An LC-MS/MS assay and complementary web-based tool to quantify and predict compound accumulation in E. coli

Abstract

Novel classes of broad-spectrum antibiotics have been extremely difficult to discover, largely due to the impermeability of the Gram-negative membranes coupled with a poor understanding of the physicochemical properties a compound should possess to promote its accumulation inside the cell. To address this challenge, numerous methodologies for assessing intracellular compound accumulation in Gram-negative bacteria have been established, including classic radiometric and fluorescence-based methods. The recent development of accumulation assays that utilize liquid chromatography–tandem mass spectrometry (LC-MS/MS) have circumvented the requirement for labeled compounds, enabling assessment of a substantially broader range of small molecules. Our unbiased study of accumulation trends in Escherichia coli using an LC-MS/MS-based assay led to the development of the eNTRy rules, which stipulate that a compound is most likely to accumulate in E. coli if it has an ionizable Nitrogen, has low Three-dimensionality and is relatively Rigid. To aid in the implementation of the eNTRy rules, we developed a complementary web tool, eNTRyway, which calculates relevant properties and predicts compound accumulation. Here we provide a comprehensive protocol for analysis and prediction of intracellular accumulation of small molecules in E. coli using an LC-MS/MS-based assay (which takes ~2 d) and eNTRyway, a workflow that is readily adoptable by any microbiology, biochemistry or chemical biology laboratory.

Introduction

Drug-resistant bacterial infections are a major public health threat, with 2.8 million antibiotic-resistant bacterial infections reported in the United States annually, resulting in >35,000 deaths and >$5.6 billion in associated healthcare costs¹. Gram-negative (GN) bacterial infections, in particular, significantly impact mortality; as one example, extended-spectrum β-lactamase-producing Enterobacteriaceae caused 197,400 infections and 9,100 deaths in 2017 (ref. ¹). To combat this challenge of drug resistance, new antibacterial classes and compounds active against novel targets are needed. Despite decades of discoveries and scores of compounds with promising activity against Gram-positive (GP) bacteria, strategic modification of compounds to broaden their spectrum of activity has not generally been successful^2,3. Outer-membrane-mediated impermeability and promiscuous efflux pumps work in concert to prevent intracellular accumulation of most compounds in GN bacteria, and as such, no new class of antibiotics active against GN pathogens⁴ has been FDA approved since the fluoroquinolones in 1968⁵.

Notably, the nature of bacterial evolution suggests that antibacterial targets present in GP bacteria will often also be good targets for GN pathogens⁶. Indeed, studies with GN permeability-defect strains show that GP-only antibiotics typically act against these bacteria^7,8, suggesting that optimization of accumulation could lead to broader-spectrum activity. While radiometric and fluorescence-based approaches have been used for quantitative assessment of intracellular accumulation of compounds inside bacteria^9–16, such studies suffer from the need for radiolabeled compounds or compounds with inherent fluorescence. Liquid chromatography–tandem mass spectrometry (LC-MS/MS)-based approaches do not have these limitations, and therefore enable the assessment of a wide scope of compounds, including non-antibiotics. Such activity-agnostic accumulation studies can provide a more general understanding of permeation trends, leading to the development of predictive guidelines describing physicochemical properties that promote intracellular accumulation and can be a critical tool in antibiotic drug discovery. Here we present a detailed protocol for evaluating intracellular compound accumulation in Escherichia coli using a recently developed LC-MS/MS-based assay and for predicting compound accumulation using the web tool eNTRyway.

Development of the protocol

We initially developed the accumulation assay to generate a broad understanding of accumulation trends in E. coli, with the goal of establishing predictive guidelines describing physicochemical traits of compounds that are likely to accumulate. The assay was used to assess the ability of >180 compounds to accumulate in E. coli¹⁷. These results, in conjunction with calculated physicochemical properties and a random forest prediction model, led to the eNTRy rules^3,17, which stipulate that a compound is most likely to accumulate in E. coli if it has an ionizable Nitrogen (primary amines are best), has low Three-dimensionality (globularity ≤0.25) and is relatively Rigid (rotatable bonds ≤5). These guidelines have since been applied to the design of compounds that achieve high accumulation and/or activity in GN bacterial species^17–27. Additionally, a retrospective analysis of the literature identified examples prior to the formalization of the eNTRy rules in which the serendipitous addition of an amine to a rigid, planar GP-only antibiotic improved the spectrum of activity, most notably the discovery of ampicillin³.

Other properties that may lead to improved accumulation include steric accessibility of the amine and high amphiphilic moment¹⁷. For the steric accessibility of the amine, it has been observed that primary amines on primary carbons often accumulate to the highest extent, whereas secondary, tertiary and quaternary amines demonstrate lower accumulation. Additionally, primary amines on secondary or tertiary carbons may demonstrate lower accumulation than comparator compounds in which the amine is more sterically accessible. This data are generally consistent with observations about the accumulation of N-alkyl guanidiniums and pyridiniums²⁶. Amphiphilic moment, which measures the distance between hydrophobic and hydrophilic portions of a compound, appears all to be positively correlated with accumulation. Side-by-side compound comparisons demonstrating these trends are shown in Extended Data Fig. 1. Others have suggested that compound accumulation in GN bacteria is limited to those with a molecular weight <600 Da²⁸.

To facilitate the implementation of the eNTRy rules, we developed an open-access web application, eNTRyway, that predicts compound accumulation in E. coli¹⁹. Furthermore, the accumulation assay has also been used to study accumulation trends of compounds with other positively charged, nitrogen-containing functional groups, identifying both guanidiniums and pyridiniums as potential motifs that can facilitate accumulation in GN bacteria²⁶.

Antibiotic efficacy (low minimum inhibitory concentration (MIC) values) in permeability and/or efflux pump deficient GN bacterial strains suggests there are numerous compounds that could be potent against GN bacteria if higher levels of intracellular accumulation could be achieved^7,8. The eNTRy rules provide guidelines to design new compounds with enhanced intracellular accumulation in E. coli and, when coupled with structure–activity relationships and crystal structure data, can be effective in the development of high-accumulating derivatives with an expanded spectrum of activity. There have now been several demonstrations of the utility of the eNTRy rules in converting GP-only compounds into versions that have GN activity²⁷. The first example was the conversion of deoxynybomycin (DNM), a GP-only gyrase inhibitor that already met the globularity and rotatable bond criteria, into 6-DNM-NH₃, through the introduction of a primary amine to the molecule¹⁷. Since then, Debio-1452, a potent GP-active fatty acid biosynthesis inhibitor, has been modified to the compound Debio-1452-NH₃, which shows improved accumulation in E. coli, and activity against a wide panel of multidrug resistant GN pathogens¹⁹. Notably, although numerous Debio-1452 derivatives were previously synthesized and evaluated for broad-spectrum activity, Debio-1452-NH₃ is the first to demonstrate useful GN activity. The scope of activity of the flavin mononucleotide riboswitch inhibitor, Ribocil-C, has similarly been improved through the addition of a primary amine¹⁸. Ribocil-C-PA shows high levels of accumulation in E. coli and activity against multidrug-resistant E. coli, E. cloacae and Klebsiella pneumoniae strains, and also validates the flavin mononucleotide riboswitch as a suitable target for GN bacteria.

Overview of the procedure

The workflow for the accumulation assay, as outlined in Fig. 1, begins with conducting control experiments, followed by evaluation of compounds in the accumulation assay, LC-MS/MS analysis and data processing. The accumulation assay procedural workflow is provided in greater detail in Fig. 2. Briefly, bacterial samples are incubated with compound for 10 min. To separate extracellular compound from the bacterial pellet, centrifugation through silicone oil is used⁹. Centrifugation through silicon oil removes the need for additional washes and generally leads to overall higher compound signal. It is important to note that the optimal density of the silicone oil is between 1.02 and 1.04 g/mL. A density of 1.01 g/mL does not provide consistent, robust separation of the bacterial pellet from the medium, whereas 1.05 g/mL is too dense for the bacterial pellet to spin through effectively. Following separation of the bacterial pellet and supernatant containing extracellular compound, LC-MS/MS analysis is used for compound quantification. A compound is designated as an ‘accumulator’ if it reaches intracellular concentrations that are statistically significantly higher than those of the negative antibiotic controls, whereas compounds that do not achieve such statistical significance are designated as ‘non-accumulators’. Importantly, here we are using ‘accumulation’ to describe the steady-state concentrations of compounds after a 10 min incubation period, reflecting the equilibrium between influx and efflux under these assay conditions.

Fig. 1 |. Stages of the accumulation workflow.

The four major stages include: (1) control experiments prior to running accumulation assay, (2) evaluation of compound accumulation in the accumulation assay, (3) quantification of intracellular accumulation using LC-MS/MS analysis, and (4) data workup and statistical analysis to identify accumulators and non-accumulators.

Fig. 2 |. Accumulation assay workflow.

In brief, bacteria are grown to mid-exponential phase (Step 10), harvested and washed (Step 11–16), incubated with compound (Steps 21, 36), extracellular compound is removed (Step 22–25), cells are lysed (Step 26–28) and intracellular compound is extracted from the pellet (Step 37–42). Samples are then submitted for LC-MS/MS analysis to quantify intracellular compound (Step 43–59). Figure adapted with permission from ref. ³.

In a complementary approach, prediction of the likelihood of compound accumulation can be performed using the open-access web tool eNTRyway, freely accessible at www.entry-way.org¹⁹. eNTRyway readily calculates properties associated with improved accumulation in E. coli¹⁷, enabling users to prioritize the construction and evaluation of compounds that fall within these guidelines. This is particularly useful for users interested in winnowing down a large collection of in silico compounds to a smaller subset of compounds that are most likely to accumulate and therefore would be prioritized for synthesis and evaluation. Users submit simplified molecular-input line-entry system (SMILES) strings describing the chemical structure of either single compounds or batches of compounds to eNTRyway to calculate the relevant properties and predict whether the compounds of interest will accumulate in E. coli. The workflow is provided in Fig. 3, and screenshots of the process are provided in Extended Data Figs 2 and 3.

Fig. 3 |. Procedure for prediction of compound accumulation in bacteria using eNTRyway.

A list of SMILES strings is generated, canonicalized using Open Babel, and submitted to eNTRyway through www.entry-way.org. Users can then prioritize the testing of compounds predicted to accumulate and use the calculated properties to better design compounds that may accumulate. Screenshots of this process are shown in Extended Data Figs 2 and 3.

Applications

The accumulation assay and eNTRy rules have already found application in understanding compound accumulation in GN bacteria and have also aided in the conversion of GP-only compounds into broad-spectrum antibiotics^17–19. The accumulation assay is a valuable tool to determine if an antibiotic in development is limited by intracellular accumulation. It also has utility in identifying how small structural modifications can affect accumulation within compound classes. The use of wild-type bacterial strains coupled with porin knockout¹⁷ or efflux pump knockout strains¹⁹ can provide information on compound mode of uptake and compound susceptibility to efflux, respectively.

Kinetics of compound accumulation

The initial assay was developed assessing compound accumulation at a 10 min time point¹⁷, but it has since been optimized such that accumulation at time points up to 2 h can be evaluated, provided cells are viable¹⁹, and as such, the kinetics of compound accumulation over time can be studied.

Use in other bacterial strains

Although this accumulation protocol is described in E. coli, it can be easily adapted to other strains of GN bacteria, and we have reported its use in Pseudomonas aeruginosa¹⁹. Although the eNTRy rules were specifically developed for E. coli, 6-DNM-NH₃, Debio-1452-NH₃ and Ribocil-C-PA display potent activity against a broader panel of GN pathogens, along with improved accumulation in these bacteria^17–19. Due to differences in outer membrane architecture, porin expression and efflux pump activity, it is likely that, in P. aeruginosa, accumulation trends may not always conform to the eNTRy rules. Further studies are needed to fully understand species-specific accumulation trends.

Comparisons with alternative methods

A number of different assays have been developed to study intracellular compound accumulation in bacteria. Accumulation assays that use radiometric detection demonstrate excellent sensitivity and low background^{10,11,14,16,29}. However, these assays require the use of radiolabeled compounds, and are impractical for the evaluation of large collections of small molecules. Fluorescence-based accumulation assays rely on the compounds of interest either being intrinsically fluorescent or amenable to the addition of a fluorescent tag^{12,13,15,30–32}. While these assays have provided important accumulation data for the fluoroquinolone and tetracycline antibiotic classes, the necessity for fluorescence is a major limiting factor³³. Additionally, as it is now well documented that minor structural modifications can impact accumulation, the addition of a fluorescent tag may not provide an accurate representation of the accumulation of the parent compound. Furthermore, depending on the detection wavelength, there can be substantial background interference when using a fluorescence-based technique^34–36.

An accumulation assay that uses LC-MS/MS as a readout for compound quantification provides distinct advantages relative to other methods since LC-MS/MS provides high sensitivity, low background and a wide dynamic range, and enables the assessment of a broad range of compounds³⁷. As such, we selected LC-MS/MS as the readout for this assay and incorporated best practices from previously described accumulation assays. Some important factors to consider in assay development include the media used, time for compound incubation, method for separating extracellular compound, and method of cell lysis. To minimize bacterial growth during the compound incubation period, phosphate-buffered saline (PBS) was selected as the medium of choice for the assay. Williams and coworkers had previously reported that a 10 min incubation period was sufficient to achieve steady-state accumulation levels for most compounds¹⁶. To remove any extracellular compound, some strategies utilize multiple washes of the cell pellet^10,38; however, this increases the potential of cell lysis and loss of compound. Instead, we adapted a method that utilizes a centrifugation through oil to separate the bacterial pellet from extracellular compound, used by Bazile and coworkers in their fluorescence-based method for detecting intracellular fluoroquinolone accumulation⁹; centrifugation through oil eliminates the need for any additional washes. For cell lysis, sonication or freeze–thaw cycles can be utilized. While the excellent method of Davis and coworkers³⁸ has some procedural similarities, the accumulation assay reported herein has notable differences. For the compound incubation step, Davis and coworkers³⁸ use high-nutrient growth media and incubate compounds with bacterial samples for 30 min, in contrast to a 10 min incubation in PBS for this assay. Most importantly, Davis and coworkers³⁸ employ four washes after compound incubation, while this method uses a facile, rapid silicone oil separation.

More importantly, the LC-MS/MS-based assay presented here has been shown to accurately differentiate between positive and negative control antibiotics and has been used to assess the accumulation of hundreds of compounds¹⁷; initial reports with other methods described herein have limitations such that they were used to test a handful of compounds (typically ten or fewer) and typically without multiple positive and negative control compounds. Further implementation of this assay will facilitate an even more comprehensive understanding of compound uptake in GN bacteria, ultimately accelerating the design of high-accumulating antibacterial agents and novel broad-spectrum antibiotics.

Experimental design

Overview of the accumulation assay workflow

Evaluating compound accumulation consists of (Fig. 1): (1) accumulation assay controls (colony-forming unit (CFU) determination, lysis validation, viability studies), (2) evaluation of compound accumulation, (3) LC-MS/MS sample analysis and (4) data analysis. We recommend performing three biological replicates.

Controls

Control experiments should be performed for each new bacterial strain that is studied. First, the number of CFUs is determined at an OD₆₀₀ = 0.55–0.60. It is important to establish the correlation of OD₆₀₀ and the number of CFUs, as a turbidity test alone is unreliable for quantifying the number of viable CFUs. The lysis method is then validated to confirm that all bacterial cells are lysed, thus ensuring all intracellular compound will be released. In addition, it is important to evaluate cellular viability upon treatment with compounds that display antibiotic activity at the relevant concentration and time point.

Positive and negative controls for the accumulation assay are critical. Ideally, these are compounds whose accumulation has been demonstrated by an independent assay and whose antibiotic spectrum of action is well characterized. As such, inclusion of GP-only antibiotics (compounds with an MIC ≤64 μg/mL for E. coli) and E. coli–active antibiotics (MIC ≤4 μg/mL) are important (control MICs presented in Table 1). Numerous GP-only antibiotics are appropriate negative controls, including vancomycin, linezolid, fusidic acid and novobiocin. Chloramphenicol and the tetracyclines serve as excellent positive controls, as they provide a robust signal and maintain cell viability under the assay conditions. Many fluoroquinolones can be used as positive controls, though due to rapid bactericidal activity and potent MICs, cellular viability should be ensured at a 10 min time point. β-Lactam antibiotics covalently bind to their targets, preventing detection in this assay. As such, β-lactams can be used as negative controls. Positive controls will typically exceed accumulation values of 300 nmol/10¹² CFUs, whereas negative controls are expected to accumulate to ≤300 nmol/10¹² CFUs. There should be a statistically significant difference between the positive and negative controls, as determined by a two-tailed t-test. Every batch of compounds tested in the assay after validation of antibiotic accumulation needs to include a positive control, such as chloramphenicol or tetracycline.

Table 1 |.

MICs of antibiotic controls used in the assay against E. coli MG1655

Compound	MIC (μg/mL)
GP-active only
Daptomycin	>512
Fusidic acid	512
Novobiocin	256
Vancomycin	256
Clindamycin	256
Mupirocin	256
Erythromycin	32
Rifampin	16
Covalently modified
Ampicillin	8
Broad-spectrum
Chloramphenicol	4
Tetracycline	2
Ciprofloxacin	0.0078

As shown in Fig. 4, statistically significant differences between these positive and negative controls indicate that intracellular accumulation, not nonspecific binding to the outer membrane, is being quantified. Additional controls to further validate this result include evaluating compound accumulation upon cotreatment with a permeabilizer, as well as evaluating compound accumulation in efflux-deficient strains of bacteria^17,19. We would expect to see an increase in whole-cell accumulation under both of those conditions. If a compound of interest is known or believed to go through a porin, evaluating accumulation in that porin knockout strain should decrease whole-cell accumulation, and can also be a useful validation¹⁷.

Fig. 4 |. Validation of compound accumulation in E. coli.

a, Compounds that demonstrate antibacterial activity against E. coli accumulate to a statistically significant extent, whereas inactive and covalently modified compounds do not accumulate to a significant extent. b, Low-accumulating controls demonstrate improved accumulation when cotreated with a permeabilizer. c, Compounds that permeate through the ompF/C porins demonstrated decreased accumulation when their regulator, ompR, is knocked out. All compounds were tested in biological triplicate, and error bars represent standard error of the mean. Data adapted with permission from ref. ³.

Bacterial strains

This assay was initially optimized using E. coli MG1655¹⁷. The protocol has since been expanded to evaluate accumulation in E. coli and P. aeruginosa clinical isolates and genetic knockout variants¹⁹. Growth time to an OD₆₀₀ = 0.55–0.60 can vary depending on the strain used, but typically ranges from 2 to 4 h. When expanding accumulation studies to other bacterial strains, growth medium, centrifugation time, compound incubation time and cell lysis method may need to be reoptimized, but our experience with P. aeruginosa suggests this will likely be straightforward. Minor changes, such as extending the centrifugation time through oil from 2 min to 5 min or extending compound incubation time from 10 min to 30 min, are often sufficient to achieve optimal results. For P. aeruginosa, growth medium was changed from LB to tryptic soy broth, but all other steps remained the same.

Compound concentration

Compounds are evaluated for accumulation at a final concentration of 50 μM to achieve intracellular concentrations that are high enough to be detected within the dynamic range of the instrument without compromising cellular viability. Treatment of cells with this compound concentration also importantly provides statistically significant differences between accumulation values observed for the active controls compared with the inactive controls.

Mass spectrometry standard solutions

For LC-MS/MS analysis, 5 mM compound stock solutions made in DMSO or ultrapure water are used. Solvent selection should be based on compound solubility. For calibration curve generation and sample analysis, a mixture of 20 samples can be run, assuming each compound has a distinct MS/MS fragmentation pattern. It is important to ensure prior to making the mixture of stock solutions that target compounds do not share the same MS/MS fragments (i.e., in the case of evaluating isomers in the accumulation assay). For isomers that do have identical MS/MS fragments, analysis must be performed with individual solutions.

Accumulation units

The units ‘nmol/10¹² CFUs’ are used to describe intracellular accumulation. This describes the average intracellular concentration across the bacterial population. It could be possible to calculate the approximate number of molecules per cell, using the known volume of E. coli cells and the known relationship between OD₆₀₀ and cell number^39,40.

Medium selection

PBS is used for the compound incubation period as it prevents cell growth during the incubation while preserving cell viability. This contrasts with using high-nutrient medium, such as lysogeny broth (LB) medium. If medium is exchanged, CFU determination will need to be performed at the end of the compound incubation period, instead of at the beginning. A main concern of using PBS is a loss of efflux pump activity. Through data acquired in an efflux pump deficient strain, it was demonstrated that efflux pumps must be active to some extent during the 10 min compound incubation, as a statistically significant increase in accumulation in efflux pump deficient strains compared with WT strains was observed¹⁹. However, it is likely that efflux pumps are more active under high-nutrient medium conditions, and therefore compounds might show lower whole-cell accumulation values under high-nutrient medium conditions; data from Davis and coworkers indicate that this is the case when quantifying salicyl-AMS accumulation³⁸. It could be possible to evaluate accumulation in PBS with a nutrient such as glucose added, as a means to increase efflux pump activity while also limiting cell growth during the assay.

Timing of assay

Earlier reports from Williams indicate that it only takes ~10 min to reach equilibrium between influx and efflux rates¹⁶, and accumulation assessments after 10 min readily differentiates positive and negative antibiotic controls. Time points of up to 2 h can also be analyzed assuming that cells maintain viability upon compound treatment in PBS over that period¹⁹.

Limitations

Although the eNTRy rules have proved to be general and actionable²⁷, eNTRyway does not incorporate two other factors that may contribute to accumulation. In addition to the presence of a primary amine, low rotatable bonds and low three-dimensionality, it was found the steric accessibility of the primary amine and the compound’s amphiphilic moment are also important¹⁷. In addition to the eNTRy parameters, these two additional factors should also be considered when evaluating a compound’s likelihood of accumulating.

Since the accumulation method utilizes LC-MS/MS, compounds that are covalently modified upon entering the cell, including the important β-lactam class of antibiotics, cannot be detected. On rare occasions, the limit of detection for a particular compound can prevent accurate quantitation of intracellular accumulation. For example, aminoglycoside antibiotics have a poor signal at low concentrations, and accumulation values of these compounds cannot be accurately determined using this assay. LC-MS/MS analysis is also restricted to bacterial population analyses, in comparison with fluorescence-based methods that have been able to achieve single-cell resolution³³. The LC-MS/MS assay relies on compound solubility in aqueous medium; if not fully soluble upon addition to PBS, compound may precipitate and appear with the bacterial pellet, leading to false positive accumulation. For compounds where solubility is a concern, users should assess solubility before testing in the assay. We have rarely had problems with solubility when testing charged species, such as acids, amines and zwitterions, but compound precipitation can be more common when evaluating neutral compounds. As would be expected, solubility trends tend to be consistent across specific compound classes. Additionally, treatment with bactericidal antibiotics, particularly fluoroquinolones, can sometimes lead to a loss in viability, even at a 10 min time point, preventing accurate quantification. Viability is especially important to assess when quantifying accumulation in strains that are genetically modified to be more susceptible to antibiotics, such as efflux pump knockout strains.

Another limitation is that the assay only reports on whole-cell accumulation and cannot differentiate between accumulation in subcellular compartments, a challenge recently highlighted in accumulation studies performed by Iyer et al.⁴¹. Although the assay readily identifies compounds that accumulate to a statistically significant extent in the cytoplasm, as the cytoplasm makes up the majority of the cell volume, it is less certain whether the assay can detect compounds that primarily accumulate in the periplasm. This uncertainty is driven by a lack of control compounds, as many compounds that have periplasmic targets, including the β-lactams and the recently reported GN-active arylomycin derivatives⁴², are covalently modified in the periplasm, preventing detection using this assay. Additionally, the periplasm accounts for ~7% of the whole-cell volume⁴⁰, indicating that, if a compound were to accumulate to a high level in the periplasm but not the cytoplasm, it is possible that the level of whole-cell accumulation would not reach levels that are statistically significant over the negative controls⁴⁰. Using established methodologies for subcellular fractionation coupled with LC-MS/MS analysis, Prochnow and coworkers recently developed an assay that can be used to distinguish between small-molecule accumulation in the cytoplasm and periplasm⁴⁰. Extensions of this method and their broader implementation will greatly improve our understanding of the subcellular distribution of compounds in GN bacteria.

Of course, accumulation values are not directly related to antibiotic activity⁴¹. Although accumulation is a prerequisite for activity for most antibiotics, a number of other factors impact MIC values, including the binding affinity of the compound to its target, and the intracellular distribution of compound.

Materials

Biological materials

• E. coli MG1655 was purchased from the American Type Culture Collection (ATCC; cat. no 700926) ▲ CRITICAL Prior to use, strains should be stored at −80 °C in the appropriate medium supplemented with 20% (vol/vol) glycerol. Under these conditions, the strains should be stable for a few years.

Reagents

• PBS 1×, without calcium or magnesium (Corning, cat. no. 21-040-CM)

• LB medium, Lennox (Millipore Sigma, cat. no. L3022)

• Bacto agar (BD Difco, cat. no. 214010)

• Dimethylsulfoxide (DMSO; Millipore Sigma, cat. no. D8418) ! CAUTION Manipulate this compound with nitrile gloves, under a fume hood.

• Ultrapure water (Millipore)

• AR20 silicone oil (Millipore Sigma, cat. no. 10836, CAS no. 63148-58-3) ▲ CRITICAL The density of the silicone oil is vital to ensuring the appropriate separation of extracellular compound from the bacterial pellet. The AR20 oil should have a density of ~1.01–1.02 to achieve this. Double check the reported density of the AR20 silicone oil prior to purchasing. Despite having the same CAS number, brands and batches will differ in their formulation, leading to oil densities ranging from 0.96 to 1.05 g/mL.

• High-temperature silicone oil (Millipore Sigma; cat. no. 175633) ▲ CRITICAL The density of the silicone oil is vital to ensuring the appropriate separation of extracellular compound from the bacterial pellet. The high-temperature silicone oil should have a density of 1.05 g/mL. Double check the reported density prior to purchasing.

• Liquid nitrogen

• Acetone (Millipore Sigma, cat. no. 179124)

• Tetracycline (Millipore Sigma, cat. no. T7660)

• Ciprofloxacin (Millipore Sigma, cat. no. PHR1044)

• Chloramphenicol (Millipore Sigma, cat. no. C1919)

• Ampicillin (Millipore Sigma, cat. no. A0166)

• Fusidic acid (Carbosynth, cat. no. AF23629)

• Mupirocin (Millipore Sigma, cat. no. 1448901)

• Clindamycin (AK Scientific, cat. no. J10047)

• Daptomycin (AK Scientific, cat. no. G627)

• Vancomycin (Millipore Sigma, cat. no. V1130)

• Rifampicin (Millipore Sigma, cat. no. R7382)

• Erythromycin (Millipore Sigma, cat. no. E5389)

• Novobiocin (Cayman Chemicals, cat. no. 18457)

• Colistin (Millipore Sigma, cat. no. C4461)

• Dry ice

• HPLC grade water (Fisher Scientific, cat. no. W5SK-4)

• HPLC grade acetonitrile (Fisher Scientific, cat. no. A998-4)

• HPLC grade methanol (Fisher Scientific, cat. no. A452-4)

• Formic acid (Sigma, cat. no. 5330020050) ! CAUTION Manipulate this compound with nitrile gloves, under a fume hood.

Equipment

• Autoclave

• 10 mL serological pipettes (Thermo Scientific, cat. no. 02-923-204)

• 25 mL serological pipettes (Thermo Scientific, cat. no. 02-923-205)

• Serological pipettor (Stellar Scientific, cat. no. MTC-PP-CON)

• 50 mL Falcon tubes (Fisher Scientific, cat. no. 14-432-22)

• 500 mL Erlenmeyer flasks (Fisher Scientific, cat. no. S63273)

• Eppendorf Scientific Shaker Flask Clamps (Fisher Scientific, cat. no. 14-278-163)

• 0.6 mL microcentrifuge tubes (Thomas Scientific, cat. no. 1149J99)

• 1.7 mL microcentrifuge tubes (Thomas Scientific, cat. no. 1149X63) ▲ CRITICAL Boil-proof microcentrifuge tubes need to be used when performing freeze—thaw cycles.

• Water bath, 3 L (VWR)

• Rattler plating beads (Zymo research, cat. no. S1001)

• Sterile spreader (VWR, cat. no. 89042-021)

• Inoculation loop (VWR, cat. no. 12000-810)

• Bunsen burner or laminar flow hood ▲ CRITICAL E. coli MG1655 is a Biosafety Level 1 bacterial strain, so it can be worked with either on the bench under sterile conditions or in a laminar flow hood.

• Shaking incubator, Innova 4200 shaker (New Brunswick Scientific, Innova 4200 shaker)

• Stationary incubator (Fisher Scientific, Isotemp incubator)

• Micro centrifuge (Eppendorf, centrifuge 5424)

• Benchtop centrifuge (Eppendorf, centrifuge 5810 R)

• −80 °C freezer (Thermo Scientific, cat. no. RDE60086FA.)

• Mass spectrometer (Sciex, 5500 QTrap)

• HPLC (Agilent, 1200 series with a degasser, an autosampler with thermostat and a binary pump)

• Nitrogen generator (Parker Hannifin, Model LCMS5000NA)

• Column (Agilent, Zorbax SB-Aq column (4.6 × 50 mm, 5 μm), cat. no. 846975-914)

• Glass vial (Macherey-Nagel, cat. no. 702283)

• Vial Cap (Macherey-Nagel, cat. no. 702732)

• Glass insert (Macherey-Nagel, cat. no. 702716)

• Pipettes (Gilson, P10, P100, P200, P1000)

• Pipette tips (Fisher Scientific, cat. no. 02-681-140, cat. no. 02-681-412)

• Syringe (Agilent, cat. no. 5190-1522)

Computational

• Sciex Analyst software (version 1.7.1)

Reagent setup

LB liquid medium

Mix 20 g LB powder in 1 L ultrapure water. Sterilize by autoclaving. This medium can be stored at room temperature (RT, 22 °C) for 1 month.

LB agar plates

LB-agar plates, 2% (wt/vol): add Bacto agar to LB medium and sterilize it by autoclaving. Pour warm medium into appropriately sized Petri dishes, and allow it to solidify. LB-agar plates can be stored at 4 °C for 1–2 weeks.

Solutions for accumulation assessment

All solutions for accumulation assessment were either prepared freshly or were stock solutions that had been stored in a −20 °C freezer and were re-evaluated for compound purity using LC-MS prior to testing.

Use the following formula to calculate the amount of compound to be weighed:

$$M=C\times V\times MW$$

where M is mass (g), C is concentration (mol/L), V is volume (L) and MW is molecular weight (g/mol).

For the compounds listed in the table below, we made 5 mM solutions. We advise making 1 mL and storing in 50 μL aliquots at −20 °C.

If you are testing different compounds, you will need to investigate compound solubility in advance. If the compounds of interest are commercially available, use the reported solubility as a guide for solvent selection. Otherwise, most organic molecules are soluble in DMSO, and most charged compounds are soluble in water. Some very hydrophobic compounds may not be soluble in either and thus may not be able to be tested in the assay.

Compound	Solvent
Tetracycline	100% DMSO
Ciprofloxacin	Ultrapure water
Chloramphenicol	100% DMSO
Ampicillin	Ultrapure water
Fusidic acid	100% DMSO
Mupirocin	100% DMSO
Clindamycin	100% DMSO
Daptomycin	Ultrapure water, supplemented with 50 mg/L calcium: CaCl2 is an appropriate salt form to use
Vancomycin	Ultrapure water
Erythromycin	100% DMSO
Rifampin	100% DMSO
Novobiocin	100% DMSO
Colistin	Ultrapure water

▲ CRITICAL Because daptomycin’s antibacterial activity is calcium dependent, calcium must be supplemented in the stock solution⁴³. Ion dependencies of your chosen compounds can be checked by performing MICs in the absence and presence of the ion of interest. This is not a concern for compounds that lack antibacterial activity.

Water with 0.1% formic acid

Mix 1 L HPLC grade water and 1 mL LC-MS grade formic acid. The solution is stored at RT for up to 2 weeks.

Acetonitrile with 0.1% formic acid

Mix 1 L HPLC grade acetonitrile and 1 mL LC-MS grade formic acid. The solution is stored at RT for up to 2 weeks.

Water with 20% methanol

Mix 80 mL HPLC grade water and 20 mL HPLC grade methanol. The solution is stored at RT for up to 1 week.

Oil preparation

A 9:1 mixture of AR20 silicone oil:high-temperature silicone oil is used. The final density of the mixture should fall between 1.02 and 1.04 g/mL. The AR20 oil should have a density of ~1.02, and the high-temperature silicone oil should have a density of ~1.05 g/mL to achieve this range. If the AR20 oil is reported to have a slightly lower density, you can increase the amount of high-temperature silicone oil proportionately. If the AR20 is reported to have a density that falls in the range of 1.02–1.04, there is no need to add the high-temperature silicone oil. A sample density calculation is shown below:

$$(\text{AR}20\text{oil density}\ast 0.9)+(\text{High}-\text{temperature oil density}\ast 0.1)=\text{Final density}$$

▲ CRITICAL Achieving a silicone oil density of 1.02–1.04 g/mL is vital to ensuring the appropriate separation of extracellular compound from the bacterial pellet.

Procedure

Prediction of high-accumulating compounds using eNTRyway ● Timing 10 min to 1 d; entirely dependent on number of compounds in batch

1. Prepare a list of SMILES strings for compounds of interest. For single compound submissions proceed to option A, and for batch compound submissions, proceed to option B. SMILES strings can easily be generated in structure drawing programs, including ChemDraw. To achieve consistent eNTRyway results, we recommend using Open Babel GUI to canonicalize the SMILES strings prior to submitting compounds on eNTRyway⁴⁴. Open Babel can be downloaded for free at https://openbabel.org/docs/dev/Installation/install.html.

   1. Single compound submissions

      1. Copy and paste the SMILES string in the ‘input’ box, making sure that the box ‘Input below’ is checked. A screenshot of the input is shown in Extended Data Fig. 2.
      2. In the center panel, check the boxes ‘Add hydrogens (make explicit)’ and ‘Canonicalize the atom order’.
      3. Click convert, and if the box ‘Output below only’ is checked, the converted SMILES string will appear in the output box. Alternatively, an output file can be created.
   2. Batch compound submissions

      1. Create a file with an appropriate format with a list of SMILES strings. .txt files work well.
      2. Select the appropriate input format, and uncheck the ‘Input below’ box. This will allow you to select the batch file.
      3. Follow Step 1A-(ii-iii), either observing the output in the output box or by creating an output file.

2. Create an eNTRyway account on http://www.entry-way.org/ and login.

3. Click on ‘Submit molecule’.

4. Submit molecules as either single compounds or in batches.

Single compound submission:

Give a molecule name, paste in the SMILES string and click ‘submit molecule’.

Batch submission:

This is done using a list of SMILES strings that can be generated in a text editor or Excel. The first column should contain the SMILES string, and the second column should contain the molecule name. Any whitespace is an acceptable spacer. Acceptable file extensions are . smi and .txt. On the eNTRyway website, select the file, provide a batch name and then submit the batch.

• 5 Monitor progress of analysis, and view results.

Single compound submission:

Click on ‘view molecules’. Once the compound has been submitted, the structure will appear with the job status listed underneath. When the job status is complete, properties will be listed including the number of rotatable bonds, globularity and whether the compound has a primary amine.

Batch submission:

Click on ‘batch results’. Upon batch submission, the job status will be listed as ‘waiting’. Upon completion, a csv file will be available for download that contains calculated properties. The length of time for batch analysis depends on the number of compounds included in the batch, as well as the degree of conformational freedom. The greater the number of potential conformers, the longer the analysis will take. Often, batch analysis can be completed in a few hours.

• 6 Identify whether compounds of interest fit suggested eNTRyway parameters. The guidelines developed indicate that a compound is most likely to accumulate in E. coli if it is relatively rigid (rotatable bonds ≤5), is relatively flat (globularity ≤ 0.25), and has an ionizable nitrogen, with primary amines being the best.

  ▲ CRITICAL STEP In addition to the listed factors, amphiphilic moment and steric accessibility of the primary amine also need to be considered. Compounds with higher amphiphilic moments tend to show improved accumulation, while sterically encumbered primary amines, such as primary amines on tertiary carbons, show reduced accumulation. Amphiphilic moment can be calculated using the program MOE. Additional details on amphiphilic moment and MOE can be found in our prior publication¹⁷. MOE is available for download at https://www.chemcomp.com/Download_Request.htm.

• 7 Prioritize compounds for further experimental validation based off their predicted likelihood of accumulating.

Preparation of the bacterial suspension ● Timing ~1.5 d

• 8 Day 1. Pick a single colony from a fresh LB-agar plate using a sterile pipette tip or sterile loop, and inoculate it in 10 mL of LB medium in a 50 mL Falcon tube. Grow three independent cultures if planning on running an experiment in triplicate.

• 9 Incubate the culture overnight in a shaking incubator at 37 °C and 250 rpm. This step takes 14–20 h.

• 10 Day 2. Dilute 2.5 mL of the overnight culture into 250 mL fresh LB, and grow to mid-exponential phase (OD₆₀₀ = 0.55–0.60) in a 37 °C shaking incubator (1:100 dilution of an overnight culture of E. coli MG1655 takes ~2 h to reach mid-exponential phase).

  ▲ CRITICAL STEP Steps 8–10 should be performed prior to all CFU determination, lysis validation, viability assessment and accumulation assays.

CFU determination assay ● Timing ~1.5 d (~1 h/rep of hands-on time); stagger bacterial culture growth if performing multiple replicates in 1 d

• 11 Harvest 200 mL of bacterial culture into four 50 mL Falcon tubes, and pellet at 3,220g at 4 °C for 10 min. Discard the supernatant.

  ? TROUBLESHOOTING

• 12 Resuspend the bacterial pellets in 40 mL total (10 mL PBS for each Falcon tube used in Step 11) sterile PBS buffer (consolidate into two Falcon tubes, 20 mL PBS/tube).

• 13 Pellet at 3220g at 4 °C for 10 min. Discard the supernatant.

• 14 Resuspend the pellets in 8.8 mL total of PBS buffer (consolidate into one Falcon tube prior to preparing into aliquots).

• 15 Prepare 875 μL aliquots of the large culture in ten 1.7 mL microcentrifuge tubes.

• 16 Equilibrate samples at 37 °C by shaking in incubator for 5 min.

  ■ PAUSE POINT Cells will not grow in PBS and will remain viable for up to 2 h.

• 17 Serially dilute the cultures in sterile Eppendorf tubes 1:10 seven times to reach 10⁻⁷ in PBS (100 μL into 900 μL; 10× seven times). Spread the last two dilutions (10⁻⁶ and 10⁻⁷) evenly on fresh LB plates using either sterile beads or a sterile spreader.

  ▲ CRITICAL STEP 100 μL of bacterial culture are added to 900 μL PBS, mixed, and then transferred to the next tube of 900 μL PBS, using a different pipette tip for each step to avoid contamination.

• 18 Incubate plates at 37 °C overnight and count colonies the following morning. Plates with 100–400 colonies should be used to determine the relationship between OD₆₀₀ and CFUs, as those numbers of colonies provide the most accurate representation of the original bacterial culture.

• 19 Calculate the number of CFUs according to the following formula:

  $$N=\frac{C\times 10}{{10}^{-D}}$$

  where N is CFUs/mL, C is the number of colonies per plate and D is the number of the 10:1 dilution. Perform the experiment in triplicate each time the plate is restreaked.

Lysis validation protocol ● Timing ~1.5 d (~1.5 h/rep of hands-on time); stagger bacterial culture growth if performing multiple replicates in 1 d

• 20 To perform lysis validation, repeat Steps 11–16 of the CFU determination assay.

• 21 Add 8.75 μL of DMSO to bacterial samples, and incubate at 37 °C with shaking for 10 min.

• 22 While incubating bacterial samples, cool 1.7 mL microcentrifuge tubes with 700 μL silicone oil mix (9:1 AR20/Sigma High Temperature) to −78 °C using a dry ice/acetone bath.

  ▲ CRITICAL STEP If preparing aliquots from a large batch of mixed oil, properly mix prior to preparing aliquots into 1.7 mL microcentrifuge tubes to ensure that the oil will have the appropriate density. Cooling the mix to −78 °C using a dry ice/acetone bath further improves separation.

• 23 Invert bacterial samples twice to ensure a mixed solution, then place 800 μL of bacterial sample on 700 μL silicone oil mix (precooled to −78 °C) in the 1.7 mL microcentrifuge tube.

• 24 Pellet samples by centrifuging at 13,000g at RT for 2 min.

  ▲ CRITICAL STEP If layering occurs correctly, the meniscus curves up. The oil and the PBS containing extracellular compound should be in distinct layers; if the oil is not formulated correctly, the layers may not be clear, or it may appear to be an emulsion. Clear separation of layers is necessary for the separation of extracellular compound.

  ? TROUBLESHOOTING

• 25 Pipette to remove the medium and oil.

  ▲ CRITICAL STEP Carefully remove the top PBS layer first, then the oil layer. Removing the oil should be done in a couple of cycles. Remove the majority using a P1000 pipette, followed by a P100 and P10 to ensure removal of as much oil as possible.

• 26 Resuspend bacteria in 200 μL ultrapure water, and transfer to a new 1.7 mL microcentrifuge tube for lysis by freeze–thawing.

  ! CAUTION Make sure to use boil-proof microcentrifuge tubes. If non-boil-proof microcentrifuge tubes are used, the pressure build-up can cause the tops to explode off.

• 27 Freeze cultures in liquid nitrogen for 3 min, then transfer to a 65 °C water bath for 3 min.

• 28 Repeat Step 27 two more times (18 min total for lysis).

• 29 To ensure that the cells have been lysed, serially dilute in sterile Eppendorf tubes to 10⁻⁶ in PBS (100 μL into 900 μL;10× six times) and spread on fresh LB plates either using sterile beads or using a sterile spreader.

• 30 Incubate plates at 37 °C overnight, and count colonies the following morning. If the lysis is successful, no cells should be viable on the LB plates.

  ? TROUBLESHOOTING

Viability determination protocol ● Timing ~1.5 d (~1 h/rep of hands-on time); stagger bacterial culture growth if performing multiple replicates in 1 d

• 31 For viability, repeat Steps 11–16 of the CFU determination protocol.

• 32 Add compound to each sample by adding 8.75 μL of a 5 mM DMSO stock to reach a final compound concentration of 50 μM. A DMSO control should be included by adding 8.75 μL of 100% DMSO to one sample. Invert tubes two times to ensure mixing.

• 33 Incubate samples at 37 °C with shaking for 10 min.

• 34 Mix cultures again, and then serially dilute them in sterile Eppendorf tubes to 10⁻⁶ in PBS (100 μL into 900 μL;10× six times) and spread on fresh LB plates either using sterile beads or using a sterile spreader.

• 35 Incubated plates at 37 °C overnight, and count colonies the following morning. All samples should be compared with a DMSO control to ensure cell viability upon antibiotic treatment.

  ? TROUBLESHOOTING

Accumulation assay protocol ● Timing ~2 h/rep; stagger bacterial culture growth if performing multiple replicates in 1 d

• 36 Repeat Steps 11–16 from the CFU determination protocol, followed by Steps 21–28 from the lysis validation protocol. The only modification is to Step 21, where you add 5 mM compound stocks made up in either DMSO or water to the bacterial samples instead of DMSO. Prepare fresh 5 mM stock solutions during the 2 h bacterial growth period to reach mid-exponential growth phase. Add the same volume (8.75 μL) to reach a final compound concentration of 50 μM. Include a positive control in each batch.

  ▲ CRITICAL STEP Prior to testing, ensure the compound of interest is soluble at 50 μM in PBS. If not soluble, the compound can precipitate in solution and pellet with the bacterial lysate, leading to false positive results.

• 37 Centrifuge to pellet the lysates at 13,000g for 2 min at RT to remove debris.

• 38 Remove 180 μL of the aqueous supernatant, and transfer to a new, labeled 1.7 mL microcentrifuge tube.

• 39 Wash the bacterial pellet by adding 100 μL of methanol to each microcentrifuge tube containing bacterial lysate, vortex the samples briefly and repellet at 13,000g at RT for 2 min.

• 40 Remove 100 μL of methanol wash, and add to the previously collected 180 μL of aqueous supernatant.

• 41 Centrifuge samples from Step 40 at 20,000g at RT for 10 min.

• 42 Carefully remove 200 μL of the resulting supernatant from the top, and submit for LC-MS/MS analysis, along with at least 50 μL of 5 mM compound stock solutions in either DMSO or water.

  ■ PAUSE POINT Samples can be kept in a −80 °C freezer for up to 1 week before LC-MS/MS analysis.

LC-MS/MS analysis

▲ CRITICAL The LC-MS/MS procedure outlines how to analyze 20 different target compounds in triplicate (60 samples total).

Preparation of individual target compound solution (5 μM) for mass spectrometry (MS) signal optimization ● Timing 0.5–1 h

• 43 Thaw the 5 mM target stock solutions (the solvent is DMSO or water) and samples at RT.

• 44 Prepare 5 μM standard solutions. To prepare 20 different standard solutions, prepare aliquots of 1 μL of 20 individual 5 mM stock solutions into 20 separate Eppendorf vials containing 1 mL 20% methanol in water.

  ▲ CRITICAL STEP It is very important to review the compounds’ chemical structures to estimate the intensity of MS signals. For compounds that are likely to have weak MS signals (i.e., those without the presence of N, −COOH, −COO- or benzyl ring, etc.), it may be better to prepare the solutions at a higher concentration (>5 μM).

  ■ PAUSE POINT Target stock solutions can be stored in a −80 °C freezer for 5 years.

Initial MS setting for manual optimization ● Timing 5 min

• 45 Set the following HPLC and MS parameters as the starting conditions:

  Settings	Parameters
  Flow setup
  Flow rate	0.3 mL/min
  Running time	5 min
  Isocratic flow	80% mobile phase A (water with 0.1% formic acid)
  	20% mobile phase B (acetonitrile with 0.1% formic acid)
  MS setup
  Source temperature	450 °C
  Electrospray ionization (ESI) voltage	5,000 V (positive mode) and −4,000 V (negative mode)
  Ion source gas 1	55 psi
  Ion source gas 2	55 psi
  Curtain gas	30 psi
  Syringe speed	7 μL/min

MS signal optimization ● Timing 2.5–3.5 h

• 46 Load ~200 μL of the 5 μM standard solution into the syringe, put it into the MS instrument’s sample infusion holder and tighten it into the position.

  ! CAUTION Syringe needle is sharp.

• 47 Start the HPLC mobile phase flow and syringe liquid flow.

• 48 Start the MS signal acquisition in full scan to find the molecular ion. Keep the ion signal stable.

  ! CAUTION The metal surface of ESI source is very hot. The spray needle is very fragile (i.d. is 100 μm).

  ? TROUBLESHOOTING

• 49 Acquire MS/MS spectrum of target’s molecular ion ([M + H]⁺ or [M − H]⁻).

• 50 Choose fragment ions from MS/MS spectrum to set multiple reaction monitoring (MRM) transitions. Optimize the collision energy (CE). The parameters for the MRM transitions of the antibiotic controls used in this assay are provided in the following table:

  Compound	ESI polarity	Parent ion (m/z)	Fragment ion (m/z)	Collision energy (eV)
  Tetracycline	Positive	445.2	154.0	80
  			410.1	25
  			241.2	85
  Chloramphenicol	Negative	321.1	256.9	−15
  			151.8	−24
  			193.9	−20
  Ciprofloxacin	Positive	332.1	288.1	26
  			314.0	35
  			231.0	50
  Ampicillin	Positive	350.1	106.1	24
  			160.1	20
  			176.1	30
  Fusidic acid	Negative	515.3	455.3	−27
  			393.3	−31
  			221.0	−50
  Mupirocin	Positive	501.1	309.1	30
  			327.1	17
  			465.1	35
  Clindamycin	Positive	425.1	126.1	60
  			377.1	26
  			355.1	38
  Daptomycin	Positive	810.9	640.8	28
  			341.3	31
  			159.0	40
  Vancomycin	Positive	725.1	144.2	19
  			100.1	80
  			1143.2	50
  Rifampin	Positive	823.4	791.4	25
  			399.3	35
  			731.3	30
  Erythromycin	Positive	734.4	576.2	28
  			158.1	50
  			576.1	60
  Novobiocin	Negative	611.3	423.1	−30
  			568.2	−30
  			380.1	−65

  ▲ CRITICAL STEP Target compound is monitored with three MRM transitions: one for quantitation and the other two for qualifiers. Different target compounds may have very different concentrations present in the real samples; therefore, to ensure the linearity of signal response in the calibration curve, it is very important to choose the MRMs with a wide range of signal intensity (the highest one; the middle one, which has five to ten times less intensity than the highest one; and the low one, which has five to ten times less intensity than the middle one). The calibration curve and MRM transitions for tetracycline are shown in Fig. 5.

• 51 Choose the MRM with the low intensity to optimize ESI source parameters.

  These parameters are typically optimized using the following tuning ranges:

  ESI source parameters	General tuning ranges
  Source temperature	250~550 °C
  Gas 1	40~65 psi
  Gas 2	40~65 psi
  Curtain gas	25~40 psi
  ESI voltage	2,000~5,500 V

• 52 Repeat Steps 46–51 for other standard solutions.

Fig. 5 |. Calibration curve and multiple reaction monitoring.

a, Calibration curve of tetracycline generated from peak area analysis using Sciex software analysis. Calibration range (tetracycline m/z 445.2 → m/z 154.0, in ESI positive mode) is 100 ng/mL ~ 30,000 ng/mL. b, MS/MS spectrum of m/z 445.2 (tetracycline). Based on the fragments, we set the following MRMs: m/z 445.2 → m/z 154.0 is chosen to quantify target compound; m/z 445.2 → m/z 410.1 and m/z 445.2 → m/z 241. 2 are qualifiers.

HPLC condition optimization and LC-MS/MS method finalization ● Timing 1.5–2.5 h

• 53 Spike 1 μL of 20 individual stock solutions into 1 mL 20% methanol solution in water in one Eppendorf vial to get one 5 μM standard mixture solution.

  ▲ CRITICAL STEP Prior to making the standard mixture solution, make sure that the 20 target compounds have distinct MS/MS fragments (thus, distinct MRM transitions). If isomers are present that have identical MS/MS fragments, check whether they can be separated by HPLC.

  If these isomers can be separated by HPLC (with different retention times), run individual standard solutions (not a mixture solution) in Step 54 to verify the corresponding isomer’s retention time.

  If these isomers cannot be separated, it means these isomers cannot be differentiated by either HPLC (retention time), MS (molecular ion) or MS/MS (fragment ions). In this case, to quantify these isomer compounds, run their standard calibration solutions separately (prepare calibration solutions with a single isomer).

• 54 After inputting the finalized MS conditions into the acquisition method, start to acquire data by injecting the standard mixture solution with the HPLC autosampler. For the antibiotics included in Table 1, the LC separation is performed on an Agilent Zorbax SB-Aq column (4.6 × 50 mm, 5 μm) with the following method:

  Total time (min)	Flow rate	Mobile phase A (%)	Mobile phase B (%)
  0	0.3 mL/min	100	0
  3	0.3 mL/min	100	0
  10	0.3 mL/min	2	98
  14.5	0.3 mL/min	2	98
  14.6	0.3 mL/min	100	0
  20	0.3 mL/min	100	0

  The HPLC running time can be extended (i.e., maintain a long period of time with 100% mobile phase B) to ensure the target compounds’ elution.

• 55 Finalize the acquisition method with optimized MS and HPLC conditions.

Sample transfer and calibration solution preparation ● Timing 1.5–2 h

• 56 Transfer 100 μL sample into the HPLC sample vial (with glass insert).

• 57 Prepare calibration mixture solutions with six different concentrations (20 nM, 100 nM, 500 nM, 2,000 nM, 10,000 nM, 30,000 nM) for calibration curve buildup. The lower limit of quantitation of each compound is where the signal-to-noise ratio is 10.

  ▲ CRITICAL STEP Prior to making the standard mixture solution, make sure that the 20 target compounds have distinct MS/MS fragments (thus, distinct MRM transitions). If isomers are present that have identical MS/MS fragments, check whether they can be separated by HPLC. If these isomers can be separated by HPLC (with different retention times), run individual standard solutions (not a mixture solution) in Step 54 to verify the corresponding isomer’s retention time. If these isomers cannot be separated, it means these isomers cannot be differentiated by either HPLC (retention time), MS (molecular ion) or MS/MS (fragment ions). In this case, to quantify these isomer compounds, run their standard calibration solutions separately (prepare calibration solutions with a single isomer).

Data acquisition ● Timing 24–26 h

• 58 Run blank, samples and calibration solutions.

Data processing ● Timing 2–4 h

• 59 Use Sciex Analyst software (version 1.7.1) with peak area for quantitation. The calibration curve is automatically plotted, where the x axis is the standard concentrations and the y axis is the MRM transition’s mass spectrometry signal’s peak area counts. Under regression options, select linear fit and 1/x weighting factor.

Accumulation data workup ● Timing 15–30 min

• 60 Accumulation values (A) can be calculated using the formula below:

  $$A=\frac{(C\ast V)}{CFUs}$$

  where C is compound concentration in nanomolar, V is final volume in milliliters, CFUs is number of CFUs for the strain of interest, and A is accumulation in nmol/10^x CFUs.

• 61 Take the average of three biological replicates. Calculate the standard deviation using the function in Excel. Calculate standard error by dividing the standard deviation by the square root of the number of biological samples.

• 62 Analyze statistical significance of accumulation in Excel using a two-sample Welch’s t-test (one-tailed test, assuming unequal variance) relative to the negative antibiotic controls.

Troubleshooting

Troubleshooting advice can be found in Table 2.

Table 2 |.

Troubleshooting table

Step	Problem	Possible reason	Solution
11	Loss of cells after centrifugation	Cells may stick to the side of the plastic tubes used during centrifugation	Use a different type of tube
		If you are using a different strain of bacteria, cells may be smaller and may not pellet as well	Increase centrifugation time, and very carefully remove the supernatant with a pipette
			Use a different type of tube
24	Incorrect separation of oil and supernatant	Oil mix is not dense enough (either oil on top or a central layer of PBS with half oil on bottom, half on top)	Slightly increase the density of your oil formulation by increasing the amount of high temperature silicone oil
		Oil mix is too dense; solution appears as an emulsion, and bacteria may not be able to pellet through	Decrease the density of your oil formulation by increasing the amount of AR20 oil
30	Cells are not lysed by freeze–thaw	Cell lysis is dependent on the architecture of the cell envelope and can vary depending on the species or strain	Use a different lysis method, such as sonication or vortexing
35	Cells are not viable upon treatment with compound	Compound is rapidly bactericidal	Asses lower compound concentrations. If cells are viable at a lower concentration of your compound of interest, such as 25 μM, evaluate whether your antibiotic controls still show a statistical difference at this concentration. If using a much lower compound concentration, incubation timing may need to be optimized
48	Lack of stable ion current due to the presence of bubbles	HPLC mobile phase transfer tubing Infusion syringe	Purge the mobile phase Push the bubble out of the syringe
	Unexpected low signal	Dirty metal surface in the ESI source	Clean the metal source with methanol and water-soaked Kimwipes
		Partially blocked ESI spray needle	Clean the needle in an ultrasonic bath with methanol, or replace with a new needle
	Difficult to find an intense molecular ion	Standard solution concentration is too low Power polarity is not optimal	Try a standard solution with a higher concentration Switch power polarity (i.e., switch from positive mode to negative mode)

Timing

• Steps 1–7, prediction of high-accumulating compounds using eNTRyway: 10 min to 1 d (entirely dependent on number of compounds in batch)

• Steps 8–10, preparation of the bacterial suspension: ~1.5 d

• Steps 11–19, CFU determination assay: ~1.5 d; ~1 h hands-on time/replicate; stagger bacterial culture growth if performing multiple replicates in 1 d

• Steps 20–30, lysis validation: ~1.5 d, ~1.5 h hands-on time/replicate; stagger bacterial culture growth if performing multiple replicates in 1 d

• Steps 31–35, viability determination: ~1.5 d, ~1 h hands-on time/replicate; stagger bacterial culture growth if performing multiple replicates in 1 d

• Steps 36–42, accumulation assay: ~2 h hands-on time/replicate; stagger bacterial culture growth if performing multiple replicates in 1 d

• Steps 43–44, preparation of individual target compound solution (5 μM) for mass spectrometry (MS) signal optimization: 0.5–1 h

• Step 45, initial MS setting for manual optimization: 5 min

• Step 46–52, MS signal optimization: 2.5–3.5 h

• Steps 53–55, HPLC condition optimization and LC-MS/MS method finalization: 1.5–2.5 h

• Steps 56–57, sample transfer and calibration solution preparation: 1.5–2 h

• Step 58, data acquisition: 24–26 h

• Step 59, data processing: 2–4 h

• Steps 60–62, accumulation data workup: 15–30 min

Anticipated results

The web application eNTRyway can be used to predict if a compound will accumulate in E. coli¹⁹. Using eNTRyway in this manner allows compounds to be prioritized (for synthesis and evaluation) that are most likely to accumulate in E. coli. As shown in Extended Data Fig. 3, the output of eNTRyway shows either a red or green checkmark next to each of the three eNTRy rules categories (primary amine, low globularity (≤0.25), and low rotatable bonds (≤5)) for individual compounds submitted. Numerical values are also provided for globularity and rotatable bonds. For batch compound submissions, a csv file is available for download that contains the functional group classification and calculated values for globularity and rotatable bonds.

Prior to running accumulation in a new bacterial strain, CFU determination, lysis validation and cellular viability upon treatment with compound need to be performed. CFU determination should be consistent across three biological replicates. The positive control for lysis validation (i.e., freeze–thaw) should demonstrate no remaining viable colonies when plated. For test compounds, cells should retain ≥90% viability during the course of the experiment. A panel of approximately ten antibiotic controls should be evaluated against the bacterial strain of interest to ensure differentiation between positive and negative controls. Negative controls are expected to accumulate to ≤300 nmol/10¹² CFUs, whereas positive controls should exceed accumulation values of 300 nmol/10¹² CFUs, to a statistically significant extent relative to the negative controls. Accumulation values for antibiotic controls in E. coli MG1655 are shown in Fig. 4, and appropriate biological replicates should be performed to ensure appropriate classification. Examples of fully processed data are provided in the source data file for Fig. 4.

For accumulation experiments on novel compounds, nine new compounds can be evaluated with one positive control. The positive control antibiotic included in each batch should not deviate a statistically significant amount from previous batches; otherwise, the samples will need to be retested. Statistical significance of any new compounds tested should be determined using a two-tailed t-test comparing the three accumulation data points from testing in biological triplicate to the panel of negative antibiotic controls. This will provide a binary classification of either a ‘non-accumulator’ or ‘accumulator’, which aids in the identification of accumulation trends in compound classes and provides an improved understanding of the relationship between whole-cell compound activity and intracellular accumulation.

While we anticipate that the majority of compounds meeting the eNTRy rules will accumulate in E. coli, there may be some exceptions. Similarly, there is still much to be learned about compound accumulation trends, and some compounds predicted by eNTRyway to not accumulate may still achieve high levels of accumulation. For example, guanidiniums and pyridiniums are not yet recognized by eNTRyway, although it has been shown that these functional groups can sometimes facilitate accumulation²⁶.

Extended Data

Extended Data Fig. 1 |. Importance of amine steric accessibility and amphiphilic moment (vsurf_A).

a, Primary amines demonstrate higher accumulation than the mono-methyl amine, di-methyl amine, tri-methyl amine and amide comparisons. Primary amines on primary carbons also show improved accumulation over primary amines on secondary or tertiary carbons. b, Increasing amphiphilic moment trends with increasing accumulation. Accumulation is reported in nmol/10¹² CFUs. Data are taken from Richter et al.¹⁷ with permission.

Extended Data Fig. 2 |. Screenshot of the ‘input’ box for SMILES strings.

SMILES strings are canonicalized using Open Babel GUI.

Extended Data Fig. 3 |. Screenshots of the process of predicting accumulation using the web tool eNTRyway.

SMILES strings are submitted to eNTRyway, and compounds are prioritized for evaluation based on how closely they meet the eNTRy rules. In the example here, both ampicillin and 6-DNM-NH₃ meet all of the criteria and are predicted to accumulate, whereas penicillin and DNM are not. A portion of this figure is taken from Parker et al.¹⁹ with permission.

Data availability

The main data discussed in this protocol are available in the supporting primary research papers (https://doi.org/10.1038/nature22308 and https://doi.org/10.1038/s41564-019-0604-5). Source data are provided with this paper. Additional requests should be addressed to the corresponding authors.