Simple Ciprofloxacin Resistance Test and Determination of Minimal

Dec 20, 2017 - Resistant bacteria are spreading worldwide, which makes fast antibiotic susceptibility testing and determination of the minimal inhibit...
1 downloads 4 Views 855KB Size
Subscriber access provided by University of Florida | Smathers Libraries

Article

Simple ciprofloxacin resistance test and determination of minimal inhibitory concentration (MIC) within two hours using Raman spectroscopy Johanna Kirchhoff, Uwe Glaser, Jürgen A. Bohnert, Mathias W. Pletz, Juergen Popp, and Ute Neugebauer Anal. Chem., Just Accepted Manuscript • DOI: 10.1021/acs.analchem.7b03800 • Publication Date (Web): 20 Dec 2017 Downloaded from http://pubs.acs.org on December 20, 2017

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 free 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 accessible to all readers and 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.

Analytical Chemistry 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 16 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

Analytical Chemistry

Simple ciprofloxacin resistance test and determination of minimal inhibitory concentration (MIC) within two hours using Raman spectroscopy Johanna Kirchhoff (1,2), Uwe Glaser (1,2), Jürgen A. Bohnert (3), Mathias W. Pletz (1,4), Jürgen Popp (1,2,5,6), Ute Neugebauer* (1,2,5,6) Affiliations 1 Center for Sepsis Control and Care, Jena University Hospital, Germany 2 Leibniz Institute of Photonic Technology, Jena, Germany 3 Institute of Medical Microbiology, Jena University Hospital, Germany and Friedrich Loeffler Institute of Medical Microbiology, Greifswald University Hospital, Germany 4 Center for Infectious Diseases and Infection’s Control, Jena University Hospital, Germany 5 Institute of Physical Chemistry and Abbe Center of Photonics, University Jena, Germany 6 Research Campus InfectoGnostics Jena, Germany Corresponding Author *E-mail: [email protected]. Phone: +49 3641 9390900 Abstract Resistant bacteria are spreading worldwide, which makes fast antibiotic susceptibility testing and determination of the minimal inhibitory concentration (MIC) urgently necessary to select appropriate antibiotic therapy in time and by this improve patient’s outcome, and at the same time avoiding inappropriate treatment as well as the unnecessary use of broad spectrum antibiotics which would foster further spread of resistant bacteria. Here, a simple and fast Raman spectroscopy-based procedure is introduced to identify antimicrobial susceptibilities and determine the MIC within only two hours total analysis marking a huge time saving compared to established phenotypic methods nowadays used in diagnostics. Sample preparation is fast and easy as well as comparable to currently established test. The use of a dielectrophoresis chip allows automated collection of the bacteria in a micron-sized region for high-quality Raman measurement directly from bacterial suspensions. The new Raman spectroscopic MIC test was validated with 13 clinical E. coli isolates that show a broad range of ciprofloxacin resistance levels and were collected from patients with blood-stream infection. Micro-Raman spectroscopy was able to detect ciprofloxacin-induced changes in E. coli after only 90 minutes interaction time. Principal component analysis as well as a simple computed ratio of the Raman marker bands at 1458 cm-1 and 1485 cm-1 show a clear concentration-dependent effect. The MIC values determined with the new Raman method are in good agreement with MICs obtained by reference methods (broth microdilution, Vitek®-2, E-test) and can be used to provide a classification as sensitive, intermediate or resistant using the clinical breakpoints provided by EUCAST. Keywords: antimicrobial resistance test, AST, MIC, Escherichia coli, ciprofloxacin, Raman spectroscopy

1 ACS Paragon Plus Environment

Analytical 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 16

Introduction The minimal inhibitory concentration (MIC) of a drug denotes the lowest concentration which prevents visible bacterial growth. It is an important parameter determined in routine medical microbiological testing in order to choose the right drug for antibiotic treatment of a patient. The European Committee on Antimicrobial Susceptibility Testing (EUCAST) defines for each species – antibiotic pair clinical breakpoints which interpret the MIC and classify the strain as sensitive or resistant. A bacterium with a MIC above the breakpoint is considered to be resistant and below to be sensitive against the respective drug. In times of increasing antibiotic resistances worldwide, a fast determination of the antibiotic susceptibility is of utmost interest to select a tailored antibiotic therapy to efficiently treat the patient with narrow spectrum antibiotics. Furthermore, recent studies suggest that determination of the exact value of the MIC (not only the determination of “sensitive” or “resistant”) can help in therapy decisions as high MIC values within the susceptible range were found to correlate with an adverse outcome for patients with susceptible Gram-negative bacteria.1 The ‘gold standard method’ for MIC testing is still the broth micro dilution (BMD) test, in which a defined volume of liquid medium is complemented with a defined concentration of the antibiotic and incubated for 16 to 20 hours with the bacteria. The MIC is read as the lowest concentration within the test series that inhibits the visible growth of the bacterium. The BMD can be carried out in 96-well plates, requiring only low material costs, however, needs significant hands-on time and demands experienced staff. A modification of the agar disk diffusion test uses test stripes which are impregnated with a concentration gradient of an antimicrobial substance (e.g. E-test from bioMérieux® and MIC test strips from Liofilchem®) to allow for quantitative antibiotic susceptibility testing (AST). The MIC is read out from the concentration scale of the strip at the edge of the bacterial growth ellipse after 16 - 24 hours incubation. The test stripes are commercially available and simple to implement, but also are relative expensive and provide results only the next day. Automated instrument systems like the Vitek®-2 from bioMérieux or Phoenix from Becton Dickinson (BD) have been developed and are used nowadays in clinical routine. Such devices are practical and highly standardized, but not available for each laboratory and still provide the susceptibility results only after 6 to 13 hours for non-fastidious bacteria like E. coli. All these commonly used AST approaches detect phenotypical resistances by observing the bacterial growth under influence of the antimicrobial agent.2,3 Some alternative strategies that found their way into clinical practice directly search for known resistance genes using polymerase chain reaction (PCR)-based techniques. While those techniques allow identification and quantification of the pathogen as well as detection of several known resistance genes within only a few hours, sometimes even directly from the clinical samples, one assay is quite expensive allowing application of the technique only in selected cases. Furthermore, no MIC values can be provided and those methods only work for known gene sequences. Even more, the presence of such genes does not always correlate with the phenotypical characteristics of the pathogen.4 Novel concepts for fast AST include matrix-assisted laser desorption ionization time-of-flight mass spectrometry (MALDI-TOF MS), microfluidics, microarrays, real-time imaging of single cells, Raman spectroscopy, fluorescence-activated cell sorting (FACS), cell lysis-based approaches and whole-genome sequencing, amongst others.5 Each of these methods has its advantages and limitations and discussing them all, would go beyond the scope of this work. Only few of the new methods can analyze bacteria directly from body fluids, most require 2 ACS Paragon Plus Environment

Page 3 of 16 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

Analytical Chemistry

upstream sample processing. Highly promising are approaches with combined microfluidic sample management and optical and/or spectroscopic readout. The MIC of S. aureus positive blood cultures could be analyzed in a microfluidic assay with antibiotic gradient within two to five hours using single cell optical readout and image analysis6. Raman spectroscopy as a readout method holds the advantage that it allows not only for AST/MICtesting, but also for the identification of the analyzed pathogen in the same assay.7 In combination with dielectrophoretic sample manipulation it was successfully applied for identification of unknown pathogens from patient’s urine samples within 35 minutes total analysis time8 and for the detection of vancomycin resistances in enterococci9 as well as for ciprofloxacin resistances in E. coli 10, both within only 3.5 hours. Using silver substrates, surface-enhanced Raman spectroscopy (SERS) could serve to determine the MIC after two hours interaction time11. Our study presents a new Raman spectroscopy-based procedure for bacterial AST and MIC determination in less than two hours total analysis time. As illustrated in Scheme 1, this would mean a huge benefit in time compared to the gold standard method and clinical routine. Using a Raman spectroscopic dielectrophoresis (DEP) setup8 bacteria from suspension are directly captured in the focus of the Raman microscope. Only 90 minutes incubation time without pre-cultivation is sufficient, and minimal sample preparation of only ten minutes is needed for subsequent spectroscopic analysis. Based on a principal component analysis, spectral marker bands were defined that indicate an effect of the fluoroquinolone ciprofloxacin. A simple ratio of the intensities of the marker bands at 1458 cm-1 and 1485 cm-1 reflects growth or growth inhibition induced by the antimicrobial drug from which the MIC can be directly read out without complicated data analysis. The principle of the new method is first demonstrated for clinical E. coli isolates with therapeutically relevant ciprofloxacin concentrations. This pathogen – drug pair is of high clinical interest as the determination of MIC values against fluoroquinolone antibiotics was found to help to predict treatment responses of fluoroquinolone susceptible E. coli urinary tract infections12. Scheme 1: Temporal advantage of AST and MIC testing by the Raman spectroscopic assay

3 ACS Paragon Plus Environment

Analytical 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 16

Experimental section Bacterial samples and cultivation conditions Ciprofloxacin-sensitive Escherichia coli (E. coli AG100) and ciprofloxacin-resistant E. coli (E. coli 3-AG100) are part of the culture collection of the Institute of Medical Microbiology, Jena University Hospital, Germany. Multidrugresistant E. coli 3-AG100 is a third step mutant (gyrA, mar) derived from E. coli K12 derivative AG100 after repeated exposure to the fluoroquinolone ofloxacin13. Bacterial samples were stored at -80°C in cryo culture medium containing 20% glycerol. For each experiment a fresh overnight culture was prepared in liquid medium or on agar plate. Liquid cultures were cultivated in tryptic soy broth (TSB, Roth GmbH) at 37°C while shaking with 160 revolutions per minute (rpm) under ambient air conditions.

Patient isolates Clinical E. coli isolates were collected from blood of sepsis patients between 2012 and 2014 at the Jena University Hospital, Germany. In the routine diagnostic laboratory positive blood cultures were detected on a BACTEC FX instrument (BD Diagnostics, Heidelberg, Germany) and the pathogen was isolated by sub-cultivation on Columbia sheep blood agar (Oxoid, Wesel, Germany). Identification of the pathogen and antibiotic susceptibility testing (AST) of the confirmed E. coli isolate was carried out using VITEK®-2 system (bioMérieux, Nürtingen, Germany). Strains were maintained at -80°C in cryobank vials (Mast Diagnostica GmbH, Reinfeld, Germany). For Raman spectroscopic MIC determination 13 isolates were randomly selected.

Antibiotic and standard AST and MIC determination with broth microdilution, E-test and Vitek®-2 Ciprofloxacin (ciprofloxacin hydrochloride, AppliChem) stock solution was prepared and subsequently sterile filtrated using a 0.2 µm pore size syringe filter (Whatman, United Kingdom). Aliquots of ciprofloxacin stock solution (10 g/L) were frozen at -20°C and fresh thawed for each measurement day. To validate and compare the results of the new Raman-based MIC assay, standard AST and MIC determination is carried out with “gold standard method” such as broth microdilution, E-test and routine diagnostic using Vitek®2. Broth microdilution (BMD) testing was performed according to the recommendation of the EUCAST guided by a Nature Protocol for MIC determination14. For categorization as susceptible, intermediate or resistant the EUCAST breakpoint tables for interpretation of MICs and zone diameters were applied15. MIC determination by agar diffusion with Liofilchem® MIC Test Strips (Liofilchem, Italy) was performed following the manufacturer’s instruction. Further experimental details are given in the Supporting Information.

4 ACS Paragon Plus Environment

Page 5 of 16 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

Analytical Chemistry

Sample preparation for Raman MIC determination Bacterial cultures were grown in TSB for 14 to 20 hours and diluted for measuring the optical density (OD) at 600 nm using a spectrophotometer Cary 60 UV-Vis (Agilent). Cultivation flasks were prepared with TSB and final ciprofloxacin concentrations ranging from 0.008 µg/mL to 8 µg/mL (see Table 1, top row). Flasks were prewarmed until inoculation. Samples for Raman MIC determination were inoculated with the overnight culture to adjust a final inoculum of 5x105 CFU/mL according to standard MIC testing. The actual inoculum was verified by determining the CFU/mL by plating diluted suspension on TSB agar plates immediately after inoculation and counting colonies the next day. The inoculated test flasks were incubated for 90 minutes at 37°C and 160 rpm shaking in a shaking incubator (Infors HT Ecotron). After 90 minutes the samples were harvested by centrifuging for 5 minutes at 2680 × g (Universal 320R, Hettich). The supernatant was discarded and the bacteria pellet resuspended in 0.5× Dulbecco's phosphate buffered saline (PBS, Merck Biochrome). Following, the bacterial sample was transferred to a smaller reaction tube and washed twice with 0.5× PBS by centrifuging at 14549 × g for 1.5 minutes using an Eppendorf centrifuge 5418. The pellet was resuspended in 0.5× PBS for instant spectroscopic analysis. Additionally, as an internal control, 100 µL of the 90 minutes sample were taken for colony counting. Parallel to the colony counting test, another aliquot of 100 µL was given into a 96-well-plate to check the turbidity after 20-24 hours by naked eye. This served as an internal test reference to proof the dilution series and the bacterial growth for all Raman experiments. The conditions for those internal tests are close to the recommendation for AST by the EUCAST for the standard broth microdilution assay with the difference that Raman samples are prepared as a broth macrodilution assay (BMaD) prior to administration to the 96-well plate.

Raman spectroscopic measurement Raman spectroscopic analysis was performed with a confocal Raman microscope CRM 300 (WITec, Ulm) coupled to a spectrograph with a diffraction grating of 600 lines/mm and a back-illuminated CCD camera (DV401 BV, Andor, Belfast) with 1024×127 pixels cooled down to -60°C. A frequency doubled continuous wave Nd:YAG solidstate laser (Exelsior 532-60) with a wavelength of 532 nm served as excitation source with a resulting laser power on the sample of 15 mW. A 60× water immersion objective (NA 1.0, Nikon) focused the laser onto the sample and collected the 180° backscattered light, which was delivered via a multimode optical fiber (100 µm core diameter) to the spectrograph. The spectral resolution was about 4 cm-1. Fast and easy collection of bacteria from suspension in a micron-sized area for the Raman analysis was achieved by using a combined dielectrophoresis (DEP) - Raman setup. Further details are found in earlier publications 8,9 and in the Supporting Information. High quality Raman spectra from 10 to 20 accumulated bacteria were collected in fast time series mode with an integration time of 1 s per individual spectrum. Of each bacterial sample three time series with 100 spectra were recorded. For a regular Raman-MIC test with a range of 6 ciprofloxacin concentrations a total of 1800 spectra were recorded. The strains E. coli AG100 and E. coli 3-AG100 were analyzed in three independent repeats of the experiment at different days within a time frame of half a year. The spectroscopic MICs of the patient isolates were determined once or twice as shown in Table 1.

5 ACS Paragon Plus Environment

Analytical 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 6 of 16

Spectral preprocessing and data analysis Spectral preprocessing and statistical analysis was performed with software GNU R, version 3.0.3 [R Core Team (2014); http://www.R-project.org/.], using the packages “hyperspec” [Beleites C, Sergo V (2014); http://hyperspec.r-forge.r-project.org], and “ggplot2” [Wickham H (Springer New York 2009)]. Spectral preprocessing included cosmic spike removal, cutting and selecting the fingerprint region (650 - 1800 cm-1) for baseline correction with a 5th order polynomial function followed by vector normalization (l2-norm). Principal component analysis (PCA) was carried out to visualize spectral differences. The Raman intensity band ratios were calculated by selecting the maximum Raman intensity value at the wavenumber region from 1456 cm-1 to 1464 cm-1, and from 1481 cm-1 to 1489 cm-1, respectively. Mean spectra and other plots were designed using software Origin Pro (OriginLab Corporation) or Excel (Microsoft Corporation).

Results and Discussion MIC testing with laboratory strains E. coli AG100 and E. coli 3-AG100 To establish optimal test conditions and validate the reproducibility of the test, two well-characterized laboratory strains13,16 were used: the ciprofloxacin-sensitive E. coli AG100 and the ciprofloxacin-resistant E. coli 3-AG100 which was derived from E. coli AG 100 and displays a multiple-antibiotic resistance (Mar) phenotype. The ciprofloxacin MIC of the sensitive E. coli AG100 is 0.032 µg/mL and the MIC of E. coli 3-AG100 is 1 µg/mL ciprofloxacin. These MICs were determined by broth microdilution assay (BMD) which is the gold standard for MIC determination and recommended by the EUCAST. Furthermore, automated Vitek®-2 analysis used in routine diagnostics as well as commercial Liofilchem® MIC test strips (E-test) provided reference MIC values (Table 1). These established phenotypical antibiotic susceptibility test (AST) methods have read-out times of 16 to 20 hours incubation time. The Vitek®-2 susceptibility result was accomplished after 9 to 13 hours in the case of E. coli. Based on previous experiments10 we have chosen 90 minutes as read-out time for our new and fast spectroscopic method. At this time point the bacteria are adapted to the medium and are in exponential growth phase, thus, we observed a strong effect between growing and inhibited bacteria.

Concentration-dependent Raman spectra after 90 minutes of ciprofloxacin treatment Raman analysis of E. coli AG100 and E. coli 3-AG100 incubated for 90 minutes with defined ciprofloxacin concentrations displayed concentration dependent spectral differences. Clearly visible effects occur in the mean spectra at several wavenumber regions (Fig. 1). Those highly reproducible effects showed very similar patterns for E. coli AG100 and E. coli 3-AG100, just under different ciprofloxacin concentrations (around the MIC of the respective strain). The spectral region between 1440 cm-1 and 1520 cm-1 appeared particularly insightful, and therefore it is presented magnified by zooming closer into this region (Fig. 1b and d). The order of intensities in the bands at the wavenumbers 1458 cm-1 and 1485 cm-1 reflects the order of the administered ciprofloxacin concentration. The band at 1458 cm-1 increases with increasing ciprofloxacin concentration and the shoulder band at 1485 cm-1 decreases. 6 ACS Paragon Plus Environment

Page 7 of 16 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

Analytical Chemistry

Figure 1: The spectral ciprofloxacin effect after 90 minutes incubation with different ciprofloxacin concentrations. Preprocessed mean spectra of E. coli AG100 (a) and E. coli 3-AG100 (c) present the same pattern in the whole spectral fingerprint region. A clear order according to the ciprofloxacin concentration in the bands around 1458 cm-1 and 1485 cm-1 is visible by zooming in on a smaller wavenumber region (b, d). Each line represents the mean spectrum of three technical replicates with 3x100 Raman spectra (532 nm excitation, 1 s integration time per spectrum).

The concentration-dependent killing effect of ciprofloxacin after already 90 minutes was also confirmed and quantified using established microbiological methods, such as examining colony forming units (CFU) from the Raman samples at the next day as wells as turbidity read-out in a microdilution assay with the Raman samples. Details are giving in the Supporting Information (Text and Fig. S-1).

Defining ciprofloxacin effect marker bands based on principal component analysis As can be seen in Figure 1, systematic changes occur in the Raman spectra as response to the drug. Unsupervised statistical methods, such as principal component analysis (PCA), are ideally suited to visualize the 7 ACS Paragon Plus Environment

Analytical 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 8 of 16

internal structure of a data set and extract wavenumbers that are most associated with the observed variance (in our case due to the action of the drug). A good separation of the spectra taken from E. coli treated with ciprofloxacin concentrations above and below the MIC was observed along the first principal component (Fig. 2a,c), indicating that the first principal component (PC1) can be used to extract the important spectral features that are associated with the action of the drug. The loadings of PC1 for the two reference strains E. coli AG100 and E. coli 3-AG100 are depicted in Figure 2b and d. Both loadings show the same maximum and minimum peak values which are associated with efficient action of the drug. Maxima indicate vibrational bands that are more pronounced in the Raman spectra of bacteria with no or low ciprofloxacin concentrations (below MIC), minima highlight vibrational bands that are associated with efficient ciprofloxacin treatment (concentrations above MIC). Most of the positive bands (789, 815, 1101 1489, 1582 cm-1) can be assigned to nucleic acid vibrations10,17,18 and are associated with an actively growing, unperturbed bacterial culture in the exponential growth phase with high DNA and RNA synthesis19. Negative bands indicate C-H deformation vibrations (1452 cm1 ) as well as protein-associated vibrations (e.g. amid III and amide I vibrations around 1279 cm-1 and 1658 cm-1, respectively)17,18. These observed spectral changes are in good agreement with previous studies10,20 and reflect the action of the fluoroquinolone drug ciprofloxacin. Pure detection of the drug can be excluded by comparing the Raman bands of the ciprofloxacin spectrum (Fig. S-3) to the loadings of PC1. Ciprofloxacin binds to the gyrase-DNA-complex and by this prevents supercoiling of the DNA. This interference with DNA replication is bactericidal. An increase in intensity of vibrational bands at positions of minima in the PC1 loading (negative values in the first PC) reflects increasing ciprofloxacin concentrations while an increase in intensity of vibrational bands at wavenumber positions of the maxima reflects active growth of the bacteria with increased amounts of nucleic acids. For automated detection of resistances and determination of the MIC from the Raman spectral data, one could use the PCA scatter plots (Fig. 2a and c). However, this would require training a robust classification model which needs to be independent of the different strains. The close vicinity of the minimum at 1452 cm-1 and the maximum at 1489 cm-1 suggests that a simple intensity ratio of the vibrational bands in the Raman spectrum associated with inhibition and growth, respectively, might lead to a fast result.

8 ACS Paragon Plus Environment

Page 9 of 16 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

Analytical Chemistry

Figure 2: PCA scatter plot (a,b) and corresponding PCA-loading of the first principal component (PC1, c, d) of the interactions of E. coli AG100 (a, c) and 3-AG100 (b,d) treated with different ciprofloxacin concentrations around the MIC. Raman band intensity ratio (e,f) of the marker bands around 1458 cm-1 and 1485 cm-1 of the strains E. coli AG100 (e) and E. coli 3-AG100 (f) represent a ciprofloxacin concentration-dependent effect. The asterisk marks the MIC from reference methods (BMD and Vitek®-2), the red line indicates the threshold for reading out the Raman-MIC. Each column represents the mean ratio derived from up to three independent experiments and the corresponding standard deviation.

Intensity ratio of spectral marker bands indicates ciprofloxacin effect after 90 minutes The marker bands selected from PCA are in excellent agreement with the spectral changes already detected with the naked eye. In Figure 1b and d it can be seen how the vibrational band around 1458 cm-1 (corresponding to the minimum of 1452 cm-1 in PC1) increases and the shoulder band at 1485 cm -1 (corresponding to the maximum at 1489 cm-1 in PC1) disappears with increasing ciprofloxacin concentration. A ratio of the Raman intensities of both vibrational bands reflects the ciprofloxacin-induced growth inhibition as can be seen in Figure 2e,f. The good correlation with PC1 values is depicted in Figure S-4a. For defining a suitable threshold for the read-out of the MIC, reference MIC values as obtained from broth microdilution assay and automated Vitek®-2 analysis were used for the two laboratory strains, i.e. MIC = 0.032 µg/mL for E. coli AG100 and MIC = 1 µg/mL for E. coli 3-AG100. Sub-inhibitory ciprofloxacin concentrations result in a smaller ratio than untreated cultures. The ratio increases with increasing ciprofloxacin concentrations and reaches band ratios higher than 1 when the bacterial cultures are treated with ciprofloxacin concentrations above the MIC (Fig. 2e,f), indicating efficient growth inhibition. Thus, the threshold was set at a ratio of 1 for simple read-out of the MIC defining the Raman9 ACS Paragon Plus Environment

Analytical 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 10 of 16

MIC as the highest concentration which has a Raman band intensity ratio smaller than 1. The correlation of the spectral band ratio and the killing rate (Fig. S-1) as biological standard is depicted in Figure S-4b.

Validation of new Raman MIC testing method with clinical E. coli isolates The new Raman spectroscopic MIC testing method was applied to E. coli strains isolated from blood cultures of patients with blood stream infection. Using the EUCAST clinical breakpoints for Enterobacteriaceae and ciprofloxacin, that are sensitive (S) ≤ 0.25 µg/mL, resistant (R) > 0.5 mg/L, and intermediate (I) = 0.5 µg/mL15, a concentration matrix was set up comprising six ciprofloxacin concentrations in the sensitive range, plus the control without ciprofloxacin, and four concentrations in the resistant range (Table 1, top row). For the initial proof-of-concept study presented in this publication, at least six concentrations including the untreated control were tested to identify the Raman-MIC. The computed ratios of the Raman bands around 1458 cm-1 and 1485 cm-1 of the clinical isolates are presented in Figure S-2. The observed graph patterns are very similar to the ones depicted in Figure 2e,f for the two laboratory strains E. coli AG100 and E. coli 3-AG100: Fast growth, corresponding to an intense vibrational band at 1485 cm-1 and a small intensity ratio (< 1) is observed for low antibiotic concentrations for most strains. An increase of the computed band ratio is observed with increasing drug concentrations above a certain ciprofloxacin concentration. Four strains (E. coli 416, 544, 554 and 579) display a band ratio < 1 for all tested ciprofloxacin concentrations (up to at least 4 mg/L ciprofloxacin), indicating that the MIC was not yet reached. Those strains are classified as ciprofloxacin-resistant (Table 1). It is remarkable that some of the patient’s isolates exhibit a much lower intensity ratio as the two laboratory strains E. coli AG100 and E. coli 3-AG100 (Fig. 2e,f) under unperturbed growth (0 µg/mL ciprofloxacin, Fig. S-2). This originates from a more intense nucleic acid band at 1485 cm-1 in the Raman spectra of the patient’s isolates indicating different metabolic activity in the different strains. In some cases (E. coli 387, 405, 416, 422, 500) lower ratios for sub-inhibitory ciprofloxacin concentrations than the ratio of the untreated control were observed, suggesting increased bacterial growth at concentration below the MIC. This was indeed reported in the literature where a short-term benefit of the bacterial fitness was observed as a result of the bacterial SOS response on sub-lethal ciprofloxacin concentrations 21. Table 1 summarizes the results from the band ratios in the concentration matrix using the above defined threshold of 1 and gives the Raman-based resistance classification as well as the Raman-based MIC values for the laboratory strains as well as the 13 patient’s isolates. For a fast visualization of the susceptibility, a common color code is used to distinguish sensitive (green), intermediate (orange) and resistant (red).

10 ACS Paragon Plus Environment

Page 11 of 16 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

Analytical Chemistry

Table 1: Visualization of Raman spectroscopic MIC test results obtained after only 90 minutes incubation time and comparison to MIC values obtained from the gold standard reference methods broth microdilution (BMD), ) indicate marker band ratios smaller than one, red Vitek®-2 and E-Test after 16-20 hours. Green tickmarks ( crosses ( ) ratios higher than one. Ratio diagrams for the clinical isolates can be found in the Supporting Information (Fig. S-2). Ciprofloxacin concentration [mg/L] E. coli strain

0

0.008 0.016 0.032 0.064 0.125 0.25 sensitive

0.5 i

1

2 4 resistant

AG100

3-AG100 387 405 407 416 422 500 539 544 545 554 579 673 683

8

Raman MIC AST [mg/L] S 0.032 S 0.032 S 0.032 R 1 R 1 R 1 S 0.064 S 0.064 R 1 S ≤ 0.125 R ≥4 R ≥8 R 2 R 1 R 1 i 0.5 R 1 R ≥4 S 0.064 S 0.125 R ≥4 R ≥4 S 0.032 S 0.032 S 0.016

BMD VITEK®-2 E-test MIC MIC MIC [mg/L] [mg/L] [mg/L] 0.032

≤ 0.25

0.032

1

1

1

0.032

≤ 0.25

0.016

0.5 0.016

0.5 ≤ 0.25

0.25 0.016

1

1

1

1

1

1

0.5

0.5

0.5

0.25

≤ 0.25

0.125

≥ 32

≥4

≥ 32

0.125

≤ 0.25

0.064

≥ 32 ≥ 32 0.032

≥4 ≥4 ≤ 0.25

≥ 32 ≥ 32 0.016

0.032

≤ 0.25

0.016

Comparison of the Raman-spectroscopic MIC values with reference methods In order to judge and validate the Raman-spectroscopic MIC values, they were compared to the results from clinical routine diagnostics (Vitek®-2), the gold standard method broth microdilution assay (BMD), quantitative MIC test stripes (E-test) as well as the broth macrodilution assay (BMaD) which was performed parallel to the Raman experiments with the same samples but with read-out the next day (after 16-24h) (Table 1 and Table S1). A very good agreement of the Raman-spectroscopic results with the results from other, established testing methods is found. In comparison to the gold standard method BMD, a correct sensitivity classification was obtained for 20 out of 25 measurements (including 6 measurements of the reference strains): 6 out of 7 sensitive strains were correctly classified as sensitive; all six resistant strains were correctly classified as resistant. Two strains (E. coli 500 and 405) are categorized as resistant by Raman spectroscopy (Raman MIC = 1 mg/L) and intermediate by BMD test and Vitek®-2 analysis (0.5 mg/L). Another strain was identified as sensitive (E. coli 539) by BMD and Vitek®-2 analysis (0.25 mg/L) and as intermediate/resistant by the new Raman 11 ACS Paragon Plus Environment

Analytical 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 12 of 16

method (0.5-1 mg/L) (Table 1, Table S-1). For those three strains (5 measurements, Table 1), the discrepancy is only one dilution step which happens to be the breakpoint. Comparing the values of the Raman-spectroscopic MIC and the BMD MIC, minor discrepancies are observed for 5 strains which however, do not affect the susceptibility classification. Deviations of the MIC values of one dilution step are also observed among the established MIC methods (e.g. between E-test and BMD for E. coli 387, 539, 405, 545, 673 and 683). As the Vitek®-2 automat has a limited MIC range in the low ciprofloxacin range (MICs below 0.25 µg/mL cannot be exactly determined), an exact comparison for finer MIC levels is not possible for the six correctly identified sensitive strains. Within the accuracy range of the reference methods (Vitek®-2, BMD and E-test), the Raman results are all in good agreement with the established methods. For three strains (E. coli 544, 554, and 579), an exact comparison is not possible as the MIC of those strains is outside of the tested range of all methods (high-level resistance). In this way, there was no resistant strain which was not recognized as such by the new Raman-spectroscopic method and with the tested data set, no false-sensitive result was obtained. In almost all cases, the Raman method predicts a slighter higher MIC value than the established methods. One reason for this one-MIC-level deviation of the Raman-spectroscopic method could be how the threshold in the Raman band ratio is defined. For three E. coli strains (E. coli 545, 673, 683) the Raman threshold of 1 is almost reached at the reference MIC value (Fig. S-2). Other discrepancies in the MIC values that are observed when the MIC is around the breakpoint (e.g. as for E. coli 405, 416, 500), could probably be explained with the short readout time of 90 minutes and a slower response of the E. coli culture when already a few mutations causing resistance are present. Ciprofloxacin resistance in E. coli is usually caused by two major mechanisms: First, decreased binding affinity of the drug by target alterations and second, reduced concentration of the drug at the target site. Target alterations are caused by chromosomal mutations in the genes of DNA gyrase or topoisomerase IV and decrease the binding affinity of the fluoroquinolone drug. Decreased accumulation is caused by both, increased efflux of the drug by increased expression of the efflux pump AcrAB-TolC or reduced influx by mutations in the porin OmpF. A clinically relevant resistance requires a combination of these mechanisms induced by several mutations.22 We assume that the isolates with a MIC around the breakpoint are better equipped with resistance mechanisms, this could be, for example, more efflux pumps or restricted influx of the drug, resulting in a slower response of the culture to the treatment. Clinically, this can result in treatment failure followed by adverse outcomes for patients with susceptible bacteria with high MIC values within the susceptible range.1 Thus, the new Raman method presented in this work, which provides the result within less than 2 hours, predicts higher MIC values for those strains than established methods. However, this might be in better agreement with clinical outcome and thus, provide valuable information for the treating physician.

Conclusion This work demonstrates the high potential of a novel and rapid Raman spectroscopic approach for antibiotic susceptibility testing (AST) and determination of the minimal inhibitory concentration (MIC) of clinical isolates in less than 2 hours analysis time. A diverse collection of 13 clinical E. coli isolates that bear a variety of gene mutations resulting in different levels of resistance were characterized with the new Raman method and the results compared to established reference tests, such as the gold standard method broth microdilution assay 12 ACS Paragon Plus Environment

Page 13 of 16 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

Analytical Chemistry

(BMD), MIC test strips (Liofilchem®), and Vitek®-2 (bioMérieux) analysis, nowadays the method of choice in clinical routine. The Raman spectroscopic MIC after only 90 minutes interaction time was found to be in good concordance with the reference MIC tests which provide the result after 6-12 hours the earliest (Scheme 1). Thus, an enormous saving in time could be achieved with the new phenotypic method. The new approach does not require pre-culturing. A pure bacterial culture as is typically applied in the routine diagnostic tests is directly inoculated into the antibiotic test tubes and for 90 minutes exposed to the drug. Sample preparation and handson-time is minimal and does not take more than ten minutes. Due to the accumulation of bacteria in defined measurement position by the dielectrophoretic force on the DEP chip within short times (several seconds), high quality Raman spectra with good signal-to-noise ratio can be acquired within seconds. The very simple and robust analysis algorithm using the computed ratio of two adjacent Raman bands would enable an almost online visualization of the Raman MIC result. The successful algorithm was demonstrated exemplarily for the fluoroquinolone drug ciprofloxacin and E. coli. In the future, we will extend the algorithm and include other antibiotics and pathogens in the analysis. While this proof-of-concept study was carried out on a planar quadrupole DEP chip, implementation into a complex microfluidic device 10 which allows for automatization of the sample preparation steps and reduction of the required sample volumes due to miniaturization, is easily possible. The ultimate limit of detection for such Raman-based methods is a single, individual bacterial cell 23. Furthermore, the new Raman-MIC algorithm was established on the same experimental set-up that enabled pathogen identification directly from patient’s urine samples within only 35 minutes total analysis time8. Thus, a combination of cultivation-independent pathogen identification and fast phenotypic resistance testing without any additional external information directly from patient’s material can become possible in the future. With this potential, the newly proposed Raman spectroscopic MIC assay could become an integral part of a novel diagnostic tool that can provide test results within a total diagnosis time of < 3 hours starting from patients’ body fluid to the susceptibility result of the responsible pathogen.

Acknowledgement We thank the staff in the routine diagnostic laboratory of the Institute for Medical Microbiology at the Jena University Hospital for the collaboration and for the performance of the Vitek®-2 analyses. We further thank Marcel Dahms for help with generation of graphs. Financial support by the BMBF via the Integrated Research and Treatment Center “Center for Sepsis Control and Care” (CSCC, FKZ 01EO1502), via the Research Campus InfectoGnostics (FKZ 13GW0096F), and by the Carl Zeiss Foundation is highly acknowledged. Furthermore, the project was supported by the Free State of Thuringia (FKZ 2015 FGI 0011 and 2016 FGI 0010) with co-financing from the European Union within the European Regional Development Fund (EFRE).

13 ACS Paragon Plus Environment

Analytical 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 14 of 16

Supporting Information The Supporting Information is available free of charge on the ACS Publications website at DOI: Experimental details on standard AST and MIC determination and DEP chips; Supplementary results (Characterization of the ciprofloxacin effect after 90 minutes with standard biological methods); Visualization of concentration-dependent killing of E. coli AG100 and E. coli 3-AG100 (Fig. S-1); Intensity ratios of the Raman bands at 1458 cm-1 and 1485 cm-1 of the clinical isolates (Fig. S-2); Raman spectrum of ciprofloxacin hydrochloride (Fig. S-3); Correlation plots of spectral band ratio and PC1 as well as killing rate (Fig. S-4); Summary table of ciprofloxacin MIC values obtained by Vitek®-2, broth microdilution, E-test, Raman spectroscopy and broth macrodilution (Table S-1).

References (1) Falagas, M. E.; Tansarli, G. S.; Rafailidis, P. I.; Kapaskelis, A.; Vardakas, K. Z. Antimicrob. Agents Chemother. 2012, 56, 4214-4222. (2) Balouiri, M.; Sadiki, M.; Ibnsouda, S. K. J. Pharm. Ana. 2016, 6, 71-79. (3) Jorgensen, J. H.; Ferraro, M. J. Clin. Infect. Dis. 2009, 49, 1749-1755. (4) Pulido, M. R.; Garcia-Quintanilla, M.; Martin-Pena, R.; Cisneros, J. M.; McConnell, M. J. J. Antimicrob. Chemother 2013, 68, 2710-2717. (5) van Belkum, A.; Dunne, W. M. J. Clin. Microbiol. 2013, 51, 2018-2024. (6) Malmberg, C.; Yuen, P.; Spaak, J.; Cars, O.; Tängdén, T.; Lagerbäck, P. PLoS ONE 2016, 11. (7) Stöckel, S.; Kirchhoff, J.; Neugebauer, U.; Rösch, P.; Popp, J. J. Raman. Spectrosc. 2016, 47, 89-109. (8) Schröder, U.-C.; Ramoji, A.; Glaser, U.; Sachse, S.; Leiterer, C.; Csaki, A.; Hübner, U.; Fritzsche, W.; Pfister, W.; Bauer, M.; Popp, J.; Neugebauer, U. Anal. Chem. 2013, 85, 10717-10724. (9) Schröder, U.-C.; Beleites, C.; Assmann, C.; Glaser, U.; Hübner, U.; Pfister, W.; Fritzsche, W.; Popp, J.; Neugebauer, U. Sci. Rep. 2015, 5, 8217. (10) Schröder, U.-C.; Kirchhoff, J.; Hübner, U.; Mayer, G.; Glaser, U.; Henkel, T.; Pfister, W.; Fritzsche, W.; Popp, J.; Neugebauer, U. J. Biophot. 2017. (11) Liu, C. Y.; Han, Y. Y.; Shih, P. H.; Lian, W. N.; Wang, H. H.; Lin, C. H.; Hsueh, P. R.; Wang, J. K.; Wang, Y. L. Sci. Rep. 2016, 6, 23375. (12) Rattanaumpawan, P.; Nachamkin, I.; Bilker, W. B.; Roy, J. A.; Metlay, J. P.; Zaoutis, T. E.; Lautenbach, E. Ann. Clin. Microbiol. Antimicrob. 2017, 16, 25. (13) Kern, W. V.; Oethinger, M.; Jellen-Ritter, A. S.; Levy, S. B. Antimicrob. Agents Chemother. 2000, 44, 814-820. (14) Wiegand, I.; Hilpert, K.; Hancock, R. E. W. Nat. Protoc. 2008, 3, 163-175. (15) EUCAST. 2017. (16) Oethinger, M.; Kern, W. V.; Jellen-Ritter, A. S.; McMurry, L. M.; Levy, S. B. Antimicrob. Agents Chemother. 2000, 44, 10-13. (17) Athamneh, A. I.; Alajlouni, R. A.; Wallace, R. S.; Seleem, M. N.; Senger, R. S. Antimicrob. Agents Chemother. 2014, 58, 1302-1314. (18) Notingher, I.; Verrier, S.; Romanska, H.; Bishop, A. E.; Polak, J. M.; Hench, L. L. Spectroscopy 2002, 16. (19) Neugebauer, U.; Schmid, U.; Baumann, K.; Ziebuhr, W.; Kozitskaya, S.; Deckert, V.; Schmitt, M.; Popp, J. Chemphyschem 2007, 8, 124-137. (20) Neugebauer, U.; Schmid, U.; Baumann, K.; Ziebuhr, W.; Kozitskaya, S.; Holzgrabe, U.; Schmitt, M.; Popp, J. J. Phys. Chem. A 2007, 111, 2898-2906. (21) Torres-Barceló, C.; Kojadinovic, M.; Moxon, R.; MacLean, R. C. Proc. Biol. Sci. 2015, 282, 20150885. 14 ACS Paragon Plus Environment

Page 15 of 16 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

Analytical Chemistry

(22) Wiedemann, B.; Heisig, A.; Heisig, P. Antibiotics 2014, 3, 341-352. (23) Lorenz, B.; Wichmann, C.; Stöckel, S.; Rösch, P.; Popp, J. Trends Microbiol. 2017, 25, 413-424.

15 ACS Paragon Plus Environment

Analytical 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 16 of 16

For TOC only:

16 ACS Paragon Plus Environment