Combined DielectrophoresisâRaman Setup for the Classification of

Subscriber access provided by George Mason University Libraries & VIVA (Virtual Library of Virginia)

Article

A combined dielectrophoresis-Raman setup for the classification of pathogens recovered from the urinary tract Ulrich-Christian Schröder, Anuradha Ramoji, Uwe Glaser, Svea Sachse, Christian Leiterer, Andrea Cszaki, Uwe Huebner, Wolfgang Fritsche, Wolfgang Pfister, Michael Bauer, Juergen Popp, and Ute Neugebauer Anal. Chem., Just Accepted Manuscript • DOI: 10.1021/ac4021616 • Publication Date (Web): 14 Oct 2013 Downloaded from http://pubs.acs.org on October 21, 2013

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 28

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

A combined dielectrophoresis-Raman setup for the classification of pathogens recovered from the urinary tract Ulrich-Christian Schröder;1, 2 Anuradha Ramoji;1, 2 Uwe Glaser;1, 2 Svea Sachse;3 Christian Leiterer;1 Andrea Csaki;1 Uwe Hübner;1 Wolfgang Fritzsche;1 Wolfgang Pfister;3 Michael Bauer;2 Jürgen Popp;1, 2, 4 Ute Neugebauer1, 2 * 1

Institute of Photonic Technology, Albert-Einstein-Straße 9, 07745 Jena, Germany

2

Center for Sepsis Control and Care, Jena University Hospital, Erlanger Allee 101, 07747 Jena,

Germany 3

Institute of Medical Microbiology, Jena University Hospital, Erlanger Allee 101, 07747 Jena,

Germany 4

Institute of Physical Chemistry and Abbe Center of Photonics, Friedrich Schiller University

Jena, Germany *corresponding author: Ute Neugebauer,

Email: [email protected],

Tel. No- +49-3641-9323364, Fax No. +49-3641-9323382

ACS Paragon Plus Environment

1


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 28

ABSTRACT Rapid and effective methods of pathogen identifications are of major interest in clinical microbiological analysis to administer timely tailored antibiotic therapy. Raman spectroscopy as a label-free, culture-independent optical method is suitable to identify even single bacteria. However, the low bacteria concentration in body fluids makes it difficult to detect their characteristic molecular fingerprint directly in suspension. Therefore, in this study Raman spectroscopy is combined with dielectrophoresis, which enables the direct translational manipulation of bacteria in suspensions with spatial non-uniform electrical fields so as to perform specific Raman spectroscopic characterization. A quadrupole electrode design is used to capture bacteria directly from fluids in well-defined micro sized regions. With live/dead fluorescence viability staining it is verified, that the bacteria survive this procedure for the relevant range of field strengths. The dielectrophoretic enrichment of bacteria allows for obtaining high quality Raman spectra in dilute suspensions with an integration time of only one second. As proof-of-principle study the setup was tested with Escherichia coli and Enterococcus faecalis, two bacterial strains that are commonly encountered in urinary tract infections. Furthermore to verify the potential for dealing with real world samples, pathogens from patient’s urine have been analyzed. With the additional help of multivariate statistical analysis a robust classification model could be built and allowed the classification of those two strains within a few minutes. In contrast the standard microbiological diagnostics base on very time-consuming cultivation tests. This setup holds the potential to reduce the crucial parameter diagnosis time by orders of magnitude.


2

Page 3 of 28

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


INTRODUCTION Urinary tract infections (UTIs) are common bacterial infections encountered in clinical practice covering a spectrum from uncomplicated UTIs in young women to life-threatening health care associated sepsis. The lifetime possibility of each woman developing an UTI is estimated to be above 40–60%.1 It is estimated that about 150 million UTIs occur yearly on a global basis causing illness costs about 6 billion dollar.2 Bacteria typically appearing in UTIs treated in hospital are Escherichia coli (causative organism in 50% of the cases), Klebsiella spp. (14 %), Enterococcus faecalis (10%), staphylococci (6%), other coliforms (4%) and Pseudomonas aeruginosa (3%).3 A significant bacteriuria in properly collected mid-stream samples of urine is defined by growth of 105 colony forming units (cfu) per milliliter of a single type of bacteria.4-7 The pathogens are in general diagnosed by urine culture; however, these conventional microbiological methods are time consuming and labor-intensive 2, 6. Identification of the causing pathogen is usually available to the treating clinician after 24 hours the earliest. In addition, it is problematic that bacteria most notably in health care associated infections show increasing resistances to antibiotics due to extensive drug use in the last decades.8 The determination of the antibiogram, i. e. the identification of antibiotic susceptibility of the pathogen for different drugs, can take another day. However, in order to initiate tailored therapies medical diagnosis needs to be rapid, inexpensive and sensitive, and pathogens as well as their antimicrobial resistances identified reliable. 2, 8 Raman spectroscopy is a technique which provides detailed and specific information on the molecular structure of the investigated specimen and has emerged in the last years as an extremely powerful method to identify cells and bacteria.9-15 The utility of Raman spectroscopy is further enhanced by a minimized sample handling which reduces costs and time:16 only small


3


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 28

sample volumes are required and labeling-steps are omitted. Therefore, a direct use of body fluids from patients without any cultivation steps becomes possible. However, this requires the retention and collection of the pathogens in micro-sized regions where the Raman signal is excited and recorded. Dielectrophoresis (DEP) is an excellent tool to fulfill these requirements by capturing dielectric nano- and microsized particles 17 such as bacteria directly in fluids 18-20 and enables their Raman spectroscopic investigations. DEP is the translational movement of neutral matter caused by polarization effects in a non-uniform electric field 21 and has been successfully applied to biosensors, cell therapeutics, drug discovery, medical diagnostics, microfluidics, nanoassembly and particle filtration within the past years.22-24 The time averaged dielectrophoretic force acting on a dielectric particle in a medium is given in dipole approximation by:25 〉 Γ ∙ ε ∙ Ref ω ∙ E ² 〈F

(1)

Γ describes the geometry of the particle and is proportional to the volume of the particle. For spherical particles Γ is proportional to the cubic radius . denotes the electric field. The DEP² which depends on the geometry of the force is proportional to the spatial gradient E electrodes and the applied voltage. !"# $ is the Clausisus-Mosotti (CM) factor, which describes the polarisation of the particle with respect to the medium. It depends on the dielectric *

functions %&∗ $ of particle and medium: %&∗ $ %& ( ) ,+ , with the circular frequency $ of the applied field, the dielectric constants %& and the conductivities -& ; the index j refers to particle and medium, i is the imaginary unit with ) . (1.


4

Page 5 of 28

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


Depending on the CM factor particles experience two regimes of DEP force. If 01!"# $ 2 0, the particles are attracted towards the electrodes, where the electric field maximizes. This is called positive DEP (pDEP). If Ref ω 4 0, a repelling force from the electrodes pushes the particles towards regions, where the electric field minimizes. This is called negative DEP (nDEP). The access to these different two regimes depends strongly on the conductivity of the surrounding medium. For high conductivities (σ > 1 S/m) as they occur under physiological conditions only the nDEP regime is accessible.26 Up to now only few studies have been reported to make use of dielectrophoretic collection of particles for Raman spectroscopic characterization. Most of these studies concern the investigation of nanostructured materials. So far one publication by Cheng et al. in 2010 concerns the use of DEP and Raman for the analysis of bacteria. 27 However, in that study the Raman effect was surface enhanced (SERS) with nanostructured gold due to weak signals. SERS enhances the Raman signal by several orders of magnitude, but implies additional surface selection rules, which lead to the observation of some differences in the vibrational modes and might also be hampered by impurities on the metal substrate.28 In this study DEP is combined with conventional micro-Raman spectroscopy in an integrated set-up to achieve a fast and easy characterization of pathogens from urine of patients with only minimal sample preparation as depicted in scheme 1: urine from patients with significant bacteriuria (> 105 bacteria of one species per milliliter) is after a short pre-treatment step administered to the chip where DEP is used to directly trap the bacteria in a defined location in space. This is achieved within a few minutes. Live/dead fluorescence viability staining can show that the organisms survive this procedure. Then, micro-Raman spectroscopy is employed to obtain spectroscopic fingerprints from the collected organisms within seconds; and finally,


5


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 28

multivariate statistical analysis techniques are employed to differentiate between the objects of interest based on their specific Raman spectra. The study focuses on Escherichia coli and Enterococcus faecalis as an example for Gram negative as well as Gram positive bacteria participating as major cause of urinary tract infections.

MATERIAL AND METHODS Bacterial samples Escherichia coli ATCC® 25922TM and Enterococcus faecalis ATCC® 29212TM are cultivated on CASO agar (ROTH GmbH) overnight at 37 °C. Five independent culture batches have been prepared, one batch per day of measurement. Bacteria are collected from one colony, centrifuged with a relative centrifugal force of 11,500 g for five minutes and the pellets are washed two times with 0.5x phosphate buffered saline (PBS, ROTH GmbH) and resuspended in same concentrated PBS. The optical density served as a measure of bacteria concentration. The absorption at 600 nm is recorded with an Agilent Cary 60 UV-Vis spectrometer and adjusted to be between 0.4 and 0.5, which corresponds to about 108 cells/ml. The conductivity of the buffer solution is measured with a Metler Toledo Seven Easy conductivity meter and is about 1 S/m. Three anonymised urine samples (2-8 ml) from three different patients with single pathogen UTIs (> 105 cells/ml, either E. coli or E. faecalis) have been provided by the Institute of Medical Microbiology and are used for testing the combined DEP-Raman device. Microbiological analysis based on cultivation and species identification has been performed for comparison by biochemical testing in Vitek II system (bioMérieux, Marcy L′Ètoile, France). For Raman analysis, the urine samples are filtered using syringe filters with 2.7 µm pore diameters ACS Paragon Plus Environment

6

Page 7 of 28

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


(Whatman Int. Ltd) to remove bigger particles such as leukocytes or epithelial cells. The filtered urine is treated like the bacteria samples from cultivation: it is centrifuged, the pellet washed twice with 0.5x PBS and finally resuspended in 0.5x PBS.

Fabrication of the DEP chips The fabrication of the DEP chips starts on wafer-level using a 4” fused-silica wafer (thickness: 1 mm, diameter 4 inch). The 4” wafer layout contains an arrangement of 60 DEP chips with a size of 10 x 10 mm², which allows variations of geometrical parameters within one fabrication run. After a pre-cleaning routine using ‘Caro's acid’ (peroxymonosulfuric acid), two thin layers of resist were spun onto the surface of the 4’’ fused silica wafer: first 200 nm ARP617.06 (Allresist GmbH) followed by 300 nm ZEP520A (Zeon Chemicals). Each resist was baked for 10 min at 200°C on a hotplate. On top of the electron beam resist a 10 nm thick conducting gold layer was evaporated to avoid charging effects during the electron beam exposure. The desired DEP design was written on the chip with the shaped beam writer SB350 OS (Vistec Electron Beam GmbH) using a beam energy of 50 keV. After the exposure the gold layer was removed and the resist was developed 60 s in ZED N50-developer and subsequently 60 s in AR600-50developer and rinsed for 20 s in isopropyl alcohol (IPA). Afterwards, a 3 nm thick titanium undercoating and a 100 nm thick gold film were deposited on top of the wafer by means of thermal evaporation under normal incidence. The lift-off -procedure was performed by an overnight-soaking in a resist remover bath followed by ultrasonic cleaning. Finally, the wafer was sliced into the individual DEP chips of 10 mm x 10 mm size each.


7


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 28

Dielectrophoresis equipment An alternating current (AC) voltage between 4 and 20 Vpp with a frequency of 1 MHz for dielectrophoretic trapping is generated with a HAMEG HMF 2550 arbitrary function generator. Tungsten wire (99.95% metals basis) for contacting the electrodes is obtained from VWR International. Dielectrophoretic force calculations have been carried out with software COMSOL Multiphysics V4.0 (COMSOL, Inc.) and the CM-factor has been plotted with Mathematica V8.0 (Wolfram Research). Fluorescence measurements Live/dead viability fluorescence staining has been carried out as reported previously 29 with a dye composition of 1 µl Syto9 (5 mMol) in 400 µl propidium iodide (75 µMol) (Invitrogen). The final concentrations on the chip are 0.3 µMol Syto9 and 1.9 µMol propidium iodide. Fluorescence signals have been detected with an AXIO Imager A1 microscope equipped with Filter Set 77 HE and camera Axio Cam (Carl Zeiss). Micro-Raman setup Raman spectroscopic analysis has been performed with a CRM 300 WITec micro-Raman setup equipped with a 600 lines/mm grating. A frequency-doubled cw Nd-YAG laser beam with an excitation wavelength of 532 nm and a power of 15 mW before passing the objective is focused on the sample via a 63x Nikon water immersion objective with a numerical aperture of 1.0. The 180° backscattered light is detected by a back illuminated ANDOR DV401 BV CCD camera with 1024 x 127 pixels cooled down to -60 °C.


8

Page 9 of 28

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


Data analysis Statistical analysis is performed with programming language ‘R’ V2.15.0 30 with inbuilt packages and using the package ‘hyperSpec’ 31. Spectra are pre-processed: baseline corrected with a 5th order polynomial, truncated to the spectral region between 600 and 1750 cm-1 and vector normalized. Afterwards, a principal component analysis (PCA) is carried out to reduce the dimension of the data set. The first ten principal components are subjected to a linear discriminant analysis (LDA) to build a classification model for the two species (trained with 600 spectra per species). This model is tested with independent data sets from three different culture batches (totally 1000 spectra per species) measured on three different days as well as with three urine samples from three patients with UTIs (in total 300 spectra). RESULTS AND DISCUSSION Design of a dielectrophoresis chip The aim is to design a dielectrophoresis chip which can capture bacteria from dilute suspensions and retain them stable in space so that high-quality Raman spectra can be recorded from the captured bacteria. In order to obtain as little as possible background signal in the Raman spectra, fused silica was used as chip substrate. It fulfills the general requirement to be insulating between the DEP electrodes and does not show significant Raman signals in the fingerprint region of the bacteria between 600 and 1800 cm-1. Urine samples have a conductivity ranging from 1 S/m to 3 S/m as was determined in measurements of urine samples from several healthy volunteers under different conditions and several patients suffering from UTI at different severities. Under these physiological conditions the real part of the Clausius-Mosotti-factor will always be negative and therefore, only negative ACS Paragon Plus Environment

9


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 28

dielectrophoretic trapping is possible. This requires a chip design with highly symmetric microsized electrodes so that the repelling forces from the electrodes will collect and capture the bacteria in one spot in the center. Therefore, on top of an insulating substrate a quadrupole electrode design is patterned with a distance of d = 40 µm between opposite positioned electrodes (figure 1b). The boundaries of the electrodes follow in the x-y plane the equations 5 678 . 6 9/2. , which are received by solving the two dimensional Laplace equation ∆=8, 5 0 (= is the potential field) with a polynomial ansatz of second order.32 20 µm and 100 µm distances between the electrodes also have been tested, but turned out to be not suitable. The chips with only 20 µm distance offer a too small collecting volume. The chips with 100 µm spacing generate a much smaller electric field gradient and therefore, require high voltages to capture the bacteria. Dielectrophoretic capture of bacteria The chip with a distance of 40 µm between opposite electrodes was used in all subsequent measurements. A droplet of 200 µl 0.5x PBS is deposited on top of the chip. Then, the nonuniform electrical field for nDEP is generated by applying an alternating peak-to-peak voltage of about 4 Vpp and 1 MHz to the electrodes via tungsten wire tips, which do not show interfering electrochemical reactions. They are arranged so that neighboring electrodes always have opposite potential. The resulting force then points towards the center of the array (figure 1c). Now, 10 µl of bacteria suspension are injected into the droplet, so that a particle stream passes the center and high efficient capturing in the center between the electrodes is achieved within a few minutes as can be seen in figure 2a where the collected bacteria form a compact, three dimensional cloud. Individual bacteria cannot be identified with the used objective. However, if


10

Page 11 of 28

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 voltage is switched off, the bacteria start to move around freely and individual bacteria can be spotted again. Furthermore, if bacteria somehow would have been destroyed in a way that extracellular debris would appear, the fluorescence viability test should have shown red dots from propidium iodide intercalating in released DNA strands and fragments. This was not the case as can be clearly seen in figure 3 of the submitted manuscript. Our experiments show that the negative dielectrophoretic trapping of bacteria in 0.5x PBS with a conductivity of about 1 S/m works best at frequencies in the order of magnitude of 1 MHz for E. coli, which are Gram negative ellipsoidal bacteria (figure 2c) as well as for E. faecalis, which are spherical Gram positive bacteria often appearing in chains of three and more bacteria. At frequencies below 500 kHz counter ion relaxation processes start to influence the DEP response 23

and below 100 kHz DEP trapping can hardly be observed in our experiments, most likely due

to competing nonlinear electro-kinetic effects. In solutions with low salt concentrations AC electro-osmosis occurs in this low frequency range 33. At frequencies higher than 10 MHz the collection gets increasingly weaker. This is in agreement with theoretical prediction (figure 2b) carried out with a Two-Shell Model 26. In this model the bacterium is considered as an ellipsoid containing the three concentric ellipsoidal main regions cytoplasm, membrane and wall with permittivities εcyt, εmem, εwall and conductivities σcyt, σmem, σwall (figure 2c). The related CM factor is plotted in figure 2b for E. coli and a medium conductivity of 1 S/m; the dielectric cell parameters are taken from Castellarnau et al.: εcyt = 49.8 ε0, σcyt = 0.48 S/m, εmem = 9.8 ε0, σmem = 259 · 10-6 S/m, εwall = 78ε0, σwall = 58 · 10-3 S/m, εm = 80 ε0; geometric parameters (figure 2c) are a = 1.5 µm, b = 0.75 µm, dmem = 8 nm, dwall = 50 nm 26. It can be seen that at high frequencies the


11


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 28

CM-factor is less intense indicating a weaker dielectrophoretic force. This force maximizes in the region below 1 MHz. The applied voltage has been adjusted to 4 Vpp. At higher voltages additional fluid movement starts to occur. According to the energy-balance equation (? ∙ ∆@ 〈- ∙ . 〉 (k is the thermal conductivity of the fluid, Tx, y, z is the temperature distribution) power is transformed into Joule heating and leads to permittivity and conductivity gradients, causing additional electrothermal induced fluid flows 34. In figure 3 an example at 14 Vpp is depicted, where these fluid whirls can clearly be observed. In the quadrupole structure they appear as four whirls symmetrically arranged around the center and also the bacteria in the center show rapid shaking. With the presented setup it is also possible to capture bacteria directly from urine samples (spiked urine samples from healthy volunteers as well as from patient’s urine samples) without any previous washing step. The conductivity of patient’s urine can vary over a high range: between 1 S/m and up to 3 S/m. When the voltage was increased up to 12 Vpp for a conductivity of about 3 S/m the power loss due to Joule heating was counterbalanced and the bacteria could be captured successfully on the chip. Viability measurements (see below) could prove that the selected parameters did not kill the bacteria. The typically yellow color of the urine has the disadvantage to cause a fluorescence background in the Raman measurements. It is therefore advisable to include a medium exchange/washing step. In the current design, this step is performed outside the chip and the washed bacteria are injected onto the chip in PBS buffer. The additional time needed for that washing procedure is only 15 minutes, which is negligible, compared to time consuming over-night cultivation steps. Further research is carried out to implement a fluidic management system so that the medium exchange can be carried out directly on the chip.


12

Page 13 of 28

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


Viability measurements If the DEP-Raman approach is supposed to be applied for classification and in later applications also for antibiotic susceptibility testing it has to be proven that the procedure does not harm the bacteria and the viability of the bacteria is not affected by the applied electric field. Cell death is mostly accompanied by a loss of cell membrane potential and a loss in integrity in the cell membrane. 35 To our knowledge a study concerning membrane damages caused by DEP has not been carried out so far. Although cell death could be measured by means of DEP 35, our method of choice to follow the influence of DEP itself is by using dead/live staining assays which are very established in biological research. The organisms are exposed to a two-color fluorescence assay consisting of the dyes Syto9 and propidium iodide. Syto9 is a green fluorescent nucleic acid staining dye labeling all bacteria, those with damaged and those with intact cell membranes. Propidium iodide is a red fluorescent nucleic acid staining dye, which only penetrates membrane damaged bacteria, causing a reduction in Syto9 fluorescence and outshines the green fluorescence.36 Between the deposition of the PBS droplet on top of the DEP chip and the injection of the bacteria as described above, 5 µl of the fluorescence staining assay are added to the PBS droplet. Subsequently, the bacteria are collected by means of dielectrophoresis. Three different peak to peak voltages have been investigated at a frequency of 1 MHz: 8 Vpp, 14 Vpp and 20 Vpp. 8 Vpp has been chosen instead of 4 Vpp because the dye increases the conductivity, and a higher voltage is needed to counterbalance power loss. After 1 h capturing and holding bacteria under DEP force the fluorescence measurements are performed with an integration time of 5 s. The results are shown in figure 3: At 8 Vpp green fluorescence indicates that all bacteria have survived the DEP procedure over one hour; at 14 Vpp clearly thermal induced fluid motion can be observed as


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

Page 14 of 28

mentioned before but the bacteria stay alive as can be seen from the bright green color; at 20 Vpp after one hour the red fluorescence dye has already penetrated the cell membrane and the red fluorescence shows that such high voltages damage the bacteria. At 20 Vpp the electrothermal whirls cannot be seen with the chosen integration time, because the cell movement is too fast at the four positions around the center. From these observations it can be concluded that the applied voltage of 4 Vpp does not alter the viability of the bacteria during the experiments.

Raman spectroscopic characterization of captured bacteria After having trapped bacteria as described above in a cloud of about 10-12 µm in diameter (figure 2a), the Raman excitation laser beam is focused inside the droplet on the dielectrophoretically collected bacteria (figure 1a). Raman spectra are recorded in the fast time series mode with a frequency of one spectrum per second and an amount of 100 spectra per sample. In total ten cultivation samples (two per day of measurement) have been investigated. Spectra with clearly visible and different Raman bands between the two species within the fingerprint region are even obtained within one second and proof how powerful the demonstrated setup works. Typical average Raman spectra from two days of measurement are depicted in figure 4a together with their standard deviation. The same spectra (600 spectra per species) have been used to train a classification model. In the Raman spectra characteristic spectral features reported previously for E. coli and E. faecalis are visible:9, 37 The most intense Raman bands appearing in the spectra from both species can be assigned to contributions from nucleic acids at 788 cm-1 (PO2- stretching, C, T), 1093 cm-1 (PO2- stretching) and 1578 cm-1 (ring stretching of guanine and adenine), proteins and peptides with the phenylalanine ring vibration at 1004 cm-1,


14

Page 15 of 28

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 amide III band at 1253 cm-1, and amide I at 1658 cm-1. CH contributions can be seen at 1341 cm-1 (CH deformation) and 1452 cm-1 (CH deformation).38-40 Statistical classification model for the discrimination of E. coli and E. faecalis To test the potential for bacterial classification, which implies the detection of small differences between the captured bacteria, a classification model is set up based on the multivariate statistical analysis methods principal component analysis (PCA) in combination with linear discriminant analysis (LDA). The data set is split into two parts: Data taken from two different days of measurement are used for training the model (600 spectra per species); data taken from three other days (giving 1000 spectra per species) are used for testing the model. The classification results of the PCA/LDA model are presented in figure 5a. In the left part of figure 5a each point represents one spectrum (indicated with the running index on the x-axis). The sign of the first linear discriminant (LD1, plotted on the y-axis) defines the predicted identity of the bacteria based on their spectra. Towards the positive LD1 axis Raman spectra of E. faecalis can found, while spectra of E. coli arrange along the negative LD1 axis. On the right part of figure 5a the relative abundance of each LD1 value is summed up and its distribution is depicted. The two different bacteria species can be very well separated with respect to LD1 and are well predicted without any misclassification. The coefficients of LD1 highlight the spectral features which lead to the differentiation of the two species (figure 5b): Main differences in the Raman spectra are identified around 748 cm-1, 1093 cm-1, 1241 cm-1, 1414 cm-1 and 1578 cm-1, which can tentatively be assigned to nucleic acid contributions, but also around 1004 cm-1, 1177 cm-1 and 1610 cm-1 which could be due to proteins.38-40 This is not surprising as nucleic acids and proteins


15


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 28

constitute the main building blocks (per dry weight) in a bacterial cell and are expected to vary between the two different species. Analysis of patient’s urine samples In order to proof the suitability for applied analysis, the presented setup was tested with “real world samples”, i. e. with urine samples from three different patients with UTIs. After a washing step, the bacteria suspensions were injected onto the integrated dielectrophoresis-Raman chip. Bacteria were captured in the electric field within a few minutes and Raman spectra were recorded with an accumulation time of 1s. Figure 4b shows the resulting mean Raman spectra taken from 100 spectra per patient sample. The same spectral features are exhibited as in the Raman spectra of bacteria from culture (figure 4a). The individual Raman spectra from the bacteria from the patient samples were subjected to classification using the PCA/LDA model described above. The results are plotted in the scatter plot and the bar plot in figure 5a: the cause of the UTI for two patients was identified to be E. coli, and for the third patient E. faecalis. Only one single spectrum out of 100 spectra from the first patient was misclassified to be E. faecalis, while the majority vote for the bacteria from that patient would give E. coli. For comparison and as gold standard method the same patient samples have also been diagnosed with standard microbiological cultivation tests. The results are in very good agreement with the Raman spectroscopic measurements: The first two patient samples were found to contain E. coli, the third sample to contain E. faecalis. One of the patients with an E. coli infection had been medicated with the antibiotic vancomycin, the other patients had been untreated at the time of sample collection. This shows that a correct identification of pathogens


16

Page 17 of 28

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


from patient’s urine samples can be achieved with the presented setup and a correct differentiation between the two common bacteria found in urinary tract infections, E. coli and E. faecalis, can be provided with a great time benefit (35 minutes) compared to classical microbiological methods (> 24h). However, it can be seen in figure 5a that the individual data points (spectra) are distributed closer to the zero position of LD1 in the LDA space. This can be explained with minor spectral differences between cultured strains and patient isolates: relative intensities variations of a few vibrational bands can be seen in figure 4. The origin of these variations could be strain and clonal diversity within the species (E. coli and E. faecalis) which was not determined in our experiments. Both, Raman spectroscopic analyses as well as microbiological analyses were only performed on species level, as is usually done in routine analysis. It is very likely that the specific bacterial strains present in the patients are not the same as were used for cultivation in the laboratory. Furthermore, it is well known, that bacteria adapt to their environment, such as chemical compositions, local temperature and illumination 41 and respond appropriately for example by switching phenotype or behavior. In relationship to Raman spectroscopy Meisel et al. have shown that Raman spectra of the same species living in different environments show variations.15 Therefore, the vibrational signature of bacteria cultivated under standardized laboratory conditions is expected to be slightly different from that of bacteria which originate directly from the urinary tract of patients. Nevertheless, the classification model is robust enough to differentiate between two prevalent pathogens encountered in urinary tract infections despite the fact that they originate from different patients with different medical history.


17


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 18 of 28

For a reliable application of this technology for the identification of various pathogens causing the UTI in the clinical routine a larger database with more bacterial strains also including wellcharacterized patient’s isolates has to be set-up as was previously done for air-borne bacteria 42.

CONCLUSIONS and OUTLOOK The present study demonstrates a powerful combination of dielectrophoresis with Raman spectroscopy, which provides a rapid, sensitive and label-free, but also easy and inexpensive method to characterize pathogens from body fluids as was shown exemplarily for urine samples from patients with UTIs. The DEP chips presented in this manuscript were produced with electron beam lithography which is a relative expensive method, but offers high flexibility for each new electrode design. Now, that the optimal design was found, the much cheaper photolithography can be used where one mask is produced and afterwards a high number of identical DEP chips can be produced at much lower costs. In order to further reduce the costs, reusing the chips would be possible. In order to avoid cross contamination from different patients, it is suggested after cleaning the chip with buffer to further treat it with oxygen plasma. This will remove all possible organic material. Based on the experience with other devices with electrodes manufactured in the clean room a reuse of 10 – 20 times should be possible. The set-up also holds the potential to investigate also other body fluids, such as ascites or cerebrospinal liquid. Gram positive and Gram negative bacteria from bacterial culture as well as from real world patient’s urine samples have been captured in suspension on the chip within a micro-sized region and then successfully differentiated with respect to their Raman signature in the fingerprint region. In total, the procedure took just a few minutes as no time-consuming


18

Page 19 of 28

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


cultivation step was necessary. Compared to standard microbiological identification methods, which in contrast need several hours till days, these promising results are crucial for reducing time in the analysis of pathogens and also offer the possibility of identifying bacteria which are hard or not all cultivable. In the presented chip design a short (max. 15 minutes) sample preparation has to be carried out prior administration of the patient’s sample onto the chip. Nevertheless, the presented method holds the potential to integrate those pre-treatment steps also on the chip: the dielectrophoretic force strongly depends on the radius of the particles as the factor Γ in equation (1) depends on the volume of the particle. That means another DEP design prior to the collection chip can separate the large mammalian cells from the small bacteria and make the filtering step dispensable. With state of the art microfluidic systems it is furthermore possible to integrate a fluidic management system on the chip which can carry out the fluid exchange in a very short time. The study further paves the way for future research on determining antibiotic resistances directly from body fluids in a fast and reliable manner ultimately leading to improvements in the time to adequate tailored antibiotic therapies.

ACKNOWLEDGMENT The financial support of the BMBF via the Integrated Research and Treatment Center "Center for Sepsis Control and Care" (CSCC, FKZ 01EO1002) and the TMWFK via MikroPLEX (FKZ PE113-1) is highly acknowledged. We thank Claudia Beleites for support in statistical analysis and Jürgen Kunert for CAD drawings.


19


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 20 of 28

REFERENCES (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)

Salvatore, S.; Salvatore, S.; Cattoni, E.; Siesto, G.; Serati, M.; Sorice, P.; Torella, M. European Journal of Obstetrics & Gynecology and Reproductive Biology 2011, 156, 131-136. Stamm, W. E.; Norrby, S. R. Journal of Infectious Diseases 2001, 183, S1-S4. Goodacre, R.; Timmins, E. M.; Burton, R.; Kaderbhai, N.; Woodward, A. M.; Kell, D. B.; Rooney, P. J. Microbiology-Uk 1998, 144, 1157-1170. Grabe, M.; T.E. Bjerklund-Johansen; H. Botto; B. Wullt; M. Çek; K.G. Naber; R.S. Pickard; P. Tenke; Wagenlehner, F.; Urology, E. A. o., Ed.; European Association of Urology "Guidelines on urological infections", 2012. Kass, E. H. J Urol 2002, 167, 1016-1019; discussion 1019-1021. Bitsori, M.; Galanakis, E. Expert Rev Anti Infect Ther 2012, 10, 1153-1164. Lee, J. B.; Neild, G. H. Medicine 2007, 35, 423-428. Andersson, D. I.; Hughes, D. Nature Reviews Microbiology 2010, 8, 260-271. Harz, A.; Rosch, P.; Popp, J. Cytometry Part A 2009, 75A, 104-113. Harz, M.; Kiehntopf, M.; Stockel, S.; Rosch, P.; Straube, E.; Deufel, T.; Popp, J. Journal of Biophotonics 2009, 2, 70-80. Krafft, C.; Dietzek, B.; Popp, J. Analyst 2009, 134, 1046-1057. Krafft, C.; Steiner, G.; Beleites, C.; Salzer, R. Journal of Biophotonics 2009, 2, 13-28. Schmitt, M.; Popp, J. Journal of Raman Spectroscopy 2006, 37, 20-28. Stockel, S.; Meisel, S.; Elschner, M.; Rosch, P.; Popp, J. Analytical Chemistry 2012, 84, 9873-9880. Meisel, S.; Stockel, S.; Elschner, M.; Melzer, F.; Rosch, P.; Popp, J. Applied and Environmental Microbiology 2012, 78, 5575-5583. Maquelin, K.; Kirschner, C.; Choo-Smith, L. P.; Ngo-Thi, N. A.; van Vreeswijk, T.; Stammler, M.; Endtz, H. P.; Bruining, H. A.; Naumann, D.; Puppels, G. J. Journal of Clinical Microbiology 2003, 41, 324-329. Lapizco-Encinas, B. H.; Rito-Palomares, M. Electrophoresis 2007, 28, 4521-4538. Bisceglia, E.; Cubizolles, M.; Mallard, F.; Vinet, F.; Francais, O.; Le Pioufle, B. Lab on a Chip 2013, 13, 901-909. Foudeh, A. M.; Didar, T. F.; Veres, T.; Tabrizian, M. Lab on a Chip 2012, 12, 32493266. Park, S.; Zhang, Y.; Wang, T. H.; Yang, S. Lab on a Chip 2011, 11, 2893-2900. Pohl, H. A. Journal of Applied Physics 1951, 22, 869-871. Gascoyne, P. R. C.; Vykoukal, J. Electrophoresis 2002, 23, 1973-1983. Pethig, R. Biomicrofluidics 2010, 4. Yang, L. J. Analytical Letters 2012, 45, 187-201. Voldman, J. Annual Review of Biomedical Engineering 2006, 8, 425-454. Castellarnau, M.; Errachid, A.; Madrid, C.; Juarez, A.; Samitier, J. Biophysical Journal 2006, 91, 3937-3945. Cheng, I. F.; Lin, C. C.; Lin, D. Y.; Chang, H. Biomicrofluidics 2010, 4. Cialla, D.; Marz, A.; Bohme, R.; Theil, F.; Weber, K.; Schmitt, M.; Popp, J. Analytical and Bioanalytical Chemistry 2012, 403, 27-54. Krause, M.; Rosch, P.; Radt, B.; Popp, J. Analytical Chemistry 2008, 80, 8568-8575. R Development Core Team, R.; R Foundation for Statistical Computing, ISBN 3-90005107-0: Vienna, Austria, 2012. Beleites, C.; Sergo, V.; Journal of Statistical Software, 2013. ACS Paragon Plus Environment

20

Page 21 of 28

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


(32) (33) (34) (35) (36) (37) (38) (39) (40) (41) (42)

Huang, Y.; Pethig, R. Measurement Science & Technology 1991, 2, 1142-1146. Green, N. G.; Ramos, A.; Morgan, H. Journal of Physics D-Applied Physics 2000, 33, 632-641. Castellanos, A.; Ramos, A.; Gonzalez, A.; Green, N. G.; Morgan, H. Journal of Physics D-Applied Physics 2003, 36, 2584-2597. Patel, P.; Markx, G. H. Enzyme and Microbial Technology 2008, 43, 463–470. Invitrogen; Molecular Probes Invitrogen, 2004. Kirschner, C.; Maquelin, K.; Pina, P.; Thi, N. A. N.; Choo-Smith, L. P.; Sockalingum, G. D.; Sandt, C.; Ami, D.; Orsini, F.; Doglia, S. M.; Allouch, P.; Mainfait, M.; Puppels, G. J.; Naumann, D. Journal of Clinical Microbiology 2001, 39, 1763-1770. Notingher, I. Sensors 2007, 7, 1343-1358. Notingher, I.; Verrier, S.; Romanska, H.; Bishop, A. E.; Polak, J. M.; Hench, L. L. Spectroscopy-an International Journal 2002, 16, 43-51. Puppels, G. J.; Demul, F. F. M.; Otto, C.; Greve, J.; Robertnicoud, M.; Arndtjovin, D. J.; Jovin, T. M. Nature 1990, 347, 301-303. Kussell, E.; Leibler, S. Science 2005, 309, 2075-2078. Rosch, P.; Harz, M.; Peschke, K. D.; Ronneberger, O.; Burkhardt, H.; Schule, A.; Schmauz, G.; Lankers, M.; Hofer, S.; Thiele, H.; Motzkus, H. W.; Popp, J. Anal Chem 2006, 78, 2163-2170.


21


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

Scheme 1: Workflow and time needed for the presented assay: the patient’s urine sample is filtered to remove larger cells such as leukocytes or epithelial cells. Medium exchange is achieved by centrifugation. The resuspended patient’s sample is injected on the combined dielectrophoresis-Raman chip where the bacteria are captured by DEP force and Raman spectra are recorded. Based on the Raman spectra the bacteria are identified and the result is available to the physician after only 35 minutes. 254x190mm (96 x 96 DPI)


Page 22 of 28

Page 23 of 28

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


Figure 1: a) The integrated dielectrophoresis Raman setup consists of a quartz glass substrate with gold microelectrodes contacted via tungsten tips. A 532 nm laser beam is focused through a 60x water immersion objective into a liquid droplet to excite the Raman signal of the captured bacteria. (b) In the dielectrophoresis design four electrodes with polynomial boundaries are symmetrically arranged to give a diametric distance of 40 µm. (c) Plot of the potential field distribution and the dielectrophoretic force (yellow arrows) in the center of the electrode array depicted in b). 352x87mm (150 x 150 DPI)



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

Figure 2: a) Negative dielectrophoretic trapping of E. coli (center) in between the four electrodes (black) in 0.5x PBS, 1 MHz and 4 Vpp. b) Real part of the CM-Factor calculated with a Two-Shell-Model for E. coli; (Ref. 26) c) Ellipsoidal Two-Shell Model for E. coli (Ref. 26) 254x92mm (150 x 150 DPI)


Page 24 of 28

Page 25 of 28

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


Figure 3: Fluorescence viability measurements, green fluorescence indicates viable cells, red fluorescence indicates non-viable cells ; DEP parameters are 1 MHz, 1 hour duration and from left to right 8 Vpp (cells stay alive), 14 Vpp (electrothermal whirls, but the cells stay alive) and 20 Vpp (the cells have died at such high voltages). 360x108mm (150 x 150 DPI)



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

Figure 4: Raman mean spectra with standard deviations of a) bacteria grown in culture and b) bacteria isolated from patient urine. (1), (3), (4): Escherichia coli, (2), (5): Enterococcus faecalis. 759x376mm (72 x 72 DPI)


Page 26 of 28

Page 27 of 28

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


Figure 5: Linear discriminant analysis; a, left) LDA scatter plot: each point represents one spectrum labeled with the index number. Two culture batches are used for training the classification model. Three culture batches and three patient samples are analyzed with this classification model. E. coli spectra are projected on negative LD1 values and E. faecalis spectra on positive LD1 values; a, right): LD1 histogram shows relative number of spectra trained and predicted for better visualization; b) LD1 coefficients which decompose the main spectral differences. 254x104mm (150 x 150 DPI)



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

Graphical abstract 81x45mm (150 x 150 DPI)


Page 28 of 28

Combined DielectrophoresisâRaman Setup for the Classification of

Recommend Documents

Combined DielectrophoresisâRaman Setup for the Classification of