Carbohydrates as New Probes for the Identification of Closely Related

Jan 12, 2015 - Mader , A.; Gruber , K.; Castelli , R.; Hermann , B. A.; Seeberger , P. H.; Rädler , J. P.; Leisner , M. Nano Lett. 2012, 12, 420– 4...
0 downloads 0 Views 2MB Size
Article pubs.acs.org/ac

Carbohydrates as New Probes for the Identification of Closely Related Escherichia coli Strains Using Surface Plasmon Resonance Imaging Emilie Bulard,†,‡,§,∥,⊥ Aurélie Bouchet-Spinelli,*,†,‡,§ Patricia Chaud,∥,⊥ André Roget,†,‡,§ Roberto Calemczuk,†,‡,§ Sébastien Fort,∥,⊥ and Thierry Livache†,‡,§ †

Univ. Grenoble Alpes, INAC-SPRAM, F-38000 Grenoble, France CEA, INAC-SPRAM, F-38000 Grenoble, France § CNRS, SPRAM, F-38000 Grenoble, France ∥ Univ. Grenoble Alpes, CERMAV, F-38000 Grenoble, France ⊥ CNRS, CERMAV, F-38000 Grenoble, France ‡

S Supporting Information *

ABSTRACT: Prevention of foodborne diseases depends highly on our ability to control rapidly and accurately a possible contamination of food. So far, standard procedures for bacterial detection require time-consuming bacterial cultures on plates before the pathogens can be detected and identified. We present here an innovative biochip, based on direct differential carbohydrate recognitions of five closely related Escherichia coli strains, including the enterohemorragic E. coli O157:H7. Our device relies on efficient grafting of simple carbohydrates on a gold surface and on the monitoring of their interactions with bacteria during their culture using surface plasmon resonance imaging. We show that each of the bacteria interacts in a different way with the carbohydrate chip. This allows the detection and discrimination of the tested bacterial strains in less than 10 h from an initial bacterial concentration of 102 CFU·mL−1. This is an improvement over previously described systems in terms of cost, easiness to use, and stability. Easily conceived and easily regenerated, this tool is promising for the future of food safety.

T

electrochemical detection of various E. coli strains has been studied by several groups using impedance monitoring. The electrode surfaces could be covered with antibodies known to recognize specific bacteria with a limit of detection (LOD) reaching 104 CFU·mL−1.6 More recently, using surfaces modified with carbohydrates7 or phages8 the LOD could be decreased to 1.2 × 102 and 8 × 102 CFU·mL−1, respectively. Another approach combining enzymatic amplification and redox cycling allowed reaching a 103 CFU·mL−1 LOD for the detection of E. coli O157:H7.9 Concerning the use of an optical transduction mode, surface plasmon resonance (SPR) has proved to be a very well-adapted method for the detection of bacteria onto microarrays.10 This technique, based on refractive index changes, monitors the interactions occurring between biomolecules (called probes) grafted on a gold surface and target entities such as proteins or whole cells within the sample in real time and with no need for labeling steps. SPR has been used to detect and identify different pathogenic bacteria in diluted samples using probes such as antibodies,11 lectins (i.e.,

he detection of bacteria in food is a major concern of public health. Centers for Disease Control and Prevention estimate that 1 in 6 Americans are infected by pathogenic food bacteria each year: for example, the Escherichia coli O157 outbreaks in 2011 led to 2138 hospitalizations.1 Other nonO157 serotypes such as E. coli O145:H28 have also caused human diseases.2,3 The detection of pathogens remains difficult. This is mainly due to the very small quantity of pathogens responsible for diseases. Most conventional detection and identification methods (i.e., immunoassays, DNA hybridization, PCR, spectrometry, etc.) require bacteria enrichment steps to increase the number of bacteria above their limit of detection.4,5 For the most used immunoassays, the detection limit is around 105 colony forming units (CFU) per milliliter. The enrichment phase consists in culturing the sample in solution for 12 to 48 h and is therefore time-consuming. During this period, consumers are not being advised to avoid eating any specific contaminated foods and the epidemic continues to spread. In the past decade, there has been an increasing interest for the conception of rapid, sensitive, and/or specific tools to quickly detect and identify very small amounts of pathogens. Particularly, label-free procedures were developed in order to overcome the problems of long operation and detection times, enrichment requirements, labeling necessity, and high cost. The © XXXX American Chemical Society

Received: October 8, 2014 Accepted: January 11, 2015

A

DOI: 10.1021/ac5037704 Anal. Chem. XXXX, XXX, XXX−XXX

Article

Analytical Chemistry proteins recognizing specifically carbohydrates),12 or bacteriophages.13 The best LODs are currently around 103 CFU· mL−1. Our team has recently described a new methodology named culture−capture−measure (CCM).14,15 The innovative aspect is that the presence of a given bacteria is revealed during the bacterial growth using surface plasmon resonance imaging (SPRi). We could that way detect as low as 20 CFU·mL−1 in the initial sample. The CCM procedure presents several advantages over the previously described SPR assays. There is no need for enrichment, and the process becomes as a consequence one step. The detection is highly specific toward a given pathogen, even in complex media like raw milk or meat. In these works, antibodies were used as probes for bacterial growth detection. Although the bacterial detection onto antibody-modified surfaces was efficient, these probes are expensive due to their production process, sensitive to dehydration (protein denaturation), and they are mostly single use probes because of the difficulty to remove softly targets from the surface. Our aim in this work is therefore to provide a new tool based on simple biomolecular probes which are able, like antibodies, to rapidly detect and identify bacterial strains using the CCM methodology. Carbohydrates can provide interesting alternatives to this aim.16,17 They are indeed relevant probes as bacteria are known to interact with carbohydrates by different ways: through adhesins first, which are particular lectins produced on the outer cell wall, and also by nonspecific bindings such as interactions with membrane transporters or electrostatic interactions.18,19 In order to minimize the complexity of the probes, we chose to work with seven mono- or disaccharides: glucose (Glc), galactose (Gal), mannose (Man), fucose (Fuc), maltose (Malt), N-acetylglucosamine (GlcNAc), and sialic acid (Neu5Ac). These carbohydrates were conjugated to pyrrole in order to allow grafting on chip using pyrrole electropolymerization. This technique, developed in our team,20−23 allows the individual electrical addressing of each carbohydrate on the biochip. The CCM monitoring of the carbohydrate−bacteria interactions could then be performed in real time without the use of any labeling step by SPRi. In this study, the analysis will be carried out using five closely related E. coli strains.

(ATCC 10798) was chosen as the nonpathogenic bacterial strain and purchased from ATCC. Bacterial Suspensions. Bacterial cells were grown overnight at 37 °C in TSB culture medium. Overnight grown bacterial culturescorresponding to 109−108 colony forming units per milliliter (CFU·mL−1)were used for SPRi experiments after serial dilutions. Bacterial counting was carried out either by McFarland turbidity measurements (Densimat apparatus, BioMérieux, Marcy l’Etoile, France) or colony counting after plating and culture on solid media TSA. For colony counting, the counting of four plates was averaged and standard deviation was calculated. Synthesis of Pyrrole−Carbohydrate Conjugates. According to a previously reported procedure,23 pyrrole− carbohydrate conjugates have been prepared from the corresponding allyl glycoside by UV-promoted radical addition of cysteamine followed by reaction with a pentafluorophenylactivated pyrrole ester. The different conjugates were isolated in 30−50% yield after purification by reversed phase SPE (solid phase extraction) on C18 Sep-Pak cartridge and characterized by 1H and 13C NMR and mass spectrometry (MS) (see Supporting Information). Final solid products were stored at 4 °C. Microarraying and SPRi. Aliquots of pyrrole−carbohydrate conjugates were dissolved in spotting buffer (50 mM phosphate buffer, pH 6.8, 50 mM NaCl, and 10% glycerol) containing 20 mM pyrrole. The optimized concentration of pyrrole−carbohydrate conjugate in the mixture was 10 mM. SPRi biochips were purchased from Horiba Scientific (ChillyMazarin, France) and are made of a high-index glass prism covered with a 2 nm thick chromium and a 50 nm thick gold layer used as a working electrode (about 2 cm 2 ). Coelectropolymerization of pyrrole and pyrrole−carbohydrate conjugates on the biochips was carried out in an automatized pipet tip (diameter 500 μm) filled with the solution to be polymerized and containing a platinum wire used as a counter electrode. The pipet tip was moved at the vicinity of the gold layer of the SPR biochip, until an electrical contact was applied between the working (gold surface) and counter (platinum wire) electrodes. The polymerization on the prism gold layer was performed with a 100 ms electric pulse at a 2.0 V bias independently of any reference electrode. After microarraying, biochips were copiously washed with water, dried, and stored in air at 4 °C. Each biochip was arrayed with quadruplicates of pyrrole−carbohydrate conjugates. Moreover, four spots of polypyrrole (Ppy) deprived of any carbohydrates were also deposited onto the gold surface to assess nonspecific SPR response. Signal measurements were performed using a SPRiLab+TM system (Horiba Scientific, Chilly-Mazarin, France). Using this commercial software, regions of interest (ROI) corresponding to individual spots on the biochips were defined. The SPRi signal was monitored with a CCD camera, and reflectivity changes (ΔR) of each ROI were followed and plotted upon time. Two devices were used: a simple chamber (1.6 mL) and a double reaction chamber (2 × 500 μL). For the double reaction chamber, two identical series of carbohydrates were arrayed on the biochip, to fit each chamber. This device allowed two experiments to be launched in parallel on the same biochip. Characterization of the Microarrays Using Lectin− Carbohydrate Recognition. Functionalized biochips were characterized by SPR measurements after injection of different lectins already known to interact specifically with carbohy-



EXPERIMENTAL SECTION Reagents. Monosaccharides, maltose, lithium bromide, sodium hydride, allyl bromide, dimethylformamide, acetonitrile, cysteamine, phosphate buffer, manganese(II) chloride tetrahydrate, sodium chloride, sodium hydroxide, glycerol, sodium dodecyl sulfate (SDS), Tween 20, and agar were purchased from Sigma-Aldrich (Saint-Quentin-Fallavier, France). Pyrrole was purchased from Acros Organics (Geel, Belgium), calcium(II) chloride from Merck (Darmstadt, Germany), tryptic soy broth medium TSB and tryptic soy agar TSA from Fluka (Saint-Quentin-Fallavier, France). UEA I (Ulex Europaeus agglutinin-I) from Ulex europaeus, PNA (peanut agglutinin) from Arachis hypogaea, WGA (wheat germ agglutinin) from Triticum vulgaris, and concanavalin A (ConA) from Canavalia ensiformis (jack bean) were purchased from Sigma-Aldrich (Saint-Quentin-Fallavier, France). E. coli O157:H7 CIP 105917 strain was purchased from the Institut Pasteur (Paris, France). E. coli O157:H7−, E. coli O55:H7, and E. coli O145:H28− strains were given by the French Agency for Food, Environmental and Occupational Health and Safety (ANSES), Food Safety Laboratory (Maisons-Alfort, France). The E. coli K12 B

DOI: 10.1021/ac5037704 Anal. Chem. XXXX, XXX, XXX−XXX

Article

Analytical Chemistry

Figure 1. (A) Structure of the different pyrrole−carbohydrates conjugates. (B) Principle of the fabrication of the carbohydrate microarrays.

drates17,24 in phosphate buffer at 25 °C. We previously performed experiments to ascertain the optimal concentration needed for a good SPRi signal-to-noise ratio. In this study, the concentrations used were 500 nM PNA, 280 nM UEA I, 50 nM WGA, and 1 μM ConA in the presence of Ca2+ and Mn2+ (1 mM). SPRi Kinetic Experiment for the Detection of Bacteria. The freshly diluted bacterial suspension of 102 CFU·mL−1 in TSB medium was deposited in the thermalized (37 °C) SPRi chamber onto the carbohydrate microarray. SPRi kinetic experiment of the bacterial growth was realized by CCM.14,15 SPRi signals were recorded in real time during the culture (1200 min) over the biochip. Data Treatment. For each experiment, an average of SPRi signal coming from quadruplicates was realized. Bulk refractive index changes were not taken into account in the data treatment; therefore, all measured SPRi changes were due to the bacterial−carbohydrate interactions. Then, we defined two typical times, the “detection time” and the “differential time”, reflecting the difference between the detection times for two different carbohydrates. SPRi signals were represented by an inverse tangent function: after smoothing kinetic curves, firstorder derivative of this function was realized. The maximum of the derivative corresponds to the inflection point of kinetic curve called “detection time”. Thus, “detection time” of distinguished and control SPRi signal were measured and the “differential time” was obtained by subtracting these two times (Supporting Information Figure S-1).The SPRi results and their associated error margins presented in this paper are based, at least, on three independent runs.



copolymer was a relevant parameter to study (Supporting Information Figure S-3, Table S-1). The formulation was optimized to 20 mM pyrrole and 10 mM pyrrole−carbohydrate conjugates in spotting buffer. The spots were 260 μm in diameter, and more than 100 spots could be addressed and analyzed on the same biochip. This electrochemical process ensures the gentle and long-lasting grafting of biomolecules on SPR biochips.22 The efficiency of the carbohydrate grafting was checked by SPRi monitoring of the recognition of each carbohydrate by different lectins, i.e., proteins specifically recognizing one (or more) carbohydrate(s).17,25 As expected, SPRi signals (Figure 2) showed an increase in the reflectivity variation correspond-

Figure 2. SPRi signals obtained from the carbohydrate spots after injection of lectins: (A) PNA at 500 nM, (B) UEA I at 280 nM, (C) ConA at 1 μM in the presence of Ca2+ and Mn2+ ions, and (D) WGA at 50 nM, in a phosphate buffer at 25 °C. Lectins were injected at t = 0.

RESULTS AND DISCUSSION

We first synthesized pyrrole−carbohydrate conjugates using simple chemistry in order to immobilize the probes on gold surfaces using electropolymerization.20−23 Then we checked the efficiency of the sugar grafting and their accessibility using lectin-specific recognition. Once these controls were validated, we evaluated the detection and identification of the growth of five different E. coli strains. Conception and Characterization of the Carbohydrate Microarray. Seven pyrrole−carbohydrate conjugates were synthesized (Figure 1A) and were grafted on the gold surface of a SPRi prism using pyrrole coelectropolymerization (Figure 1B). Interactions with carbohydrates are known to be multivalent, and therefore, the density of carbohydrates in the

ing to the recognition of specific spots by lectins. Indeed, the UEA I only interacted with the fucose spots, the PNA with the galactose ones, the WGA with the N-acetylglucosamine and sialic acid spots in a lesser extent, and the ConA with the mannose and maltose spots. In this last test, it is important to note that ConA does not interact with glucose because of the βconfiguration of this carbohydrate on the biochip.26 For ConA, a stabilization of the signal was observed when the protein− carbohydrate interaction reached its equilibrium. For the other lectins, the signals decreased slightly after reaching a maximum. This corresponds to the dynamics of carbohydrate−lectin interactions.17 C

DOI: 10.1021/ac5037704 Anal. Chem. XXXX, XXX, XXX−XXX

Article

Analytical Chemistry

bacterial strains, the SPRi signals presented an exponential growth which is in agreement with the kinetics of bacterial growth. It sometimes occurred in two phases: (i) a small plateau, observed for E. coli O157:H7, or a shoulder, observed for E. coli O157:H7−, O55:H7, and O145:H28− and (ii) a second plateau reached after a second exponential growth. At the end of SPRi acquisition, the biochips were completely recovered by a bacterial biofilm. It seems that the shoulder corresponds to the bacterial stationary phase. The second exponential growth could be attributed to the growth of the biofilm until it reaches its maximal thickness, due to the limitation of nutrients in the medium, or to the SPRi dynamic range limit. When we compare the signals obtained for a given bacteria on the different spots of the biochip, we see that, after one overnight incubation with E. coli O55:H7 or E. coli K12 culture (Figure 4, parts B and D, respectively), SPRi revealed a homogeneous growth of bacteria in the medium over the biochip without any particular interaction between any carbohydrate and the strains. On the contrary, in the case of E. coli O157:H7 (Figure 4C), the percentage of reflectivity variation ΔR(%) increased significantly earlier onto galactose plots (red curve) than on the other carbohydrates. This means that galactose interacts more specifically to E. coli O157:H7. In the case of E. coli O157:H7− (Figure 4E), an earlier SPRi signal was obtained for the bacterial interaction with sialic acid (dark blue curve) but not with galactose. In the case of E. coli O145:H28− (Figure 4F) mannose (light blue curve), galactose, and sialic acid interacted earlier with bacteria than the other carbohydrates. Therefore, the carbohydrate biochip we conceived allows the qualitative distinction between these closely related strains. The reflectivity amplitudes were variable from one bacterium to the other because they depend on the number of bacteria close to the surface and also on the contribution of one bacterium to the SPRi signal (particularly if bacteria interact through their pili or core bacterial surface receptors). In the cases of E. coli O55:H7 and E. coli O157:H7, without taking into account any particular SPRi signal, the global SPRi reflectivity was very low (below 20%), whereas for E. coli K12, E. coli O157:H7− and E. coli O145:H28−, the reflectivity was higher (around 40−60%), suggesting that SPRi amplitude could not be used as a relevant criterion for discriminating the interactions of those bacteria with carbohydrates. In order to evaluate rigorously these interactions, we chose as a test criterion the time scale and defined two typical times: the “detection time” and the “differential time”. The detection time was measured at the inflection point of the kinetic curves (see Experimental Section for details). We pointed out the differential time between curves that distinguished among the others and the control curves (i.e., polypyrrole and the other carbohydrates). Differential time is then the difference between two detection times. Data are summarized in Table 1 and reveal the reproducible interactions between three different strains and particular carbohydrates. E. coli O157:H7 interacted systematically with galactose at 529 ± 26 min, that is to say 76 ± 32 min before the other tested carbohydrates. Sialic acid interacted particularly, 47 ± 20 min earlier, with E. coli O157:H7− when compared to the controls. Specific interactions of galactose and mannose with E. coli O145:H28− were correlated with an earlier detected exponential growth (78 ± 15 and 53 ± 11 min, respectively) than with the other carbohydrates. Affinity of E.

Finally, the here-described biochips showed an efficient grafting of all the carbohydrates and revealed their accessibility on the polymer surface toward lectins and then possibly toward bacteria. Regenerability of the Biochips. After each SPRi experiment, microarrays could be regenerated with soft washing with SDS 2% followed by sodium hydroxide (0.02 M), which are common reactants in biomedical laboratories. After deposition of lectins or bacteria, the microarrays were again characterized by SPRi measurements after injection of different lectins in phosphate buffer. Biochips exhibited the same lectin− carbohydrate binding specificity indicating that lectin− carbohydrate or bacteria−carbohydrate interactions did not alter the carbohydrate microarray and that immobilized carbohydrates were not nutrients for bacteria. The biochips could be used at least five times without any loss of performance in bacterial detection. Real-Time Monitoring of the Interactions between Carbohydrates and Bacteria during the Culture. We have investigated the interactions of various E. coli bacterial strains with the carbohydrates grafted on the microarrays during the culture of bacteria by CCM (Figure 3A).14,15 We chose to work

Figure 3. (A) Schematic view of the SPRi culture−capture−measure method and (B) standard kinetics of bacterial growth.

with four pathogenic bacterial strains: E. coli O157:H7, E. coli O157:H7−, a subserotype without flagella, E. coli O145:H28−, a close phenotype, and E. coli O55:H7, an ancestor of O157 and O148 strains. The nomenclature O refers to the O-antigen located in the lipopolysaccharide in the external membrane, and the nomenclature H refers to the presence of an H-antigen (H7 is a variant of this antigen) on the flagella or to the absence of flagella (symbolized by a minus: H7−, H28−). As a control, nonpathogenic E. coli K12 was selected. SPRi plots reflect the global change of the refractive index in the medium which results from the presence of bacteria 100− 300 nm above the gold surface (i.e., length of the evanescent wave interacting with the gold surface plasmon).17 Then, the exponential growth of SPRi reflectivity observed derives from the bacterial growth in the vicinity of spots. Experimental kinetic curves were similar to standard bacteria growth (Figure 3B) occurring in three different stages: first, the lag phase, and finally the stationary phase (growth stops because of the lack of nutrients and overpopulation). Figure 4 shows the kinetic curves corresponding to the SPRi signals collected from the carbohydrate biochip (average signals of quadruplicates) after deposition of the bacterial suspension at 102 CFU·mL−1 in TSB. Concerning the general shape of the SPRi plots, for all the D

DOI: 10.1021/ac5037704 Anal. Chem. XXXX, XXX, XXX−XXX

Article

Analytical Chemistry

Figure 4. SPRi signals obtained from the spots of the carbohydrate microarray, after deposition of the bacterial suspension at 102 CFU·mL−1 in TSB, in the case of (A) buffer (control), (B) E. coli O55:H7, (C) E. coli O157:H7, (D) E. coli K12, (E) E. coli O157:H7−, with focus on the beginning of the interactions, and (F) E. coli O145:H28−, with focus on the beginning of the interactions.

Table 1. Detection Times and Differential Times between Distinguished SPRi Signals and the Other Ones at the Inflection Point of SPRi Signals Obtained, after Deposition of the Bacterial Suspension at 102 CFU·mL−1 in TSBa bacterial strains

a

E. coli O157:H7

E. coli O157:H7−

interacting carbohydrate

galactose

sialic acid

mannose

E. coli O145:H28− galactose

sialic acid

detection time (min) differential time/control spots (min)

529 ± 26 76 ± 32

295 ± 7 47 ± 20

448 ± 8 78 ± 15

473 ± 4 53 ± 11

494 ± 21 32 ± 17

Data were obtained from the SPRi signal averages of three independent runs.

Table 2. Fingerprint of the Carbohydrate Microarray Towards Five E. coli Bacteriaa bacterial strains

a

carbohydrates

E. coli K12

E. coli O157:H7

E. coli O157:H7−

E. coli O55:H7

E. coli O145:H28−

Glc, Malt, Fuc, GlcNAc Gal Man sialic acid

− − − −

− ++ − −

− − − +

− − − −

− ++ ++ +

++, differential time >50 min; +, 50 min > differential time >5 min; −, differential time