomal reaction mixture containing 50 mM potassium phosphate pH 7.4, 3 mM MgCl₂, 5 mM glucose-6-phosphate, 1 unit mL⁻¹ glucose-6-phosphate dehydrogenase, 1 mM NADP⁺, and 0.5 mg mL⁻¹ total microsomal protein.³⁷ Reaction mixtures were incubated for 24 h at 35 °C with gentle rocking on a Cole-Parmer tilt shaker table. Library plates were then centrifuged at 4000g for 30 min at 4 °C in a Sorvall RT6000 refrigerated centrifuge and 800 μL of the supernatants transferred to sterile 96-deep-well cluster tube plates. To the residues was added 200 μL of DMSO, and the samples were mixed thoroughly. Library plates were centrifuged again at 4000g for 30 min, and 300 μL of the supernatants was removed and combined with the first extracts. The resulting extracts were frozen at −80 °C and dried under a strong vacuum (<50 μmHg) in a Genevac Quatro centrifugal concentrator. These “pre-met” (PM) library samples were then reconstituted in 200 μL of DMSO to provide a 1 mM PM FDA working library. The original FDA library was similarly replicated and diluted with DMSO to obtain a 1 mM “un-met” (UM) FDA working library. Both UM and PM working libraries were stored in U-bottom polypropylene storage plates (Becton Dickinson, catalog # 351190) at −80 °C. Samples of wells containing microsomally metabolized drug controls from both UM (un-metabolized control) and PM (metabolized sample) plates were analyzed by LC-MS/MS to provide a relative measure of metabolism. The percent metabolism of these control drugs was 52, 55, 60, 66, 95, and 100% for tolbutamide, dextromethorphan, chlorzoxazone, phenacetin, diclofenac, and coumarin, respectively. These controls demonstrate that the metabolism conditions employed in this study were sufficient to achieve a relatively high degree of metabolism.

UM/PM vs −/+Cef Library Screen against MRSA.

Four sets of library screening plates were prepared for the following screens: UM−Cef, UM+Cef, PM−Cef, PM+Cef. Two sets of UM plates and two sets of PM plates were first prepared from working library samples (4 μL @ 1 mM) in 384-well microtiter plates (Corning, catalog # 3680) using a Biomek 3000 liquid handing workstation. Plates were frozen at −80 °C and dried as described above. To each well in each set was added 20 μL of cation-adjusted Mueller-Hinton (CAMH) broth containing 4000 cfu of MRSA (ATCC 43300) and containing either no Cef for −Cef screens or +8 μg mL⁻¹ Cef (equal to 1/4 to 1/2× MIC) for +Cef screens. These additions were performed using an Integra Viaflo Assist automated multichannel pipetter in a Labconco BSL-2 biosafety cabinet. Plates were incubated for 48 h at 35 °C. Fresh CAMH broth (10 μL) was added to the wells of these four sets of plates, followed by incubation for 2 h at 35 °C to restart active cell growth. To the wells of these plates was then added 6 μL of 100 μg mL⁻¹ resazurin (sodium salt).^18,38,39 The plates were incubated for another 2 h at 35 °C, and the A₆₁₀/A₅₀₀ absorbance ratio (Promega Technical Bulletin TB317) was measured in a Molecular Devices SpectraMax M5 multimode microplate reader.

Hit Picking and Minimum Inhibitory Concentration (MIC) Determination.

Library screening data was processed and analyzed using Matlab scripts (The Mathworks, Natick, MA). Based on the values for known active and inactive antibacterial agent controls (Supplementary Figure 1), a cutoff value between active and inactive compounds was selected and lists of active wells in each screening set (UM−Cef, UM+Cef, PM−Cef, PM+Cef) were generated. These lists were merged to give a pooled hit list. Rows were added to this pooled hit list to include known active and inactive antibiotics containing wells as controls. MICs were determined by hit picking 2 μL samples from both UM and PM working plates (two sets from each) into the first columns of 384-well plates (four sets total, for UM−Cef, UM +Cef, PM−Cef, and PM+Cef MIC determinations). These samples were then serially diluted in steps of two across the plates with DMSO using an Integra Viaflo Assist automated multichannel pipetter. The last column was left blank (DMSO only). These plates were frozen at −80 °C and dried under a strong vacuum as described above. To each well in each set was added 20 μL of cation-adjusted Mueller-Hinton (CAMH) broth containing 4000 cfu of MRSA (ATCC 43300) and containing either no Cef for −Cef MICs or 8 μg mL⁻¹ Cef for +Cef MICs. (This provided MIC plates with 100 μM as the highest test agent concentration.) Plates were incubated for 48 h at 35 °C. Fresh CAMH broth (10 μL) was added to the wells of these four sets of plates, followed by incubation for 2 h at 35 °C to restart active cell growth. To the wells of these plates was then added 6 μL of 100 μg mL⁻¹ resazurin. The plates were incubated for another 2 h at 35 °C, and the A₆₁₀/A₅₀₀ absorbance ratio, measured as described above. MICs were determined using a cutoff midway between known active and inactive samples. All MICs were determined at least in triplicate and at least in quadruplicate for MIC_min ≤ 25 μM to ensure reproducibility.

Scaled up Capecitabine Metabolism Reaction and Active Metabolite Purification and Identification.

Capecitabine demonstrated significant antibacterial activity only after metabolism (Table 2). To identify its active metabolite(s), a scaled-up metabolism reaction was performed on 1 mL of 400 μM capecitabine using the reaction conditions described above for PM library preparation for 24 h. To this reaction mixture was then added 4 mL of 80% ice-cold isopropanol/200 mM acetic acid, the sample mixed well, and microsome debris pelleted at 4000g for 30 min at 4 °C. The supernatant was collected, and the pellet re-extracted with 2 mL of 67% isopropanol/200 mM acetic acid. The combined extracts were frozen at −80 °C and dried in a Genevac Quatro centrifugal concentrator. The residue was dissolved in 0.4 mL of 5% acetonitrile/95% 10 mM ammonium acetate and syringe filtered to make a sample for semipreparative HPLC. The sample was purified by semipreparative HPLC on a Kromasil C18 column (3.0 × 150 mm², 5 μm particle size, catalog # K08670646). Fractions were collected at 1 min intervals in a 96-well microtiter plate (MidSci, catalog # TP92096). The flow rate was 250 μL min⁻¹, and the purification gradient was 3% B for 0–5 min, 3–13% B for 5–65 min, 13–100% B for 65–75 min; solvent A = 10 mM ammonium acetate pH 6.5, solvent B = 30% solvent A with 70% acetonitrile, and solvent C = 100% acetonitrile.

Identification of Active Capecitabine Metabolites.

Purified HPLC fractions (4 μL each) were dispensed into a fresh sterile 96-well plate (MidSci, catalog # TP92096). This plate was frozen at −80 °C and dried in a Genevac Quatro centrifugal concentrator. To each well was added 100 μL of CAMH broth containing 4000 cfu of MRSA and plates incubated for 48 h at 35 °C. Fresh CAMH broth (50 μL) was added, plates incubated at 35 °C for 2 h, 30 μL of 100 μg mL⁻¹ resazurin added, plates incubated an additional 2 h, and the A₆₁₀/A₅₀₀ ratio used to identify active fractions. Active fractions were analyzed by liquid-chromatography tandem mass spectrometry (LC-MS/MS) on an AB Sciex 3200 QTrap mass spectrometer coupled to a Shimadzu UPLC system and high-resolution mass spectrometry (HRMS) on a Thermo Q Exactive Plus mass spectrometer for compound identification. Compound identifications were confirmed by comparison with authentic commercial samples.

Spectrum of Activity of Capecitabine Metabolites and Related Agents.

MICs were determined for capecitabine, its microsomal metabolites (DFCR, DFUR), its in vivo ultimate active anticancer metabolite (5-fluorouracil), and several related compounds (floxuridine, 5-fluorocytidine, 5-fluorouridine, carmofur, and gemcitabine) against a panel of bacterial strains (Table 3), as described above. The strains tested were MRSA strains F-182 (ATCC 43300), N315 (BEI NR-45898), HI022 (BEI NR-30550), MN8 (BEI NR-45918), TCH70 (BEI HM-139), RN1 (BEI MR-45904), and COL (BEI NR-45906), one strain of VRE (clinical), and one strain of E. coli Seattle 1946 (ATCC 25922).

Checkerboard Assays to Confirm Synergy with Cefoxitin.

A number of agents demonstrated apparent synergistic activity with cefoxitin (upper entries in Table 2). Checkerboard assays³⁴ were performed to confirm synergy for the top ranked agents: floxuridine, gemcitabine, novobiocin, and rifaximin. Checkerboard assays were performed in 96-well plates from DMSO stocks and dried under a vacuum similarly to the MIC determinations described above. To these plates was added 100 μL of CAMH broth containing 4000 cfu of MRSA to each well, and the plates were incubated for 48 h at 35 °C. Fresh CAMH broth (50 μL) was added, plates incubated at 35 °C for 2 h, 30 μL of 100 μg mL⁻¹ resazurin added, plates incubated an additional 2 h, and resazurin absorbance ratio was measured as described above. All checkerboard assays were performed in triplicate.