Rapid Identification of Pseudomonas spp. via Raman Spectroscopy

Dec 25, 2015 - (3) Iron is an essential element for virtually all living organisms, because it is involved in many key metabolic processes within the ...
2 downloads 8 Views 1MB Size
Subscriber access provided by Penn State | University Libraries

Article

Rapid identification of Pseudomonas spp. via Raman Spectroscopy Using Pyoverdine as Capture Probe Susanne Pahlow, Stephan Stöckel, Sibyll Pollok, Dana CiallaMay, Petra Roesch, Karina Weber, and Juergen Popp Anal. Chem., Just Accepted Manuscript • DOI: 10.1021/acs.analchem.5b02829 • Publication Date (Web): 25 Dec 2015 Downloaded from http://pubs.acs.org on December 27, 2015

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 10

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

Rapid identification of Pseudomonas spp. via Raman Spectroscopy Using Pyoverdine as Capture Probe Susanne Pahlow1,2, Stephan Stöckel1,2, Sibyll Pollok3, Dana Cialla-May1,2,4, Petra Rösch1,2, Karina Weber1,2,4* and Jürgen Popp1,2,4 1

Institute of Physical Chemistry and Abbe Center of Photonics, Friedrich Schiller University Jena, Helmholtzweg 4, 07743 Jena, Germany 2 InfectoGnostics Research Campus Jena, Center for Applied Research, Philosophenweg 7, Jena, 07743, Germany 3 Ernst-Abbe-Hochschule, Carl-Zeiss-Promenade 2, 07745 Jena, Germany 4 Leibniz Institute of Photonic Technology, Albert-Einstein-Straße 9, 07745 Jena, Germany ABSTRACT: Pyoverdine is a substance, which is excreted by fluorescent pseudomonads in order to scavenge iron from their environment. Due to specific receptors of the bacterial cell wall the iron loaded pyoverdine molecules are recognized and transported into the cell. This process can be exploited for developing efficient isolation and enrichment strategies for members of the Pseudomonas genus, which are capable of colonizing various environments and also include human pathogens like P. aeruginosa and the less virulent P. fluorescens. A significant advantage over antibody based systems is the fact that siderophores like pyoverdine can be considered as ‘immutable ligands’, since the probability for mutations within the siderophore uptake systems of bacteria is very low. While each species of Pseudomonas usually produces structurally unique pyoverdines, which can be utilized only by the producer strain, cross reactivity does occur. In order to achieve a reliable identification of the captured pathogens, further investigations of the isolated cells are necessary. In this proof of concept study we combine the advantages of an isolation strategy relying on ‘immutable ligands’ with the high specificity and speed of Raman microspectroscopy. In order to isolate the bacterial cells pyoverdine was immobilized covalently on planar aluminum chip substrates. After capturing, single cell Raman spectra of the isolated species were acquired. Due to the specific spectroscopic fingerprint of each species the bacteria can be identified. This approach allows a very rapid detection of potential pathogens, since time-consuming culturing steps are unnecessary. We could prove that pyoverdine based isolation of bacteria is fully Raman compatible and further investigated the capability of this approach by isolating and identifying P. aeruginosa and P. fluorescens from tap water samples, which are both opportunistic pathogens and can pose a threat for immunocompromised patients.

The detection of bacterial pathogens like P. aeruginosa is a topic of high interest both in life science and environmental research, because microbial contamination can cause considerable harm in various environments, for example the water supply system. Regarding the detection of Pseudomonas spp. it is of particular importance to identify the exact species, because the Pseudomonads are a genus of great diversity and include many environmentally and medically relevant species, which differ significantly in terms of their pathogenicity. For preventing the spread of pathogenic bacteria such as P. aeruginosa sensitive detection methods have to be developed. A timely identification of the bacteria enables taking appropriate measures, so that further harm can be prevented. This illustrates, why the time, which is needed until the potential pathogens can be reliably identified, is such a crucial parameter in microbial testing. Thus, conventional methods, which often depend on culturing and take up to several days, need to be replaced by faster and more convenient alternatives. Two widely used and very efficient alternatives to cell culturing, the current gold standard of bacterial detection, are nucleic acid based approaches and immunoassays. With regard to sample preparation steps and technical requirements, anti-

body based methods are often even more convenient and rapid than nucleic acid based assays. Lateral flow tests, dipstick and agglutination assays can be easily performed by untrained personnel, also outside a specialized laboratory. 1 The antibodies used for such assays are capable of highly specific target recognition. This specificity, which is retained even in very complex surroundings, makes them highly valuable for multiple applications in biosensing. Despite these obvious and considerable advantages of immunoassays, the use of antibodies can entail some difficulties. Bacteria are constantly mutating and adapting to their environment. Due to this fact it is possible that the surface antigens which are responsible for the recognition by the immunoglobulines are also subject to mutation and the bacteria can no longer be detected or, even worse, might be mistaken for another species. This issue has led to the investigation of alternative capture probes for microorganisms, which also have a high specificity but whose recognition elements are less likely to mutate. Among these alternatives are glycanes and siderophores, which are also referred to as ‘immutable ligands’.2 This term can be considered justified, because these surface structures are directly related to the pathogens virulence. If a mutation actually appears, the patho-

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

gens would lose either their virulence or an important competitive advantage over other species. Siderophores are secreted by bacteria, plants and fungi in order to acquire iron from their environment. They are low molecular weight compounds (typically 500 – 1500 kDa), which are able to bind Fe(+III) with extremely high affinity. 3 Iron is an essential element for virtually all living organisms, because it is involved in many key metabolic processes within the cell like oxygen transport, DNA synthesis and electron transport.4 Even though iron is one of the most abundant elements of the earth crust, it has a very low bioavailability in the range of 10-9 – 10-18 M 5, which is far from the requirements of optimal microbial growth. In the human body the iron level is also strictly regulated, because in its free form it is toxic for cells. Most of the iron is tightly bound to proteins, like hemoglobin, ferritin or enzymes, resulting in a total free serum Fe(+III) concentration of 10-24 M.6 Consequently, strategies for scavenging iron from their surrounding or their host can be considered of great importance for the microbe’s survival and also its virulence.7 Recently it has been reported that siderophores make very promising capture probes for various microorganisms and that they are valuable components in bioassays aiming for the identification of bacteria.2a, b, 8 Next to the previously mentioned advantages over antibodies further benefits of siderophores are that it is possible to synthesize them chemically in some cases, as it was demonstrated for staphyloferrin A by Pandey et al.8f and that they are less sensitive than proteins regarding temperature and denaturizing conditions. An additional very interesting feature of siderophores might be the possibility to selectively enrich viable bacterial cells. So far this concept has been successfully established by Wolfenden et al. for the detection of Escherichia coli and by Kim et al. for Yersinia enterocolitica both using the desferrioxamine B iron complex.8d, e If siderophores are employed as capture probes for bacteria, it has to be taken into account that bacteria can not only bind the siderophores produced by their own species (endogenous siderophores) but also those secreted by foreign species (exogenous siderophores or xenosiderophores). 9 While it is possible to identify a bacterial species by analyzing the specific set of siderophores they are able to bind 8e, further characterization is necessary if only one siderophore is used for capturing. All studies conducted so far would have to rely on an array of siderophores in order to identify the bacterial species.2b, c, 8a Usually identification is supposed to be achieved by analyzing the pattern of bound bacteria on a chip with multiple capture probes. Within this contribution we demonstrate how Raman spectroscopy can be combined with siderophore based capturing for specific identification of bacterial species. We used pyoverdine produced by Pseudomonas fluorescens DSM 50090 for isolating various Pseudomonas species from tap water samples and subsequently acquired the Raman spectra of the captured bacteria, which then are used for assigning the cells to the correct species. Since different species of Pseudomonas can be captured with PVD50090 further characterization of the isolated bacteria is required for proper identification. Especially because the captured microorganisms have a differing pathogenicity, such further characterization is important. While P. aeruginosa is a well-known human pathogen, which is especially dangerous for immunocompromised patients, the other

Page 2 of 10

species are essentially harmless, even though P. fluorescens10 might also pose a threat for critically ill people. Raman microspectroscopy is a valuable tool for identifying microorganisms, which has been successfully employed for various applications.11 Since the investigation of single cells is possible, time-consuming cultivation steps can be omitted entirely. However, combining this technique with enrichment or isolation strategies is desirable, because in many cases the bacterial cells are concealed in a sample matrix, which usually will influence the Raman measurements and hamper the correct identification. Furthermore, there might be only a very low number of bacteria present in the sample, necessitating the utilization of enrichment procedures. In our approach we introduce a chip based strategy for isolation and enrichment of the bacterial cells, which comes along with the possibility for automated sample preparation.

EXPERIMENTAL SECTION Chip Preparation. The fabrication of the planar Raman chips as well as their modification with (3-glycidyloxypropyl) trimethoxysilane (GOPS) was performed as previously described. 12 Ferric pyoverdine from Pseudomonas fluorescens DSM 50090 was purchased from Sigma-Aldrich (Taufkirchen, Germany). For the immobilization of the pyoverdine Fe(+III) complex on the aluminum surface of the Raman chips, a solution of 0.05 mM was prepared in 0.1 M sodium phosphate buffer with pH 8.0. 0.3 µl were applied per spot. The chips were incubated in a humidity chamber at 4 °C over night and then washed twice with 1x phosphate buffered saline (PBS) for 5 min. After briefly rinsing with distilled water, the chips were then dried using pressurized air. In order to minimize unspecific binding the chips were blocked with 5% fetal calf serum in 0.1 M phosphate buffer. Cultivation of Bacteria and Isolation Experiments. P. aeruginosa DSM 22644, P. aeruginosa (environmental isolate), P. stutzeri DSM 5190 and Escherichia coli DSM 423 (Leibniz Institute DSMZ - German Collection if Microorganisms and Cell Cultures, Braunschweig, Germany) were cultured on lysogeny broth (Carl Roth GmbH & Co. KG, Karlsruhe, Germany) agar plates at 37 °C. P. fluorescens DSM 50090, P. fluorescens DSM 50106, P. chlororaphis DSM 50083 and P. psychrophila DSM 17535 were grown accordingly at 26 °C. For optimal results we incubated the cells, after reactivating them from a cryogenic culture, for at least three days, preparing a fresh plate every 24 hours. For the isolation experiments the bacterial cells were harvested from the plates after 24 h and washed three times with tap water, which was previously sterilized by filtration (membrane filter 0.22 µm, Carl Roth GmbH & Co. KG, Karlsruhe, Germany). Finally, the cells were suspended in tap water and the concentration was adjusted to approximately 109 cells/ml by measuring the optical density (OD) at 600 nm. For determining the limit of detection the samples were further diluted. Prior to applying the suspensions with the bacteria on the pyoverdine modified chip surface, the bacteria were incubated in tap water for 3 to 6 h. These samples were either applied directly on the chip or stored at 4 °C for later isolation experiments. In a typical experiment 100 µl of the sample were applied on the chip and incubated for up to 30 min. In order to remove unspecifically bound cells, the chips were then washed extensively with PBS buffer containing 0.01 % Tween 20. Finally the chips were briefly rinsed with distilled water. Before acquiring the microscopic images the captured bacteria cells were stained with

ACS Paragon Plus Environment

Page 3 of 10

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

100 µl of a carbol gentian violet solution (Carl Roth GmbH & Co. KG, Karlsruhe, Germany). Raman Measurements. The Raman measurements were carried out under ambient conditions using a Raman microscope setup (BioParticle Explorer, rap.ID Particle Systems GmbH, Berlin, Germany) with 532 nm excitation wavelength and a 100x magnification objective. More detailed information on the system can be found elsewhere. 11a-e The integration time per spectrum was 15 s as in the afore-mentioned references in order to achieve a sufficient signal-to-noise ratio. Data Preprocessing and Multivariate Analysis. Preprocessing of the spectral data as well as training of the selflearning machines to build a classification model and the validation thereof were conducted by using the open source software Gnu R with in-house developed scripts.13 The first preprocessing step consisted of a background elimination in each Raman spectrum. This background was calculated for each spectrum via the statistics-sensitive nonlinear iterative peak-clipping algorithm (SNIP) with a fourth order clipping filter, which is implemented in the package Peaks. 14 Subsequently, cosmic spikes in the Raman spectra were removed by comparing two consecutive spectra of one and the same cell and employing a robust variant of the upper-bound spectrum algorithm.15 The spectra afterwards were subjected to a wavenumber calibration with a Raman spectrum of acetaminophen as daily standard.16 Eventually, the spectral region under consideration was limited to the so-called fingerprint region (4371792 cm-1). After a further background correction, all spectra were scaled by means of a vector normalization. Based on the preprocessed data, a support vector machine (SVM) classification model using a radial basis function kernel was calculated by resorting to the package e1071.17 As has been shown previously, SVMs are well suited to classify and identify different bacterial species based on Raman spectra.18 A leave-onebatch-out cross-validation (LOBOCV) allowed to estimate the optimal kernel parameters cost=4 and gamma=0.001439194. To this end, each set of Raman spectra of one single batch (separately cultured biological replicate) is set aside once as test data while training is performed using all the other batches. 41 batches, each represented by approximately 50 Raman spectra were available for the LOBOCV. Thus, the generalization error could be estimated. Finally, Raman spectra of 6 other separately cultured, independent batches were taken as identification set. The model was thus used to predict this independent test data set to simulate an application of the procedure under realistic conditions, when the model would face ‘real-word’ samples.

RESULTS AND DISCUSSION Surface Functionalization of Aluminum with Pyoverdine. The aim of this study is to provide a new Raman compatible isolation strategy for Pseudomonas spp. using pyoverdine as capture probe. In order to enable Raman measurements directly on the chip substrate, which is utilized for the isolation of the bacterial cells, the material should not be Raman active itself. A suitable choice fulfilling this criterion is aluminum. Furthermore surface modification of this metal can be performed quite easily. We employed aluminum sputtered as a 210 nm thick layer on silicon wafers. 12 For covalently tethering the pyoverdine molecules to the metal surface, we chemically activated the latter one with an organosilane, which provides reactive epoxy groups. Pyover-

dines are generally composed of a dihydroxyquinoline chromophore, an acyl side chain and a peptidic chain, the sequence of which is depending on the strain. 19 The pyoverdine we employed for our experiments, was produced by P. fluorescens DSM 50090 and possesses a primary amine group due to the lysine within the peptide chain (see Figure 1a). This residue was used to immobilize the pyoverdine iron complex on the epoxy modified chip surface (see Figure 1b). The hydroxyl groups from the serine entities might also react with epoxy groups. However, this is rather unlikely since this reaction requires much more alkaline pH conditions than the ones used in the coupling protocol. Previously Renard et al. reported a similar approach for covalently attaching pyoverdine to porous silica20, while Doorneweerd et al. used microcontact printing for surface modification of their sensor with a pyoverdine bovine serum albumin complex.8c

Figure 1. a) Structure of the pyoverdine iron complex of P. fluorescens DSM 50090. b) Scheme illustrating the proposed isolation strategy for Pseudomonas spp.: The (3-glycidyloxypropyl)trimethoxysilane modified surface (1) is functionalized with an pyoverdine iron complex (2). Subsequently the chip can be used for capturing various Pseudomonas species (3).

Isolation of Pseudomonas spp. with a Pyoverdine Modified Chip. Under iron restricted conditions, many bacteria produce iron chelating molecules, referred to as siderophores. 5 These compounds possess an extraordinary high affinity to Fe(+III). Once the siderophores have bound the iron, the bacteria take up the iron chelate complex via specific receptors. The expression of siderophores and the corresponding receptors for iron uptake is regulated by the Ferric Uptake Regulator (FuR) protein and is dependent on the iron level in the environment. Since iron is an essential element for almost all microorganisms many species are able to not only take up their own iron loaded siderophores, but also those of other species, which offers them an advantage in the competition for iron. 21 This is also true for the fluorescent pseudomonads group. Next to their outer membrane receptors for pyoverdine, they can express further proteins, which enable them to use

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 10

Figure 2. Microscopic images of crystal violet stained cells after isolation with the pyoverdine modified chip. For each species a suspension with a concentration of 109 cells/ml was used.

various siderophores produced by other microorganisms. Interestingly, the uptake of pyoverdine is usually strain specific. The specificity arises from the individual sequence of the amino acids in the peptide chain of the pyoverdine molecule, which varies from strain to strain. The analysis of the peptide chain of pyoverdine can be even exploited for identifying Pseudomonas strains, especially for typing of different P. aeruginosa strains.22 Even though different members of the Pseudomonas genus can be distinguished by the amino acid sequence in the peptide chain of the pyoverdines, cross reactivity occurs more often than it was initially assumed.23 As a consequence this uptake of foreign siderophores has to be taken into account, if assays aiming for the identification microorganisms are established using siderophores as capture probes. Within this study we take advantage of the pyoverdine cross reactivity and demonstrate how several Pseudomonas species can be captured with one type of pyoverdine. As a first step towards developing a new Raman compatible isolation strategy, experiments with different Pseudomonas strains were performed using ferripyoverdine immobilized on an aluminum chip surface. An overview of the outcome of the isolation experiments for each species used in this study is given in Figure 2. After the capturing procedure the immobilized cells were stained with crystal violet. For each species an area of the pyoverdine modified chip surface with a typical cell density after the siderophore mediated isolation is shown. For the experiments a rather high concentration of 109 cells/ml was used, because by doing so the pyoverdine modified areas can be easily visualized in the microscopic images (see Figure 3b and Figures S-1 – S-5). As expected P. fluorescens DSM 50090 did bind to its cognate pyoverdine (PVD50090) and may be isolated with our chip based approach. Apparently, the immobilization strategy does not impair recognition of the ferrisiderophore by its receptor. Also, the well characterized PAO 1 strain of P. aeruginosa (DSM 22644) can be isolated

with immobilized PVD50090. This result was to be expected, since previous studies proved that the pyoverdine-Fe transporter (FpvA) of P. aeruginosa DSM 22644 is able to recognize PVD50090 and many others. 23b Furthermore, we successfully isolated P. chlororaphis DSM 50083. Because the pyoverdine’s peptidic chain of this strain is identical with the one of P. fluorescens DSM 50090 (see Table 1), the accordant reactivity seems well founded. Moreover, these results are in perfect agreement with a study of Hohnadel and Meyer, who also found that the three strains, mentioned above, are able to incorporate each other’s pyoverdines.23c We were further able to isolate P. psychrophila DSM 17535, P. fluorescens DSM 50106 and an environmental strain of P. aeruginosa with our system. Attempts to isolate P. putida DSM 291 and P. stutzeri DSM 5190 failed. P. putida is not able to use the heterologous PVD50090 for an efficient iron uptake23c, so it is not surprising that this strain cannot be captured with this type of pyoverdine. P. stutzeri is a non pyoverdine producing species and therefore lacks the corresponding receptors necessary for the uptake. The few cells of P. putida and P. stutzeri, which are observed on the chip surface in Figure 2, are due to unspecific binding. It should be noted that the pyoverdines produced by P. aeruginosa strains can be each assigned to three structurally different types.27 With the here introduced isolation strategy only P. aeruginosa strains belonging to the group I are accessible. In order to prove that the binding of the bacterial cells to the chip surface is actually due to the pyoverdine modification, only six out of eight square aluminum fields were spotted with the siderophore iron complex. In Figure 3b microscopic images of a chip are shown after isolation of P. psychrophila. As expected the cells are predominantly located on the fields with the pyoverdine modification. The circular shape of the area

ACS Paragon Plus Environment

Page 5 of 10

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

where the ferric pyoverdine was applied can be clearly recognized, which also indicates that the immobilization of the pyoverdine was successful. On the unmodified aluminum fields only few cells can be found due to unspecific binding. Further images for other species can be found in the supporting information of this article (Figure S-1 – S-5). Table 1. Peptidic sequences of the pyoverdines from species investigated in this study. Cyclic structures are indicated in brackets. D-amino acids are underlined. Species/ strain P. fluorescens DSM 50106 ATCC 17826 P. fluorescens DSM 50090 ATCC 13525 P. aeruginosa DSM 22644 ATCC 15692 P. aeruginosa env. isolate P. putida DSM 291 ATCC 12633 P. psychrophila DSM 17535 P. stutzeri DSM 5190 ATCC 17588 P. chlororaphis DSM 50083 ATCC 9446

Peptide sequence

Reference

Ser-Lys-Gly-FoOHOrn-Ser-Ser-Gly[Orn-FoOHOrn-Ser]

23b

Ser-Lys-Gly-FoOHOrn-[Lys-FoOHOrnSer]

24

Ser-Arg-Ser-FoOHOrn-[Lys-FoOHOrnThr-Thr]

25

undetermined

-

Asp-Lys-OHAsp-Ser-Thr-Ala-Glu-Ser[OHOrn]

26

undetermined

-

non pyoverdine producing organism (produces desferrioxamine E)

-

Ser-Lys-Gly-FoOHOrn-[Lys-FoOHOrnSer]

26

Within the course of our study we found that cells, cultivated on lysogeny broth (LB) agar plates and subsequently incubated in tap water, can be successfully isolated using immobilized ferripyoverdine, even though LB agar is an iron rich medium, which is not optimal for the stimulation of siderophore production. We assume that the combination of the cultivation on plates and the tap water treatment are sufficient to induce expression of the pyoverdines and the corresponding receptors for their uptake. Isolation experiments with cells grown in liquid LB medium until the exponential phase were not successful. The growth conditions in liquid medium are obviously favorable for the cells. If it is desired to use cells grown in liquid medium, either iron deficient media (i.e. succinate medium or cas amino acid medium) or a strong iron chelator like Ethylenediamine-di(o-hydroxyphenylacetic acid) (EDDHA) have to be employed. Raman Measurements of Pseudomonas spp. Isolated with a Pyoverdine Modified Chip. Raman spectroscopy allows analyzing bacteria on single cell level. In order to identify a bacterial species a database has to be established first. For this purpose numerous single cell spectra of the species of interest have to be recorded. Based on these data a statistical model is created, which allows analyzing spectra of cells isolated from unknown samples and assigning them to the correct species. For applying this technique in combination with sample preparation procedures, it has to be ensured that the treatment of the cells does not interfere with the Raman based identification. For example, dominant spectral features of the capture molecules present in the bacterial Raman spectra would hamper the identification process. As a first step towards building a database for Pseudomonas spp., the Raman compatibility of the pyoverdine based isola-

tion strategy has to be ensured. For this purpose, we recorded spectra of five different Pseudomonas species and one Escherichia coli strain on the pyoverdine modified substrates. Furthermore background spectra of the pyoverdine modified areas and also spectra of unmodified fields were acquired. Except from a very slight fluorescence background no distinct Raman bands, originating from the surface modifications, were observed in the background spectra (Figure 4). Even though pyoverdine contains a dihydroxyquinoline chromophore, its fluorescence is quenched, because the iron complex was employed for the experiments. A more detailed depiction of the background spectra, as well as a reference spectrum for ferric pyoverdine can be found in the supporting information in Figure S-6. The mean spectra of E. coli and the different Pseudomonas species shown in Figure 4 resemble typical Raman spectra of bacterial cells. A Raman spectrum of a cell is quite complex, since all components of the cell will contribute to the Raman signal, resulting in a superposition of various spectra. Characteristic features like the broad band centered at 2900 cm-1, which results from C-H-stretching vibrations of CH2 and CH3 groups, as well as a band at 1448 cm-1 which can be assigned to the deformation vibrations of CH2 and CH3 groups were observed. Furthermore contributions of proteins like the amide I and amide III bands at 1665 and 1245 cm -1 and the sharp phenylalanine band at 1004 cm-1 can be found in the spectra. Also, as expected, the bands resulting from nucleic acids (i.e. 780, 1101 and 1575 cm-1) are present.

Figure 3. a) Schematic display of the isolation protocol for demonstrating the specific capture of Pseudomonas spp. due to the pyoverdine modification. b) Microscopic images of a pyoverdine modified Al field and an unmodified field after incubation with P. psychrophila shown in different magnifications.

With regard to a possible application of this approach, a database with spectra of bacterial cells, preconditioned in tap water, was established. Next to five Pseudomonas spp. E. coli

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

was included in the database as well, because it is a possible contaminant in drinking water. Even though E. coli cannot be captured with pyoverdine, because it lacks the appropriate receptors, the isolation experiments showed that a few cells can stick on the chip surface due to unspecific adsorption. This was also observed for P. stutzeri and P. putida. For each species at least four independent batches with approximately 50 single cell spectra were investigated for the database. The database served as training ground for an SVM algorithm to build a classification model that learned how to discriminate the 2165 Raman spectra of the 6 analyzed bacterial species originating from 41 single batches. A leave-one-batch-out cross-validation (LOBOCV) was performed to tune the kernel parameters of the employed radial basis kernel and to validate the classifier. By holding out all the Raman spectra of one batch and refitting the model with the residual Raman spectra of the 40 other batches, a minimal correlation between the model data and the held-out data is achieved, because the Raman spectra of one and the same batch were not shared between the test set and the surrogate model on each fold. After the determination of the kernel parameters, 85.5% of the Raman spectra were correctly assigned to their respective class in the course of a LOBOCV (see Table 2).

Page 6 of 10

our approach. The isolated cells then were identified using the fully-trained classifier. The dataset contained 333 Raman spectra from six samples of P. aeruginosa and P. fluorescens isolated from tap water. By facing the SVM model with this completely independent dataset, a real-world-scenario is simulated, when “unknown” samples are to be identified. The outcome is given in

Figure 4. Mean Raman spectra of E. coli and different Pseudomonas spp. recorded on pyoverdine modified aluminum substrates. For each mean spectrum approximately 50 single cell spectra were used, the light grey areas indicate the standard deviation.

Identification of P. aeruginosa and P. fluorescens in Tap Water. In order to determine the limit of detection tap water samples were spiked with different concentrations of P. fluorescens DSM 50090 and the environmental isolate of P. aeruginosa. The number of cells in the samples was determined via plate counting. After the chip based isolation procedure, the Raman measurements were performed. For a reliable identification at least 40 cells were measured from each sample. We found that down to a concentration of 5 x 10 3 CFU/ml and a sample volume of 100 µl this minimum number of cells can be easily achieved, because enough bacteria can be isolated with

ACS Paragon Plus Environment

Page 7 of 10

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 3: 302 of 333 Raman spectra were correctly assigned to the respective Pseudomonas species giving an overall accuracy of 90.7% and specificity of 100% for E. coli. The batchwise sensitivities ranged from 84.2% for sample #04 (P. fluorescens with 105 CFU/ml) to 96.7% for sample #03 (P. aeruginosa with 103 CFU/ml), which corresponds well with the achieved accuracy from LOBOCV. P. aeruginosa (species-wise sensitivity 93.2%) was generally better identified than P. fluorescens (87.3%), while a correlation between the batch-wise sensitivities and the initial bacterial load of the batches was not found. All in all, a reliable identification of the bacteria was achieved in each of the six test samples with concentrations down to 5x 103 CFU/ml. The actual concentration span for P. aeruginosa in real water samples is very wide and ranges from 1 – 2,400 CFU/ml.28 In order to also be able to analyze samples with very low cell numbers, the implementation of a prior enrichment step, for example via filtration of a larger volume, would be suitable. Table 2. Classification results of the LOBOCV for a SVM. identified as

true Ecol

Paer

Pchl

Pflu

Ppsy

Pstu

specificity [%]

Ecol

262

2

2

3

6

1

99.3

Paer

15

347

8

16

14

38

94.8

Pchl

0

6

165

10

2

1

99.0

Pflu

3

27

3

335

5

20

96.7

Ppsy

7

11

2

13

369

10

97.6

Pstu sensitivity [%]

1

40

10

28

9

374

94.9

91.0

80.1

86.8

82.7

91.1

84.2

85.5

Ecol, E. coli; Paer, P. aeruginosa; Pchl, P. chlororaphis; Pflu, P. fluorescens; Ppsy, P. psychrophila; Pstu, P. stutzeri

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

Table 3. Results of the identification after processing six test samples.

Page 8 of 10

AUTHOR INFORMATION Corresponding Author *e-mail: [email protected]

true

*phone +49-3641-206 309 P. aeruginosa

P. fluorescens

Notes #01

#02 5

#03 4

#04 3

#05 5

The authors declare no competing financial interest.

#06 4

3

5 x 10 5 x 10 5 x 10 5 x 10 5 x 10 5 x 10 CFU/ml CFU/ml CFU/ml CFU/ml CFU/ml CFU/ml identified as E. coli

0

0

0

0

0

0

P. aeruginosa

68

51

59

7

0

5

P. chlororaphis

1

0

0

0

0

0

P. fluorescens

3

2

1

48

39

37

P. psychrophila

2

0

0

0

0

0

P. stutzeri

2

1

1

2

3

1

sensitivity [%]

89.5

94.4

96.7

84.2

92.7

86.0

CONCLUSION For the first time a siderophore based isolation strategy for bacterial cells was successfully combined with Raman microspectroscopy enabling the detection of microbial contamination in tap water. Our innovative approach allows isolation of P. aeruginosa and other pseudomonads using the ferric complex of the fluorescent pigment pyoverdine. Even though the structure of the pyoverdines slightly varies from strain to strain, cross reactivity does occur and demands further characterization of the isolated cells in order to unambiguously identify the actual species. Within this study, we demonstrate that Raman microspectroscopy is a suitable option for a reliable identification of the Pseudomonas species. Six independent tap water samples spiked with P. fluorescens and P. aeruginosa were successfully identified using an SVM model. Overall an accuracy of 90.7 % is achieved, which is sufficient for identifying the type of microbial contamination. Single cells can be investigated individually and even remain intact for further analysis, if desired. With our chip based isolation approach, time consuming cultivation steps can be avoided entirely. Concentrations as low as 5x 10 3 CFU/ml can be readily detected within two hours, which allows an appropriate intervention in case a human pathogen was identified.

ASSOCIATED CONTENT Supporting Information Further microscopic images of isolated Pseudomonas spp. and a reference spectrum of pyoverdine are supplied in the supporting information. This material is available free of charge via the Internet at http://pubs.acs.org.

ACKNOWLEDGMENT Funding of the research projects BioInter (13022-715) by the Development Bank of Thuringia and the European Union (EFRE) as well as InfectoGnostics (13GW0096F), Fast Diagnosis (13N11350) and JCBI 2.0 (03IPT513Y) by the Federal Ministry of Education and Research (BMBF), Germany is gratefully acknowledged. Furthermore funding of the collaborative research center ChemBioSys (SFB 1127) by the Deutsche Forschungsgemeinschaft (DFG) is highly acknowledged. We would like to thank Uwe Hübner and Konstantin Kirsch for fabricating the chip substrates.

REFERENCES 1. P. P. Banada, A. K. Bhunia, in Principles of Bacterial Detection: Biosensors, Recognition Receptors and Microsystems, ed. M. Zourob, S. Elwary, A. Turner. Springer: New York, 2008, pp 567-602. 2. (a) H. Y. Kim, C. R. Estes, A. G. Duncan, B. D. Wade, F. C. Cleary, C. R. Lloyd, W. R. Ellis, L. S. Powers, Ieee Engineering in Medicine and Biology Magazine 2004, 23. 122-129; (b) H. Y. Mason, C. Lloyd, M. Dice, R. Sinclair, W. Ellis, L. Powers, Biosensors & Bioelectronics 2003, 18. 521-527, DOI: 10.1016/s0956-5663(03)00010-1; (c) A. K. Adak, J. W. Boley, D. P. Lyvers, G. T. Chiu, P. S. Low, R. Reifenberger, A. Wei, Acs Applied Materials & Interfaces 2013, 5. 64046411, DOI: 10.1021/am4016236. 3. (a) R. C. Hider, X. Kong, Natural Product Reports 2010, 27. 637-657, DOI: 10.1039/b906679a; (b) G. Schwarzenbach, K. Schwarzenbach, Helvetica Chimica Acta 1963, 46. 1390-1400, DOI: 10.1002/hlca.19630460434. 4. (a) M. Miethke, M. A. Marahiel, Microbiology and Molecular Biology Reviews 2007, 71. 413–451, DOI: 10.1128/mmbr.00012-07; (b) P. T. Lieu, M. Heiskala, P. A. Peterson, Y. Yang, Molecular aspects of medicine 2001, 22. 1-87, DOI: 10.1016/s0098-2997(00)00006-6; (c) S. Fardeau, C. Demailly-Mullie, A. Dassonville-Klimpt, N. Audic, P. Sonnet, Abstracts of Papers of the American Chemical Society 2011, 242. 695705. 5. R. Saha, N. Saha, R. S. Donofrio, L. L. Bestervelt, Journal of Basic Microbiology 2013, 53. 303-317, DOI: 10.1002/jobm.201100552. 6. K. N. Raymond, E. A. Dertz, S. S. Kim, Proceedings of the National Academy of Sciences of the United States of America 2003, 100. 3584-3588, DOI: 10.1073/pnas.0630018100. 7. (a) J. M. Meyer, A. Neely, A. Stintzi, C. Georges, I. A. Holder, Infection and Immunity 1996, 64. 518-523; (b) S. E. Dale, A. DohertyKirby, G. Lajoie, D. E. Heinrichs, Infection and Immunity 2004, 72. 2937, DOI: 10.1128/iai.72.1.29-37.2004. 8. (a) S.-M. Wu, Z.-L. Zhang, X.-D. Wang, M.-X. Zhang, J. Peng, Z.-X. Xie, D.-W. Pang, Journal of Physical Chemistry C 2009, 113. 9169-9174, DOI: 10.1021/jp901221h; (b) N. Bugdahn, F. Peuckert, A. G. Albrecht, M. Miethke, M. A. Marahiel, M. Oberthuer, Angewandte Chemie-International Edition 2010, 49. 10210-10213, DOI: 10.1002/anie.201005527; (c) D. D. Doorneweerd, W. A. Henne, R. G. Reifenberger, P. S. Low, Langmuir 2010, 26. 15424-15429, DOI: 10.1021/la101962w; (d) M. L. Wolfenden, R. M. Sakamuri, A. S. Anderson, L. Prasad, J. G. Schmidt, H. Mukundan, Advances in Biological Chemistry 2012, 2. 396-402, DOI: 10.4236/abc.2012.24049; (e) Y. Kim, D. P. Lyvers, A. Wei, R. G. Reifenberger, P. S. Low, Lab on a Chip 2012, 12. 971-976, DOI: 10.1039/c2lc20904g; (f) R. K. Pandey, G. G. Jarvis, P. S. Low, Organic & Biomolecular Chemistry 2014, 12. 1707-1710, DOI: 10.1039/c3ob41230j; (g) T. Zheng, E. M. Nolan, Metallomics 2012, 4. 866-880, DOI: 10.1039/c2mt20082a. 9. (a) M. Miethke, T. Kraushaar, M. A. Marahiel, FEBS Letters 2013, 587. 206-213, DOI: http://dx.doi.org/10.1016/j.febslet.2012.11.027;

ACS Paragon Plus Environment

Page 9 of 10

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

(b) A. D'Onofrio, J. M. Crawford, E. J. Stewart, K. Witt, E. Gavrish, S. Epstein, J. Clardy, K. Lewis, Chemistry & Biology 2010, 17. 254-264, DOI: http://dx.doi.org/10.1016/j.chembiol.2010.02.010; (c) W. Rabsch, G. Winkelmann, Biol Metals 1991, 4. 244-250, DOI: 10.1007/BF01141188. 10. B. S. Scales, R. P. Dickson, J. J. LiPuma, G. B. Huffnagle, Clinical microbiology reviews 2014, 27. 927-948. 11. (a) S. Stöckel, S. Meisel, M. Elschner, F. Melzer, P. Rösch, J. Popp, Anal Bioanal Chem 2015, 407. 787-794, DOI: 10.1007/s00216-0147906-5; (b) S. Meisel, S. Stöckel, P. Rösch, J. Popp, Food Microbiology 2014, 38. 36-43, DOI: http://dx.doi.org/10.1016/j.fm.2013.08.007; (c) A. Silge, W. Schumacher, P. Rösch, P. A. Da Costa Filho, C. Gérard, J. Popp, Systematic and Applied Microbiology 2014, 37. 360-367, DOI: http://dx.doi.org/10.1016/j.syapm.2014.05.007; (d) S. Kloß, B. Kampe, S. Sachse, P. Rösch, E. Straube, W. Pfister, M. Kiehntopf, J. Popp, Analytical Chemistry 2013, 85. 9610-9616, DOI: 10.1021/ac401806f; (e) S. Stöckel, S. Meisel, M. Elschner, P. Rösch, J. Popp, Angewandte Chemie-International Edition 2012, 51. 5339-5342, DOI: 10.1002/anie.201201266; (f) S. Pahlow, S. Meisel, D. Cialla-May, K. Weber, P. Rösch, J. Popp, Advanced Drug Delivery Reviews 2015. doi:10.1016/j.addr.2015.04.006, DOI: doi:10.1016/j.addr.2015.04.006. 12. S. Pahlow, S. Kloss, V. Blättel, K. Kirsch, U. Hübner, D. Cialla, P. Rösch, K. Weber, J. Popp, Chemphyschem 2013, 14. 3600-3605, DOI: 10.1002/cphc.201300543. 13. R Development Core Team. R Foundation for Statistical Computing: Vienna, Austria, 2011. 14. (a) M. Morhác, Nuclear Instruments and Methods in Physics Research Section A: Accelerators, Spectrometers, Detectors and Associated Equipment 2009, 600. 478-487; (b) M. Morhác, Peaks: Peaks, R package version 0.2. 2008. 15. D. M. Zhang, K. N. Jallad, D. Ben-Amotz, Applied Spectroscopy 2001, 55. 1523-1531, DOI: 10.1366/0003702011953757. 16. M. M. Carrabba, in Handbook of Vibrational Spectroscopy. John Wiley & Sons, Ltd, 2006, vol. 1. 17. E. Dimitriadou, K. Hornik, F. Leisch, D. Meyer, A. Weingessel, e1071: Misc Functions of the Department of Statistics (e1071), TU Wien. 2011. 18. (a) I. Espagnon, D. Ostrovskii, R. Mathey, M. Dupoy, P. L. Joly, A. Novelli-Rousseau, F. Pinston, O. Gal, F. Mallard, D. F. Leroux, Journal of Biomedical Optics 2014, 19. DOI: 10.1117/1.jbo.19.2.027004; (b) S. Kloß, P. Rösch, W. Pfister, M. Kiehntopf, J. Popp, Anal. Chem. 2015, 87. 937-943, DOI: 10.1021/ac503373r. 19. J. M. Meyer, Archives of Microbiology 2000, 174. 135-142, DOI: 10.1007/s002030000188. 20. G. Renard, M. Mureseanu, A. Galarneau, D. A. Lerner, D. Brunel, New Journal of Chemistry 2005, 29. 912-918, DOI: 10.1039/b500302b. 21. P. Cornelis, C. Baysse, S. Matthijs, in Pseudomonas: genomics and molecular biology, ed. P. Cornelis. Horizon Scientific Press, 2008. 22. (a) R. Fuchs, M. Schafer, V. Geoffroy, J. M. Meyer, Current topics in medicinal chemistry 2001, 1. 31-57, DOI: 10.2174/1568026013395542; (b) S. Kilz, C. Lenz, R. Fuchs, H. Budzikiewicz, Journal of Mass Spectrometry 1999, 34. 281-290, DOI: 10.1002/(SICI)1096-9888(199904)34:43.0.CO;2-M. 23. (a) P. Cornelis, S. Matthijs, Environmental Microbiology 2002, 4. 787-798, DOI: 10.1046/j.1462-2920.2002.00369.x; (b) J. Greenwald, M. Nader, H. Celia, C. Gruffaz, V. Geoffroy, J.-M. Meyer, I. J. Schalk, F. Pattus, Molecular Microbiology 2009, 72. 1246-1259, DOI: 10.1111/j.1365-2958.2009.06721.x; (c) D. Hohnadel, J. M. Meyer, Journal of Bacteriology 1988, 170. 4865-4873; (d) P. Cornelis, Applied Microbiology and Biotechnology 2010, 86. 1637-1645, DOI: 10.1007/s00253-010-2550-2. 24. C. Linget, P. Azadi, J. K. Macleod, A. Dell, M. A. Abdallah, Tetrahedron Letters 1992, 33. 1737-1740, DOI: 10.1016/s00404039(00)91719-2. 25. G. Briskot, K. Taraz, H. Budzikiewicz, Liebigs Annalen Der Chemie 1989. 375-384. 26. P. Demange, S. Wendenbaum, A. Bateman, A. Dell, J. M. Meyer, M. A. Abdallah, in Iron, Siderophores, and Plant Diseases, ed. T. R. Swinburne. Springer US, 1986, vol. 117, pp 131-147. 27. (a) P. Cornelis, D. Hohnadel, J.-M. Meyer, Infection and immunity 1989, 57. 3491-3497; (b) J.-M. Meyer, A. Stintzi, D. De Vos, P. Cornelis, R. Tappe, K. Taraz, H. Budzikiewicz, Microbiology 1997, 143. 35-43; (c) D. De Vos, M. De Chial, C. Cochez, S. Jansen, B. Tümmler, J.M. Meyer, P. Cornelis, Archives of Microbiology 2001, 175. 384-388, DOI: 10.1007/s002030100278.

28. K. D. Mena, C. P. Gerba, in Reviews of Environmental Contamination and Toxicology Vol 201. Springer, 2009, pp 71-115.

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 10

for TOC only

ACS Paragon Plus Environment

10