Subscriber access provided by UNIVERSITY OF TOLEDO LIBRARIES
B: Biophysics; Physical Chemistry of Biological Systems and Biomolecules
Measuring Drug-Induced Changes in Metabolite Populations of Live Bacteria: Real Time Analysis by Raman Spectroscopy Paul R. Carey, Grant R. Whitmer, Michael J Yoon, Michael N Lombardo, Marianne Pusztai-Carey, Hossein Heidari-Torkabadi, and Tao Che J. Phys. Chem. B, Just Accepted Manuscript • DOI: 10.1021/acs.jpcb.8b03279 • Publication Date (Web): 24 May 2018 Downloaded from http://pubs.acs.org on May 24, 2018
Just Accepted “Just Accepted” manuscripts have been peer-reviewed and accepted for publication. They are posted online prior to technical editing, formatting for publication and author proofing. The American Chemical Society provides “Just Accepted” as a service to the research community to expedite the dissemination of scientific material as soon as possible after acceptance. “Just Accepted” manuscripts appear in full in PDF format accompanied by an HTML abstract. “Just Accepted” manuscripts have been fully peer reviewed, but should not be considered the official version of record. They are citable by the Digital Object Identifier (DOI®). “Just Accepted” is an optional service offered to authors. Therefore, the “Just Accepted” Web site may not include all articles that will be published in the journal. After a manuscript is technically edited and formatted, it will be removed from the “Just Accepted” Web site and published as an ASAP article. Note that technical editing may introduce minor changes to the manuscript text and/or graphics which could affect content, and all legal disclaimers and ethical guidelines that apply to the journal pertain. ACS cannot be held responsible for errors or consequences arising from the use of information contained in these “Just Accepted” manuscripts.
is published by the American Chemical Society. 1155 Sixteenth Street N.W., Washington, DC 20036 Published by American Chemical Society. Copyright © American Chemical Society. However, no copyright claim is made to original U.S. Government works, or works produced by employees of any Commonwealth realm Crown government in the course of their duties.
Page 1 of 13 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60
The Journal of Physical Chemistry
Measuring Drug-Induced Changes in Metabolite Populations of Live Bacteria: Real Time Analysis by Raman Spectroscopy Paul R. Carey*1, Grant R. Whitmer1, Michael J. Yoon1, Michael N. Lombardo2, Marianne Pusztai-Carey1, Hossein Heidari-Torkabadi1†, Tao Che1‡ AUTHOR ADDRESS 1
Department of Biochemistry, Case Western Reserve University School of Medicine, 10900 Euclid Ave, Cleveland OH, 44106
2
Department of Pharmaceutical Sciences, University of Connecticut, 69 North Eagleville Road, Storrs CT, 06269
ABSTRACT: Raman difference spectroscopy is shown to provide a wealth of molecular detail on changes within bacterial cells caused by infusion of antibiotics or hydrogen peroxide. E. coli strains paired with chloramphenicol, DHFR propargylbased inhibitors, meropenem, or hydrogen peroxide provide details of the depletion of protein and nucleic acid populations in real time. Additionally, other reproducible Raman features appear and are attributed to changes in cell metabolite populations. An initial candidate for one of the metabolites involves population increases of citrate, an intermediate within the TCA cycle. This is supported by the observation that a strain of E. coli without the ability to synthesize citrate, gltA, lacks an intense feature in the Raman difference spectrum that has been ascribed to citrate. The methodology for obtaining the Raman data involves infusing the drug into live cells, then washing, freezing, and finally lyophilizing the cells. The freeze-dried cells are then examined under a Raman microscope. The difference spectra, [cells treated with drug] minus [cells without treatment] are time-dependent and can yield population kinetics for intracellular species in vivo. There is a strong resemblance between the Raman difference spectra of E. coli cells treated with meropenem and those treated with hydrogen peroxide.
INTRODUCTION At the clinical level, treating bacterial infections is becoming increasingly problematic due to the rise of bacterial resistance 1, 2 coupled with the slow pace of development of novel antibiotics. Although new compounds are being reported 3, 4 and considerable effort is being expended on development, progress is hindered by lack of knowledge in the molecular details of drug–cell interactions. For example, the development of rules for compound-membrane penetration are hindered by the lack of availability of facile methods to generate kinetics for compound infusion and efflux. 5, 6 Also while there are ongoing studies on how intracellular drugs perturb cell metabolism7, 8, there is not a wealth of information for live cells. In two recent accounts we developed a Raman-based protocol that, in favorable cases, allows us to follow drug uptake in live bacteria and to measure the population of drug molecules in the cell. 9, 10 The protocol exposes bacteria, under growth conditions, to known concentrations of drug molecules. At set times, aliquots are filtered to remove free compound, washed, frozen and freeze-dried. Thus, at those times, the cell-drug complexes are immortalized. The freeze dried cells yield high quality Raman spectra. By subtracting the Raman spectra ([cells + drug] – [untreated cells]) the difference spectra have contributions from the spectrum of the drug (and its reaction products). In a related experiment, by creating a standard plot of drug intensity vs. concentration, the intracellular populations of drug molecules could be calculated and followed as a function of infusion time. In many of our early Raman difference spectra, we saw both positive and negative peaks that were not associated with the infused drug. They were difficult to assign. However, further experiments over the past three years have enabled us to assign many “new peaks” with a high degree of probability. Many are due to changes in populations of major cell components, e.g. nucleic acids and proteins. Other features are beginning to be understood, and they are due to substantial changes in metabolite populations. Moreover, the non-drug features in the difference spectra often appear when bands from the drug within the cell are too weak to be observed. Although other workers have obtained Raman spectra from bacteria,11, 12 even down to the level of single cells,13 we believe the protocol described here provides the first data from live
ACS Paragon Plus Environment
The Journal of Physical Chemistry 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60
Page 2 of 13
drug-cell complexes during infusion. The complexes are immortalized by flash freezing and freeze-drying. Difference spectroscopy then provides Raman data at each time point during the infusion process. Three classes of drugs infusing into E. coli are investigated here; chloramphenicol, (an inhibitor of the 50S ribosome subunit), propargyl-linked inhibitors of dihydrofolate reductase (DHFR) carrying the diaminopyrimidine pharmacaphore, and the carbapenem meropenem that acylates and blocks the activity of penicillin-binding proteins. In addition, we tested the effect of exogenous hydrogen peroxide on the cells. This set of experiments was undertaken because we have earlier, unpublished data that show that H2O2 treatment brings about the appearance of “new bands” that are similar to those seen during antibiotic treatment. We now compare, in detail, the products of the reactions of intracellular antibiotics with the effects of hydrogen peroxide. The ability to follow populations of intracellular molecules, or their effects on metabolism, in real time has the potential to provide a rational ranking of drug efficacy and may lead to a better understanding, at the molecular level, of the causes of cell mortality. A major thrust of the present paper is to suggest that the peaks originating from changes in cell metabolism offer a novel and powerful means of probing metabolism and for assessing drug efficacy. METHODS E. coli WT K-12 strain was utilized for most experiments described. Strains BW25113 and NR 698 were utilized for the experiments shown in Figure 2; ATCC 25922 for Figures 1, 3, 4, and 5; and MG 1655 for Figure 6. The exceptions were for experiments with the propargyl-based compound and the “citrate minus” strain described below. The inhibitor UCP1038 was gift of the late Dr. Amy Anderson and Dr. Dennis Wright. For the tests of the 1038 UCP compound, MICs were determined according to the Clinical and Laboratory Standards Institute's Performance Standards for Antimicrobial Susceptibility Testing. The microdilution broth assay was performed using the E. coli strains BW25113 from the Coli Genetic Stock Center at Yale University and NR698, generously provided by Dr. Thomas Silhavy of Princeton University and Dr. Daniel Kahne of Harvard University14, 15, 16. Assays were run at an inoculum of 5 x 105 CFU/mL in Isosensitest Broth (Oxoid). The MIC was defined as the lowest concentration of inhibitor to visually inhibit growth following an 18 hour incubation at 37 °C. The “citrate knockout” E. coli strain MG1655 plus gltA::cam gal-76::Tn10, was kindly provided by Dr. James Imlay of the University of Illinois. Protocol for freeze dried samples. Typically, 100 mLs of cells were grown in Mueller Hinton broth at 37°C until OD600 reached 0.6 to 0.8. Six sterile Falcon tubes were each loaded with 10 mLs of cells and were kept at 37°C in a shaker. The drug being utilized was made up in a stock solution, typically at 10 mM. X µM of stock solution was added to each of five tubes to give the required final concentration of drug. At 5 separate time points (2 to 60 minutes) each 10 mL drug/cell solution was vacuum filtered through a 0.22 µm Durapore filter (Millipore), in order to remove broth components and cell fragments. The cells were washed with isotonic NaCl then the filter was immediately transferred to a Falcon tube and frozen in liquid nitrogen. The tube was then placed on a freeze dryer over-night. For Raman spectroscopy the freeze dried cells were placed on aluminum foil which was seated on the stage of a Raman microscope17. Approximately 40 mW of Kr+ 647.1 nm irradiation was used to generate the Raman spectrum of the cells. Raman spectra were collected on the CCD detector for 100 x 1 sec exposures. This procedure generates the Raman spectrum from all the cells in the focal volume of the laser beam. Data collection and processing were performed using HoloGRAMS and GRAMS/AI7 software (ThermoGalactic, Salem, NH). Raman difference spectra were obtained using GRAMS/AI’s spectral subtract feature using the formula [Raman spectrum of treated cells at time t] minus [Raman spectrum of untreated cells at time zero]. Raman peak assignments, mostly for the nucleic acid and protein components of the cells, are listed in Tables 1 and 2 in Supplementary Information. They are based on several decades of Raman analysis of the purified components. Some key papers for the assignments are listed in reference 18.18 The Raman data shown in this publication were collected between 2015 and March 2018. In each set of experiments typically 4 data sets were collected from untreated freeze dried cells and from cells treated with antibiotic, at each time point. The sets with the highest signal-to-noise ratio were used to generate Raman difference spectra [Raman spectrum treated cells] minus [Raman spectrum untreated cells]. For each spectrum the 1450 cm-1 band was taken as an internal intensity standard; this is due to the deformation modes of every -CH2- group in each cell sample. Care was taken to exactly superimpose the 1450 peaks from treated and untreated cells, this process is illustrated in Figure 1. Each experiment was repeated 4 times, sometimes by different group members in the list of authors. For experiments where an accurate baseline can be drawn, e.g. Figs.1 and 2 intensity reproducibility is in the range 2-4%, with less accuracy and Figures 3-6. For this reason, we only discuss intense positive or negative peaks in the latter.
ACS Paragon Plus Environment
2
Page 3 of 13 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60
The Journal of Physical Chemistry RESULTS CHLORAMPHENICOL
Figure 1. Raman spectra of freeze dried untreated E. coli (green) and E. coli treated with 3.2 µg/ml chloramphenicol for 15 minutes [in growth medium at 37 C (red). The difference trace is shown in red below.
Chloramphenicol inhibits protein synthesis by binding to the 23S rRNA of the 50S ribosomal subunit. It has a MIC of 16 µg/mL against WT E. coli. The top traces in Figure 1 are a superposition of the Raman spectra of freeze-dried E. coli untreated (green) and freeze-dried cells that prior to freeze-drying were exposed for 15 minutes to 3.2 µg/mL of chloramphenicol in growth medium at 37 C (red). The peak at 1448 cm-1 is due to vibrational modes from all the CH2 groups in the cells and is taken as an internal intensity standard. The spectral envelopes in Figure 1 are made up of the contributions from the Raman spectra of all the molecules in each bacterium. Thus, many thousands of vibrations give rise to the observed profile. However, a few groups can be picked out because they are good “Raman scatterers” and have high Raman intensity. For example, the peak at 1572 cm-1 is mostly from guanine ring vibrations (overlapped by a minor contribution from an adenine mode). Also, the sharp feature at 1002 cm-1 is a phenyl ring vibration, likely having major contributions from the phenylalanine groups in proteins. A second strategy to reduce the complexity is to subtract the spectra of two closely matched samples where one sample differs from the other by a known perturbation. This is illustrated in Figure 1 where the spectrum of the untreated cells is subtracted from that of the cells exposed to the drug, and gives rise to the lower signal to noise ratio difference trace in Figure 1. We focus on the “negative” peaks since they have the highest signal/noise ratio. The negative traces at 1668 and 1241 cm-1 are due to the protein amide modes known as amide I and amide III, respectively. The sharp negative peak at 1002 cm-1 is assigned to phenylalanine residues embedded in the cells’ proteins. Thus, one immediate conclusion from the difference spectrum is that treating the cells with chloramphenicol for 15 minutes has resulted in a loss of intracellular protein. Ap-
ACS Paragon Plus Environment
3
The Journal of Physical Chemistry 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60
Page 4 of 13
proximating the intensities of amide I and III traces in the two superimposed spectra by peak heights gives the estimate that the total protein concentration in cells has fallen by 11±2%. This conclusion is consonant with chloramphenicol’s role in blocking protein synthesis. The second suite of relatively intense negative bands are seen at 1572, 1482(shoulder), 1332, and 724 cm-1, which are pri-
Figure 2. Adapted from A. Raman spectrum of the DHFR inhibitor UCP 1038 in DMSO. Solvent peaks are near 1040, 707 and -1 676 cm . B. Raman difference spectrum of UCP 1038 inside WT E. coli cells after 10 minutes of infusion. The intra-cell drug 6 17 population is ≈ 10 per cell . C. Raman difference spectrum of UCP 1038 inside E. coli NR 698 following 2±1 minutes expo6 sure to infusion. This E. coli strain has a severely compromised membrane . D. Raman difference spectrum of UCP 1038 inside E. coli NR 698 following 60 minutes exposure.
marily due to G, G, A and A ring modes, respectively. In addition, the 806 and 780 cm-1 negative peaks have major contributions from RNA and DNA backbone phosphodiester stretches, respectively 19, 20. Measuring the changes in relative intensities of the RNA and DNA backbone stretches in the super-imposed spectra shows that the RNA and DNA contents of the cells have decreased 22 ± 2% upon chloramphenicol treatment. Although we can readily detect the effects of chloramphenicol diffusing into the bacteria we do not see a clear signal from the intra-cell drug itself. The drug does not have a very intense Raman spectrum and this, possibly combined with a low population of intracellular molecules, may preclude detection. PROPARGYL-LINKED COMPOUNDS AS DIHYDROFOLATE REDUCTASE INHIBITORS Dihydrofolate reductase (DHFR) has long been a target for inhibition in the battle against pathogenic bacterial cells, and also against mammalian cells in anti-proliferative therapy. The groups of Drs. Amy Anderson (deceased) and Dennis Wright (University of Connecticut) designed, synthesized and assayed a large number of novel inhibitors 21, 22 of the kind seen in Figure 2. They feature the classic pharmacaphore ”diaminopyrimidine” (ring 1) bound to a second ring system via a propargyl link. As a result of their highly conjugated nature, these compounds possess an intense (but non-resonance)
ACS Paragon Plus Environment
4
Page 5 of 13 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60
The Journal of Physical Chemistry
Raman spectrum and in previously published work, we were able to detect these compounds within bacteria and measure their populations, interactions and obtain preliminary kinetics for their uptake10. Figure 2 compares the Raman data for UCP 1038 (Figure 2) in DMSO (trace A) with the Raman difference data for the drug inside WT E. coli cells after 10 minutes of infusion in growth medium containing 10 µg/mL of the drug (trace B). The spectra are essentially identical. However, when UCP 1038 is taken up for a period of 2±1 minutes by E.coli NR 698, which has a severely compromised outer membrane14, 15, 16, the difference spectrum is greatly changed (trace C). The three UCP 1038 ring modes can still be identified near 1562, 1597 and 1632 cm-1, but these are joined by intense “new bands” near 1687, 1285, 1216 and 772 cm-1 (as well as several less intense features). To summarize, UCP 1038 penetrates WT E.coli with only minor perturbations to its Raman spectrum, and no other major changes are seen in the Raman difference spectrum. It has a MIC of 5µg/mL. For E.coli NR698, however, intense nondrug Raman bands appear within the time resolution of our experiments of 2 minutes. This combination has a MIC
