Journal of Applied Microbiology ISSN 1364-5072
ORIGINAL ARTICLE
Surface plasmon resonance-based bacterial aerosol detection E.V. Usachev1, O.V. Usacheva2 and I.E. Agranovski1 1 Griffith School of Engineering, Griffith University, Brisbane, Qld, Australia 2 Department of molecular genetics, The D.I. Ivanovsky Institute of Virology of The Ministry of Health and Social Development of The Russian Federation, Moscow, Russia
Keywords bioaerosol detection, E. coli, first alert bioaerosol monitoring device, personal sampler, surface plasmon resonance. Correspondence Igor E. Agranovski, Griffith School of Engineering, Griffith University, Brisbane, 4111 Queensland, Australia. E-mail: [email protected] 2014/1273: received 23 June 2014, revised 15 August 2014 and accepted 27 August 2014 doi:10.1111/jam.12638
Abstract Aims: In the area of bioaerosol research, rapid methods for precise detection attracted much interest over last decades. One of such technologies operating in nearly real-time mode without any specific labelling is known as surface plasmon resonance (SPR). Recently, we validated a SPR protocol in conjunction with our earlier developed personal bioaerosol sampler for rapid detection of airborne viruses. Considering that the biological interaction between targeted micro-organism and corresponding antibody is strongly related to sizes of targeted micro-organisms, this research is vital validating suitability of SPR technique for bacterial aerosol detection, as characteristic size of bacteria is 2–3 orders of magnitude larger than sizes of common viruses. The combination of SPR with portable air sampling instrumentation could lead to the development of portable bioaerosol monitor. Methods and Results: This study is focussed on the SPR technology application for direct detection of common environmental bacterial strain— Escherichia coli. The detection limit of developed SPR techniques based on utilization of a planar gold sensor chip functionalized with polyclonal antibody via NeutrAvidin junction for sensing of bacterial cells was found to be 15 9 103 CFU ml1, which corresponds to the limit of detection in the air to be 219 9 104 CFU l1 for 1 min of sampling time. Conclusions: The technology was found fully suitable for rapid and reliable detection of large size bacterial aerosols. Low magnitude of the limit of detection looks very promising for sensitive detection of highly pathogenic airborne bacteria in the ambient air. Significance and Impact of the Study: The suggested technology based on a simple model organism is one of the first attempts to develop a real-time monitor for reliable detection of airborne bacteria. The outcomes would be of strong interest of professionals involved in monitoring and/or control of pathogenic airborne bacteria, including Legionella, Mycobacterium tuberculosis and Bacillus anthracis.
Introduction Bioaerosols originated from natural or artificial environments could potentially posses significant hazard for public health. The bioaerosol constituents may include bacteria, fungi, viruses, allergens, bacterial endotoxins and Journal of Applied Microbiology © 2014 The Society for Applied Microbiology
other components (Douwes et al. 2003). As a result of exposure to bioaerosols, the health problems arise, including infectious diseases (Ayres et al. 2009; PerezPadilla et al. 2009), organic dust toxic syndrome (Kabir et al. 2012), allergies (Cullinan et al. 2000) and other impairments (Trout et al. 2001). The bioaerosol assessment 1
Real-time bacterial aerosol detection
methodology for rapid and precise airborne pathogen detection in different environments has received considerable attention in recent years. One of the main parts for bioaerosol detection is representative air sampling. Previously developed air sampling technology (Agranovski et al. 1999), based on a method involving air bubbling through porous medium, submerged into a liquid layer was realized in a personal bioaerosol sampler (Agranovski et al. 2002). The sampler performance was validated for representative monitoring of airborne fungi, bacteria and viruses (Agranovski et al. 2004a, 2005a,b). Collected bioaerosol particles could be analysed by traditional time-consuming culture-based methods. However, for the last decades, a significant scientific interest was attracted towards development of methods for nearly real-time continuous detection and quantification of biological airborne particles. A number of various detection approaches and instruments have been developed, including ultraviolet aerodynamic particle sizer spectrometer (UVAPS), bioaerosol mass spectrometer (BAMS), ATP bioluminescence-based method and nanoPCR technique (Agranovski et al. 2004b; Russell et al. 2005; Xu and Yao 2013 and Park et al. 2014). Surface plasmon resonance (SPR)-based technique, representing a group of immunoreaction-based molecular techniques, enables data acquisition in nearly real-time mode. Moreover, this technology allows label-free and selective analyte detection. The SPR is an optical phenomenon where SPR sensors detect the refractive index (RI) changes due to analyte mass accumulation on the surface. The key role in the SPR detection is the ligand/analyte pair. The target analytes such as proteins, small organics, microbes, viruses or toxins are reacting with specific ligand (e.g. antibodies) immobilized on the sensor surfaces (Homola et al. 2005; Singh and Hiller 2006). As the SPR technology was recognized by scientific world as a powerful molecular tool, a number of SPR-based instruments and sensors have been developed. The Spreeta SPR biosensors developed by Texas Instruments are unique in that all of the optoelectronic components are incorporated into a small moulded plastic chip of a fingertip size. The Spreeta-based sensors’ noise level and resolution are comparable with other available SPR brands (Chinowsky et al. 2003). In the area of bioaerosol detection, such miniature technology is highly attractive because it enables portable bioaerosol monitor development. Bacterial detection from different sources by SPR was intensively studied over the last decade. Different foodborne pathogens such as Escherichia coli O157:H7, bacteria from genus Salmonella, Yersinia, Listeria and others were successfully detected by the SPR in direct assay format (Oh et al. 2005; Hearty et al. 2006; Waswa et al. 2006, 2
E.V. Usachev et al.
2007; Lan et al. 2008). However, the data reported in the literature are quite contradictive with the sensitivity limit ranging from 101 to 105 CFU ml1 for bacterial targets. In addition, very limited data are available for the SPR application in bioaerosol research. Recently, we have shown applicability of the SPR for viral aerosol detection on the MS2 bacteriophage commonly used in bioaerosol research. The sensitivity of the procedure for 1 min of sampling was about 107 PFU per litre of air and slightly varied for different sensor types used (Usachev et al. 2013a,b, 2014). This study is a logical continuation of our previous evaluation of the SPR technology applicability in conjunction with personal bioaerosol sampler for real-time bioaerosol detection. The aim of the present work was a conceptual evaluation of the SPR technology application to the direct bacterial aerosol detection on the common environmental bacterial model—E. coli. Special attention was given to the SPR procedure sensitivity of intact and physically stressed bacteria. Based on the nature of biological interaction between targeted micro-organism and corresponding antibody, capturing efficiency is strongly related to the size of targeted micro-organism. On this basis, an additional research was required to validate suitability of SPR technique for bacterial aerosol detection, as characteristic size of bacteria (in order of 1 lm) is 2–3 orders of magnitude larger than the size of common viruses, causing significantly different interaction compared to antibody-virus pair interaction. Materials and methods Cells The E. coli bacterium commonly presented in the environment was used in this research. E. coli K12 strain was obtained from the Southern Biological (Australia). Bacterial culture was maintained on nutrient agar (OXOID, England) plates. Cultures for the assays were grown overnight in 2YT medium broth (Bacto Tryptone 16 g; Yeast Extract 10 g; NaCl 5 g; Distilled water 1 l) at 37° C with aeration. Then, cells were collected and washed three times with phosphate-buffered solution (PBS, pH = 74). The bacterial samples were made out of stock suspension by serial decimal dilutions in sterile PBS. The number of colony-forming units (CFU) for each dilution was confirmed by plating 100 ll of diluted sample onto the nutrient agar plates and incubated overnight at 37°C. The colonies formed on the plates were then counted. Sensor preparation Based on data for E. coli detection from different sources (Dudak and Boyaci 2007; Waswa et al. 2007), the planar Journal of Applied Microbiology © 2014 The Society for Applied Microbiology
E.V. Usachev et al.
Real-time bacterial aerosol detection
injected at a flow rate of 15 ll min1 followed by 120 s of chip washing by running buffer solution. Then, the sensor regeneration was undertaken by 50 ll of regeneration solution injected at 50 ll min1 flow rate. Finally, the sensor and injection port were washed/cleaned over 120 s interval by running buffer solution. In each assay, the first analyte (PBS) injection served as a blank (zero) reference. The bacterial samples were tested in each assay in order of increased concentrations. For the sensor’s regeneration purpose, three types of regeneration solution were tested. The first solution was 50 mmol orthophosphoric acid in water (Smith et al. 2009), the second was 100 mmol NaOH with 1% Triton X-100 (Waswa et al. 2007), and the third was 1% SDS (Dudak and Boyaci 2007). All data acquired from the SPR assays were analysed using Qdat software ver. 1.0.0.24 (ICx Nomadics, Inc., Stillwater, OK).
gold Spreeta chip was employed. The chip surface functionalization was performed by adsorption of NeutrAvidin molecules onto the gold film with consequent irreversible binding with biotinylated rabbit polyclonal anti-E. coli antibodies. To reduce a nonspecific signal, the 2% BSA was used to saturate unoccupied gold surface. In this study, the sensor preparation and all SPR analyses were conducted at 25°C using a dual-channel, semiautomated SPR platform SensıQ (SensıQ Technologies, Inc., Oklahoma City, OK). The Spreeta-based planar gold sensor chip SS01 from SensıQ Technologies was first briefly cleaned with absolute ethanol and dried with airstream before assembly. The cleaned sensor was installed into a flow cell, and the system was primed with running buffer (PBS) and normalized by the software (ver. A.05). The sensor functionalization was performed in a single run with running buffer at a flow rate of 50 ll min1. Firstly, the gold surface was cleaned by injecting 100 ll of cleaning solution (100 mmol NaOH with 1% Triton X-100 in water) as depicted in segment 1 of Fig. 1. This was followed by rinsing with DI water (segment 2 in Fig. 1). The sensing channel was saturated by 150 ll of NeutrAvidin with the concentration of 100 lg ml1 discretely injected by a flow rate of 5 ll min1. To block unoccupied sites on a sensor surface, both channels were saturated with 150 ll of 2% BSA (bovine serum albumin) in PBS at the same flow rate. Finally, the sensing channel was charged with biotinilated rabbit polyclonal antiE. coli antibody (Abcam, Cambridge, UK) by discrete injection of 150 ll (100 lg ml1) at 5 ll min1.
Experimental setup Bacterial aerosols were generated by a 3 jet Collison nebulizer (BGI, Inc., Waltham, MA) from cells suspension and supplied into a rotating aerosol chamber. To achieve appropriate (detectable) bioaerosol concentration in the chamber, the initial bacterial suspension with the concentration of 95 9 106 CFU ml1 was used for nebulization. The nebulizer was operated at a flow rate of 6 l min1 of dry and filtered compressed air. The air temperature and humidity were continuously monitored and controlled throughout the entire procedure (T = 23–25°C and RH = 45–55%). To trace any unfavourable variations in aerosol supply, a real-time optical aerosol spectrometer (Model 4705, Aeronanotech, Moscow, Russia) was used. The rotating aerosol chamber used in this study was described in detail in our previous publications (Pyankov
SPR assay All SPR experiments were carried out in an assay format as per following protocol. After 120 s of the sensor equilibration/normalization step, 100 ll of the analyte was 4000
R
3500
Response, RU
3000 1640
2500 R
2000 1500
BSA
2090 RU
1000 500 Figure 1 The planar gold surface chip functionalizing with rabbit polyclonal antiEscherichia coli antibody. 1—washing solution; 2—DI water; R—running buffer.
RU
R
R 1
0 0
Journal of Applied Microbiology © 2014 The Society for Applied Microbiology
R
Antibody
NeutrAvidin
2 500
1000
1500
2000
2500
3000
3500
4000
4500
Time, s
3
Real-time bacterial aerosol detection
E.V. Usachev et al.
et al. 2012; Usachev et al. 2013a,b). Three ‘bubbler’ type personal bioaerosol samplers (Agranovski et al. 2005a) were charged with 50 ml of collecting liquid each (PBS, pH = 74) and stationary installed inside the chamber in a specially designed holder. For each experimental run, the nebulizer was charged with bacterial suspension and operated to produce relatively stable aerosol concentration in the chamber as was continuously monitored by the optical aerosol spectrometer (approx. 3 min were required to equilibrate concentration in the chamber). Then, the personal bioaerosol samplers were switched on and operated at the airflow rate of 4 l min1. Aliquots of collecting liquid from the samplers (100 ll) were acquired from all three devices after 1, 10 and 25 min time periods and analysed by the SPR procedure as described above. The final results were corrected to account for sampling related dilution of the air in the chamber (12 l min1) by HEPA (high-efficiency particulate air filter HEPA-CUP 75; GE Healthcare, RYDALMERE, NSW, Australia)- filtered ambient air. For the material balance of the bacteria being sampled at particular time periods, the fluorescent dye (fluorescein sodium salt, C20H10Na2O2, Fluka AG, Buchs, Switzerland) with the concentration of 05 mg ml1 in the nebulizer (corresponding to 1326 9 108 FU ml1) was used. The procedure is described in detail by Usachev et al. (2014) for the viral aerosol study. Briefly, the ratio of the bacterial concentration (expressed in CFU ml1) by the sample fluorescence intensity (expressed in arbitrary fluorescence units FU ml1) of the initial suspension in the nebulizer has been established as 716 9 102 CFU FU1. That enables estimating the total amount of airborne bacteria collected for each particular experimental run to be identified by measuring the fluorescence intensity of an aliquot of the collecting liquid acquired from corresponding samplers at required time periods.
Results In the present study the SPR technology was evaluated for direct bacterial aerosol detection. Figure 1 illustrates the sensor functionalization process. The fresh sensor was installed into SensiQ instrument, then primed with running buffer (PBS) and normalized according to the manufacturer’s recommendations. After rinsing of the sensor’s channels with regenerating solution (100 mmol NaOH and 1% Triton X-100) and deionized water, the NeutrAvidin was discretely injected into the sensing channel. The surface saturation was observed at 2090 RU level (arbitrary unit of the machine’s signal response). The unoccupied surface blockage was achieved by passing BSA solution through both channels (2% BSA in PBS). Finally, the sensing channel was charged with antibody at a level of 1640 RU. Testing of different regeneration solutions revealed the optimal (stable results with high reproducibility over 50 runs) solution for particular settings to be 100 mmol NaOH and 1% Triton X-100. The sensing range and detection limit of the technique were evaluated. Overnight cell culture was washed thrice with PBS. Serial tenfold dilutions of bacterial cells in PBS were prepared and tested in the assay format for 100 ll analyte injections with the flow rate of 15 ll min1 following regeneration by injection of 50 ll of previously described regeneration solution over 1-min time period. Each bacterial dilution was tested in triplicate. As is seen from Fig. 2, the response was increased with an increase of analyte concentration. All plotted curves represent the difference between the reference and sensing channels responses. Besides, in the assay format, all curves were referenced against responses obtained for pure PBS, which was taken as zero. Based on the response data for 10-second dissociation step, the analyte detection was in the range of four orders of magnitude from 15 9 104 to 15 9 108 CFU ml1 (Fig. 3). The error bars represent
140 1·5E+8
Response, RU
120 100 80 60 1·5E+7 40 20 0 0
100
200
300 Time, s
4
400
1·5E+6 1·5E+5 1·5E+4 500
Figure 2 The raw data obtained from detection of Escherichia coli decimal dilutions. The bacterial concentrations are expressed in CFU ml1. Running buffer was used as zeroed negative control. Each dilution was tested in triplicate, but only 1 curve per dilution is plotted.
Journal of Applied Microbiology © 2014 The Society for Applied Microbiology
E.V. Usachev et al.
Real-time bacterial aerosol detection
80
140 120
y = –1E–14x2 + 3E–06x + 4·8637 R2 = 0·9974
70 60 Response, RU
100 Response, RU
y = –3E–15x2 + 1E–06x + 3·6177 R2 = 0·9919
80 60 40
50 40 30 20
20 10 1E+4
1E+7 1E+5 1E+6 1E+8 Bacterial concentration, CFU ml–1
1E+9
Figure 3 Intact Escherichia coli cells detection range. Measurement results of the bacterial binding (mean values of 10 s dissociation step) for three repeats vs bacterial decimal dilutions. Polynomial trend line with equation is plotted.
standard deviations of at least 3 measurements undertaken for each dilution. A single-factor ANOVA test revealed statistically significant difference in response to the bacterial concentrations tested (q < 005). A high correlation (R2 value for a quadratic fit) between the average normalized response values and bacterial concentration was observed with the minimal detectable concentration of 15 9 104 CFU ml1 (confirmed by 5 replication test). Also, due to fast chip surface saturation occurring at high microbial concentrations in the liquid, the correlation of the results at >108 CFU ml1 becomes poor. The effect of the physical freeze/thawing stress onto sensitivity of direct bacteria SPR detection was investigated. Here, the bacterial suspension was washed with PBS and then several times freezed (at 20°C) and thawed (at +37°C) following production of serial tenfold dilutions on PBS for SPR assay. As is seen from Fig. 4, the response values were highly correlated with bacterial concentration growth (R2 = 099). The trend is almost perfectly quadratic fitted. Each dilution was tested in triplicate and standard deviations were estimated. The statistical significance of the response difference between each dilution points was confirmed with a single-factor ANOVA test (q < 005). Because of this physical treatment, the sensitivity of SPR detection was improved by one order of magnitude with the LOD corresponding to 15 9 103 CFU ml1. Figure 5 shows a comparison of the binding of nontarget bacteria (Bacillus subtilis, Pseudomonas sp.), viruses Journal of Applied Microbiology © 2014 The Society for Applied Microbiology
0 1E+2
1E+3
1E+4
1E+5
1E+6
1E+7
1E+8
Bacterial concentration, CFU ml–1 Figure 4 Physically stressed Escherichia coli cells detection range. Measurement results of the bacterial binding (mean values of 10-s dissociation step) for three repeats vs bacterial decimal dilutions. Polynomial trend line with equation is plotted.
Response, RU
0 1E+3
90 80 70 60 50 40 30 20 10 0 0
E· coli
non E· coli 100
200 300 Time, s
400
500
Figure 5 Selectivity test of the SPR sensor. Non-Escherichia coli — MS2 and T4 bacteriophages, Pseudomonas sp. and Bacillus subtilis. All microbes at concentrations above 107 CFU or PFU per ml, respectively.
(T4 and MS2 bacteriophages) and E. coli K12 strain to the antibodies against E. coli. The association and dissociation steps presented clearly show the specific immunoreaction of antibody/antigen pair used in this study. All microbes were used at concentration higher than 107 CFU or PFU per ml. All data were referenced against the blank injection of running buffer, as it was used as analyte diluent. Besides, the specificity of the antibodies used was also confirmed by ELISA assay. Finally, we challenged the SPR-based approach for direct bacterial aerosol detection. The bacteria (washed with PBS) were aerosolized and sampled over 25-min 5
Real-time bacterial aerosol detection
25
10
6
1
4 2 0 0
50 100 150 200 250 300 350 400 450 Time, s
Figure 6 The raw data obtained from SPR detection of Escherichia coli aerosol samples collected for 1, 10 and 25 min. Sampling liquid (PBS) was used as zeroed negative control.
period. The results obtained for bacterial aerosols collected by the personal sampler and analysed by the SPR technique are shown in Fig. 6. Here, response curves represent the E. coli cells binding with immobilized antiE. coli antibody for 1, 10 and 25 min of the sampler operation for collection of airborne micro-organisms. As is seen in Fig. 7, the SPR response was found to be in excellent correlation with the sampling time. The R2 value for a logarithmic fit was about 0999. In particular, highly consistent results were obtained for two longer sampling time periods, where standard deviation did not exceed 10%. The ratio of the bacteria to the fluorescence in the nebulizer suspension was found to be 716 9 102 CFU FU1. The fluorescence of liquid corresponding to 1-min sampling time was found to be 2445 9 104 FU ml1. Based on this finding, the minimal detectable concentration of bacteria collected for 1 min of air sampling was calculated to be 175 9 103 CFU ml1, which corresponds to the LOD of airborne bacteria at around 219 9 104 CFU per litre of air. Discussions and conclusions An approach similar to one described by Waswa et al. (2007) was used in this project for the SPR sensor functionalization. In addition, a referencing channel saturated with bovine serum albumin to avoid nonspecific sorption on the gold surface (Fig. 1) was used. The results obtained for charging efficiency were comparable with the previously reported ones for high capacity COOH5 sensor type (Usachev et al. 2014). Three chip regeneration solutions were initially challenged and the best reproducibility of the results was obtained for the solution described by Waswa et al. (2007) with slightly reduced NaOH concentration. In contrast, use of other two solutions led to significant 6
10
10
8
Response,RU
Response, RU
E.V. Usachev et al.
5 y = 1·3731In(x) + 5·7821 R 2 = 0·9993 0
0
5
10
15
20
25
Sampling time, min Figure 7 Bacterial aerosol SPR detection over 25 min sampling time. Escherichia coli concentration in nebulizer was 95 9 106 CFU ml1. Error bars represent standard deviation of at least 3 experimental runs.
losses of the SPR response even after very few runs. These findings could be explained by stability of the antibodies used in this project for particular regeneration settings. The sensitivity of the technique for detection of micrometre-sized microbial bioaerosol was firstly explored by using decimal dilutions of intact bacterial cells, which were minimally treated by thrice washing with PBS only. The LOD of the technique was found to be approx. 15 9 104 CFU ml1 (Fig. 3). This finding is in good agreement with the results presented by Leonard et al. (2004) (105 CFU ml1) and quite distant from the outcomes reported by Dudak and Boyaci (2007) (101 CFU ml1), Waswa et al. (2007) (102 CFU ml1) and Subramanian et al. (2006) (103 CFU ml1). This discrepancy could be explained by some difference in instrumentation involved and ligand/analyte pair properties. In addition, comparing present results with our previous data reported for viral bioaerosol detection, it is evidently seen that the sensitivity of the approach is much higher for larger sized analyte corresponding to bacterial strains. As was reported by Taylor et al. (2006), some decrease of lower limit of detection (LOD) of bacterial strains by SPR instruments could be achieved by optimization of the sample preparation step. It was evidently shown that bacteria treated with detergent lysing solution were detected at two orders of magnitude lower LOD. Considering that after aerosolization bacteria could change its morphological and immunogenic properties due to substantial physical stress occurring in the air environment (desiccation, high/low temperature, UV radiation, etc.), we investigated the effect of the physical freeze/thawing stress onto sensitivity of direct bacteria SPR detection. Our data revealed the LOD improvement by one order of magnitude (Fig. 4). The difference in the results for untreated and stressed bacteria samples could be Journal of Applied Microbiology © 2014 The Society for Applied Microbiology
E.V. Usachev et al.
explained by possible cell lysis as the result of treatment. In this case, the analyte is changing its morphological and possibly immunogenic properties. Moreover, the SPR sensitivity difference for intact and treated cells could be also explained by the polyclonal nature of the antibodies used. The treatment of complex target (whole bacterial cell) could alternate active immunoreacting pairs. In the previous studies, we have shown applicability of SPR approach in conjunction with the personal bioaerosol sampler for the detection of airborne virus (Usachev et al. 2013a,b, 2014). The combined sampling and detection technology was capable of providing selective and almost real-time detection of airborne virus in a manner of minutes. The high viral concentrations in the air could be detected in
ORIGINAL ARTICLE
Surface plasmon resonance-based bacterial aerosol detection E.V. Usachev1, O.V. Usacheva2 and I.E. Agranovski1 1 Griffith School of Engineering, Griffith University, Brisbane, Qld, Australia 2 Department of molecular genetics, The D.I. Ivanovsky Institute of Virology of The Ministry of Health and Social Development of The Russian Federation, Moscow, Russia
Keywords bioaerosol detection, E. coli, first alert bioaerosol monitoring device, personal sampler, surface plasmon resonance. Correspondence Igor E. Agranovski, Griffith School of Engineering, Griffith University, Brisbane, 4111 Queensland, Australia. E-mail: [email protected] 2014/1273: received 23 June 2014, revised 15 August 2014 and accepted 27 August 2014 doi:10.1111/jam.12638
Abstract Aims: In the area of bioaerosol research, rapid methods for precise detection attracted much interest over last decades. One of such technologies operating in nearly real-time mode without any specific labelling is known as surface plasmon resonance (SPR). Recently, we validated a SPR protocol in conjunction with our earlier developed personal bioaerosol sampler for rapid detection of airborne viruses. Considering that the biological interaction between targeted micro-organism and corresponding antibody is strongly related to sizes of targeted micro-organisms, this research is vital validating suitability of SPR technique for bacterial aerosol detection, as characteristic size of bacteria is 2–3 orders of magnitude larger than sizes of common viruses. The combination of SPR with portable air sampling instrumentation could lead to the development of portable bioaerosol monitor. Methods and Results: This study is focussed on the SPR technology application for direct detection of common environmental bacterial strain— Escherichia coli. The detection limit of developed SPR techniques based on utilization of a planar gold sensor chip functionalized with polyclonal antibody via NeutrAvidin junction for sensing of bacterial cells was found to be 15 9 103 CFU ml1, which corresponds to the limit of detection in the air to be 219 9 104 CFU l1 for 1 min of sampling time. Conclusions: The technology was found fully suitable for rapid and reliable detection of large size bacterial aerosols. Low magnitude of the limit of detection looks very promising for sensitive detection of highly pathogenic airborne bacteria in the ambient air. Significance and Impact of the Study: The suggested technology based on a simple model organism is one of the first attempts to develop a real-time monitor for reliable detection of airborne bacteria. The outcomes would be of strong interest of professionals involved in monitoring and/or control of pathogenic airborne bacteria, including Legionella, Mycobacterium tuberculosis and Bacillus anthracis.
Introduction Bioaerosols originated from natural or artificial environments could potentially posses significant hazard for public health. The bioaerosol constituents may include bacteria, fungi, viruses, allergens, bacterial endotoxins and Journal of Applied Microbiology © 2014 The Society for Applied Microbiology
other components (Douwes et al. 2003). As a result of exposure to bioaerosols, the health problems arise, including infectious diseases (Ayres et al. 2009; PerezPadilla et al. 2009), organic dust toxic syndrome (Kabir et al. 2012), allergies (Cullinan et al. 2000) and other impairments (Trout et al. 2001). The bioaerosol assessment 1
Real-time bacterial aerosol detection
methodology for rapid and precise airborne pathogen detection in different environments has received considerable attention in recent years. One of the main parts for bioaerosol detection is representative air sampling. Previously developed air sampling technology (Agranovski et al. 1999), based on a method involving air bubbling through porous medium, submerged into a liquid layer was realized in a personal bioaerosol sampler (Agranovski et al. 2002). The sampler performance was validated for representative monitoring of airborne fungi, bacteria and viruses (Agranovski et al. 2004a, 2005a,b). Collected bioaerosol particles could be analysed by traditional time-consuming culture-based methods. However, for the last decades, a significant scientific interest was attracted towards development of methods for nearly real-time continuous detection and quantification of biological airborne particles. A number of various detection approaches and instruments have been developed, including ultraviolet aerodynamic particle sizer spectrometer (UVAPS), bioaerosol mass spectrometer (BAMS), ATP bioluminescence-based method and nanoPCR technique (Agranovski et al. 2004b; Russell et al. 2005; Xu and Yao 2013 and Park et al. 2014). Surface plasmon resonance (SPR)-based technique, representing a group of immunoreaction-based molecular techniques, enables data acquisition in nearly real-time mode. Moreover, this technology allows label-free and selective analyte detection. The SPR is an optical phenomenon where SPR sensors detect the refractive index (RI) changes due to analyte mass accumulation on the surface. The key role in the SPR detection is the ligand/analyte pair. The target analytes such as proteins, small organics, microbes, viruses or toxins are reacting with specific ligand (e.g. antibodies) immobilized on the sensor surfaces (Homola et al. 2005; Singh and Hiller 2006). As the SPR technology was recognized by scientific world as a powerful molecular tool, a number of SPR-based instruments and sensors have been developed. The Spreeta SPR biosensors developed by Texas Instruments are unique in that all of the optoelectronic components are incorporated into a small moulded plastic chip of a fingertip size. The Spreeta-based sensors’ noise level and resolution are comparable with other available SPR brands (Chinowsky et al. 2003). In the area of bioaerosol detection, such miniature technology is highly attractive because it enables portable bioaerosol monitor development. Bacterial detection from different sources by SPR was intensively studied over the last decade. Different foodborne pathogens such as Escherichia coli O157:H7, bacteria from genus Salmonella, Yersinia, Listeria and others were successfully detected by the SPR in direct assay format (Oh et al. 2005; Hearty et al. 2006; Waswa et al. 2006, 2
E.V. Usachev et al.
2007; Lan et al. 2008). However, the data reported in the literature are quite contradictive with the sensitivity limit ranging from 101 to 105 CFU ml1 for bacterial targets. In addition, very limited data are available for the SPR application in bioaerosol research. Recently, we have shown applicability of the SPR for viral aerosol detection on the MS2 bacteriophage commonly used in bioaerosol research. The sensitivity of the procedure for 1 min of sampling was about 107 PFU per litre of air and slightly varied for different sensor types used (Usachev et al. 2013a,b, 2014). This study is a logical continuation of our previous evaluation of the SPR technology applicability in conjunction with personal bioaerosol sampler for real-time bioaerosol detection. The aim of the present work was a conceptual evaluation of the SPR technology application to the direct bacterial aerosol detection on the common environmental bacterial model—E. coli. Special attention was given to the SPR procedure sensitivity of intact and physically stressed bacteria. Based on the nature of biological interaction between targeted micro-organism and corresponding antibody, capturing efficiency is strongly related to the size of targeted micro-organism. On this basis, an additional research was required to validate suitability of SPR technique for bacterial aerosol detection, as characteristic size of bacteria (in order of 1 lm) is 2–3 orders of magnitude larger than the size of common viruses, causing significantly different interaction compared to antibody-virus pair interaction. Materials and methods Cells The E. coli bacterium commonly presented in the environment was used in this research. E. coli K12 strain was obtained from the Southern Biological (Australia). Bacterial culture was maintained on nutrient agar (OXOID, England) plates. Cultures for the assays were grown overnight in 2YT medium broth (Bacto Tryptone 16 g; Yeast Extract 10 g; NaCl 5 g; Distilled water 1 l) at 37° C with aeration. Then, cells were collected and washed three times with phosphate-buffered solution (PBS, pH = 74). The bacterial samples were made out of stock suspension by serial decimal dilutions in sterile PBS. The number of colony-forming units (CFU) for each dilution was confirmed by plating 100 ll of diluted sample onto the nutrient agar plates and incubated overnight at 37°C. The colonies formed on the plates were then counted. Sensor preparation Based on data for E. coli detection from different sources (Dudak and Boyaci 2007; Waswa et al. 2007), the planar Journal of Applied Microbiology © 2014 The Society for Applied Microbiology
E.V. Usachev et al.
Real-time bacterial aerosol detection
injected at a flow rate of 15 ll min1 followed by 120 s of chip washing by running buffer solution. Then, the sensor regeneration was undertaken by 50 ll of regeneration solution injected at 50 ll min1 flow rate. Finally, the sensor and injection port were washed/cleaned over 120 s interval by running buffer solution. In each assay, the first analyte (PBS) injection served as a blank (zero) reference. The bacterial samples were tested in each assay in order of increased concentrations. For the sensor’s regeneration purpose, three types of regeneration solution were tested. The first solution was 50 mmol orthophosphoric acid in water (Smith et al. 2009), the second was 100 mmol NaOH with 1% Triton X-100 (Waswa et al. 2007), and the third was 1% SDS (Dudak and Boyaci 2007). All data acquired from the SPR assays were analysed using Qdat software ver. 1.0.0.24 (ICx Nomadics, Inc., Stillwater, OK).
gold Spreeta chip was employed. The chip surface functionalization was performed by adsorption of NeutrAvidin molecules onto the gold film with consequent irreversible binding with biotinylated rabbit polyclonal anti-E. coli antibodies. To reduce a nonspecific signal, the 2% BSA was used to saturate unoccupied gold surface. In this study, the sensor preparation and all SPR analyses were conducted at 25°C using a dual-channel, semiautomated SPR platform SensıQ (SensıQ Technologies, Inc., Oklahoma City, OK). The Spreeta-based planar gold sensor chip SS01 from SensıQ Technologies was first briefly cleaned with absolute ethanol and dried with airstream before assembly. The cleaned sensor was installed into a flow cell, and the system was primed with running buffer (PBS) and normalized by the software (ver. A.05). The sensor functionalization was performed in a single run with running buffer at a flow rate of 50 ll min1. Firstly, the gold surface was cleaned by injecting 100 ll of cleaning solution (100 mmol NaOH with 1% Triton X-100 in water) as depicted in segment 1 of Fig. 1. This was followed by rinsing with DI water (segment 2 in Fig. 1). The sensing channel was saturated by 150 ll of NeutrAvidin with the concentration of 100 lg ml1 discretely injected by a flow rate of 5 ll min1. To block unoccupied sites on a sensor surface, both channels were saturated with 150 ll of 2% BSA (bovine serum albumin) in PBS at the same flow rate. Finally, the sensing channel was charged with biotinilated rabbit polyclonal antiE. coli antibody (Abcam, Cambridge, UK) by discrete injection of 150 ll (100 lg ml1) at 5 ll min1.
Experimental setup Bacterial aerosols were generated by a 3 jet Collison nebulizer (BGI, Inc., Waltham, MA) from cells suspension and supplied into a rotating aerosol chamber. To achieve appropriate (detectable) bioaerosol concentration in the chamber, the initial bacterial suspension with the concentration of 95 9 106 CFU ml1 was used for nebulization. The nebulizer was operated at a flow rate of 6 l min1 of dry and filtered compressed air. The air temperature and humidity were continuously monitored and controlled throughout the entire procedure (T = 23–25°C and RH = 45–55%). To trace any unfavourable variations in aerosol supply, a real-time optical aerosol spectrometer (Model 4705, Aeronanotech, Moscow, Russia) was used. The rotating aerosol chamber used in this study was described in detail in our previous publications (Pyankov
SPR assay All SPR experiments were carried out in an assay format as per following protocol. After 120 s of the sensor equilibration/normalization step, 100 ll of the analyte was 4000
R
3500
Response, RU
3000 1640
2500 R
2000 1500
BSA
2090 RU
1000 500 Figure 1 The planar gold surface chip functionalizing with rabbit polyclonal antiEscherichia coli antibody. 1—washing solution; 2—DI water; R—running buffer.
RU
R
R 1
0 0
Journal of Applied Microbiology © 2014 The Society for Applied Microbiology
R
Antibody
NeutrAvidin
2 500
1000
1500
2000
2500
3000
3500
4000
4500
Time, s
3
Real-time bacterial aerosol detection
E.V. Usachev et al.
et al. 2012; Usachev et al. 2013a,b). Three ‘bubbler’ type personal bioaerosol samplers (Agranovski et al. 2005a) were charged with 50 ml of collecting liquid each (PBS, pH = 74) and stationary installed inside the chamber in a specially designed holder. For each experimental run, the nebulizer was charged with bacterial suspension and operated to produce relatively stable aerosol concentration in the chamber as was continuously monitored by the optical aerosol spectrometer (approx. 3 min were required to equilibrate concentration in the chamber). Then, the personal bioaerosol samplers were switched on and operated at the airflow rate of 4 l min1. Aliquots of collecting liquid from the samplers (100 ll) were acquired from all three devices after 1, 10 and 25 min time periods and analysed by the SPR procedure as described above. The final results were corrected to account for sampling related dilution of the air in the chamber (12 l min1) by HEPA (high-efficiency particulate air filter HEPA-CUP 75; GE Healthcare, RYDALMERE, NSW, Australia)- filtered ambient air. For the material balance of the bacteria being sampled at particular time periods, the fluorescent dye (fluorescein sodium salt, C20H10Na2O2, Fluka AG, Buchs, Switzerland) with the concentration of 05 mg ml1 in the nebulizer (corresponding to 1326 9 108 FU ml1) was used. The procedure is described in detail by Usachev et al. (2014) for the viral aerosol study. Briefly, the ratio of the bacterial concentration (expressed in CFU ml1) by the sample fluorescence intensity (expressed in arbitrary fluorescence units FU ml1) of the initial suspension in the nebulizer has been established as 716 9 102 CFU FU1. That enables estimating the total amount of airborne bacteria collected for each particular experimental run to be identified by measuring the fluorescence intensity of an aliquot of the collecting liquid acquired from corresponding samplers at required time periods.
Results In the present study the SPR technology was evaluated for direct bacterial aerosol detection. Figure 1 illustrates the sensor functionalization process. The fresh sensor was installed into SensiQ instrument, then primed with running buffer (PBS) and normalized according to the manufacturer’s recommendations. After rinsing of the sensor’s channels with regenerating solution (100 mmol NaOH and 1% Triton X-100) and deionized water, the NeutrAvidin was discretely injected into the sensing channel. The surface saturation was observed at 2090 RU level (arbitrary unit of the machine’s signal response). The unoccupied surface blockage was achieved by passing BSA solution through both channels (2% BSA in PBS). Finally, the sensing channel was charged with antibody at a level of 1640 RU. Testing of different regeneration solutions revealed the optimal (stable results with high reproducibility over 50 runs) solution for particular settings to be 100 mmol NaOH and 1% Triton X-100. The sensing range and detection limit of the technique were evaluated. Overnight cell culture was washed thrice with PBS. Serial tenfold dilutions of bacterial cells in PBS were prepared and tested in the assay format for 100 ll analyte injections with the flow rate of 15 ll min1 following regeneration by injection of 50 ll of previously described regeneration solution over 1-min time period. Each bacterial dilution was tested in triplicate. As is seen from Fig. 2, the response was increased with an increase of analyte concentration. All plotted curves represent the difference between the reference and sensing channels responses. Besides, in the assay format, all curves were referenced against responses obtained for pure PBS, which was taken as zero. Based on the response data for 10-second dissociation step, the analyte detection was in the range of four orders of magnitude from 15 9 104 to 15 9 108 CFU ml1 (Fig. 3). The error bars represent
140 1·5E+8
Response, RU
120 100 80 60 1·5E+7 40 20 0 0
100
200
300 Time, s
4
400
1·5E+6 1·5E+5 1·5E+4 500
Figure 2 The raw data obtained from detection of Escherichia coli decimal dilutions. The bacterial concentrations are expressed in CFU ml1. Running buffer was used as zeroed negative control. Each dilution was tested in triplicate, but only 1 curve per dilution is plotted.
Journal of Applied Microbiology © 2014 The Society for Applied Microbiology
E.V. Usachev et al.
Real-time bacterial aerosol detection
80
140 120
y = –1E–14x2 + 3E–06x + 4·8637 R2 = 0·9974
70 60 Response, RU
100 Response, RU
y = –3E–15x2 + 1E–06x + 3·6177 R2 = 0·9919
80 60 40
50 40 30 20
20 10 1E+4
1E+7 1E+5 1E+6 1E+8 Bacterial concentration, CFU ml–1
1E+9
Figure 3 Intact Escherichia coli cells detection range. Measurement results of the bacterial binding (mean values of 10 s dissociation step) for three repeats vs bacterial decimal dilutions. Polynomial trend line with equation is plotted.
standard deviations of at least 3 measurements undertaken for each dilution. A single-factor ANOVA test revealed statistically significant difference in response to the bacterial concentrations tested (q < 005). A high correlation (R2 value for a quadratic fit) between the average normalized response values and bacterial concentration was observed with the minimal detectable concentration of 15 9 104 CFU ml1 (confirmed by 5 replication test). Also, due to fast chip surface saturation occurring at high microbial concentrations in the liquid, the correlation of the results at >108 CFU ml1 becomes poor. The effect of the physical freeze/thawing stress onto sensitivity of direct bacteria SPR detection was investigated. Here, the bacterial suspension was washed with PBS and then several times freezed (at 20°C) and thawed (at +37°C) following production of serial tenfold dilutions on PBS for SPR assay. As is seen from Fig. 4, the response values were highly correlated with bacterial concentration growth (R2 = 099). The trend is almost perfectly quadratic fitted. Each dilution was tested in triplicate and standard deviations were estimated. The statistical significance of the response difference between each dilution points was confirmed with a single-factor ANOVA test (q < 005). Because of this physical treatment, the sensitivity of SPR detection was improved by one order of magnitude with the LOD corresponding to 15 9 103 CFU ml1. Figure 5 shows a comparison of the binding of nontarget bacteria (Bacillus subtilis, Pseudomonas sp.), viruses Journal of Applied Microbiology © 2014 The Society for Applied Microbiology
0 1E+2
1E+3
1E+4
1E+5
1E+6
1E+7
1E+8
Bacterial concentration, CFU ml–1 Figure 4 Physically stressed Escherichia coli cells detection range. Measurement results of the bacterial binding (mean values of 10-s dissociation step) for three repeats vs bacterial decimal dilutions. Polynomial trend line with equation is plotted.
Response, RU
0 1E+3
90 80 70 60 50 40 30 20 10 0 0
E· coli
non E· coli 100
200 300 Time, s
400
500
Figure 5 Selectivity test of the SPR sensor. Non-Escherichia coli — MS2 and T4 bacteriophages, Pseudomonas sp. and Bacillus subtilis. All microbes at concentrations above 107 CFU or PFU per ml, respectively.
(T4 and MS2 bacteriophages) and E. coli K12 strain to the antibodies against E. coli. The association and dissociation steps presented clearly show the specific immunoreaction of antibody/antigen pair used in this study. All microbes were used at concentration higher than 107 CFU or PFU per ml. All data were referenced against the blank injection of running buffer, as it was used as analyte diluent. Besides, the specificity of the antibodies used was also confirmed by ELISA assay. Finally, we challenged the SPR-based approach for direct bacterial aerosol detection. The bacteria (washed with PBS) were aerosolized and sampled over 25-min 5
Real-time bacterial aerosol detection
25
10
6
1
4 2 0 0
50 100 150 200 250 300 350 400 450 Time, s
Figure 6 The raw data obtained from SPR detection of Escherichia coli aerosol samples collected for 1, 10 and 25 min. Sampling liquid (PBS) was used as zeroed negative control.
period. The results obtained for bacterial aerosols collected by the personal sampler and analysed by the SPR technique are shown in Fig. 6. Here, response curves represent the E. coli cells binding with immobilized antiE. coli antibody for 1, 10 and 25 min of the sampler operation for collection of airborne micro-organisms. As is seen in Fig. 7, the SPR response was found to be in excellent correlation with the sampling time. The R2 value for a logarithmic fit was about 0999. In particular, highly consistent results were obtained for two longer sampling time periods, where standard deviation did not exceed 10%. The ratio of the bacteria to the fluorescence in the nebulizer suspension was found to be 716 9 102 CFU FU1. The fluorescence of liquid corresponding to 1-min sampling time was found to be 2445 9 104 FU ml1. Based on this finding, the minimal detectable concentration of bacteria collected for 1 min of air sampling was calculated to be 175 9 103 CFU ml1, which corresponds to the LOD of airborne bacteria at around 219 9 104 CFU per litre of air. Discussions and conclusions An approach similar to one described by Waswa et al. (2007) was used in this project for the SPR sensor functionalization. In addition, a referencing channel saturated with bovine serum albumin to avoid nonspecific sorption on the gold surface (Fig. 1) was used. The results obtained for charging efficiency were comparable with the previously reported ones for high capacity COOH5 sensor type (Usachev et al. 2014). Three chip regeneration solutions were initially challenged and the best reproducibility of the results was obtained for the solution described by Waswa et al. (2007) with slightly reduced NaOH concentration. In contrast, use of other two solutions led to significant 6
10
10
8
Response,RU
Response, RU
E.V. Usachev et al.
5 y = 1·3731In(x) + 5·7821 R 2 = 0·9993 0
0
5
10
15
20
25
Sampling time, min Figure 7 Bacterial aerosol SPR detection over 25 min sampling time. Escherichia coli concentration in nebulizer was 95 9 106 CFU ml1. Error bars represent standard deviation of at least 3 experimental runs.
losses of the SPR response even after very few runs. These findings could be explained by stability of the antibodies used in this project for particular regeneration settings. The sensitivity of the technique for detection of micrometre-sized microbial bioaerosol was firstly explored by using decimal dilutions of intact bacterial cells, which were minimally treated by thrice washing with PBS only. The LOD of the technique was found to be approx. 15 9 104 CFU ml1 (Fig. 3). This finding is in good agreement with the results presented by Leonard et al. (2004) (105 CFU ml1) and quite distant from the outcomes reported by Dudak and Boyaci (2007) (101 CFU ml1), Waswa et al. (2007) (102 CFU ml1) and Subramanian et al. (2006) (103 CFU ml1). This discrepancy could be explained by some difference in instrumentation involved and ligand/analyte pair properties. In addition, comparing present results with our previous data reported for viral bioaerosol detection, it is evidently seen that the sensitivity of the approach is much higher for larger sized analyte corresponding to bacterial strains. As was reported by Taylor et al. (2006), some decrease of lower limit of detection (LOD) of bacterial strains by SPR instruments could be achieved by optimization of the sample preparation step. It was evidently shown that bacteria treated with detergent lysing solution were detected at two orders of magnitude lower LOD. Considering that after aerosolization bacteria could change its morphological and immunogenic properties due to substantial physical stress occurring in the air environment (desiccation, high/low temperature, UV radiation, etc.), we investigated the effect of the physical freeze/thawing stress onto sensitivity of direct bacteria SPR detection. Our data revealed the LOD improvement by one order of magnitude (Fig. 4). The difference in the results for untreated and stressed bacteria samples could be Journal of Applied Microbiology © 2014 The Society for Applied Microbiology
E.V. Usachev et al.
explained by possible cell lysis as the result of treatment. In this case, the analyte is changing its morphological and possibly immunogenic properties. Moreover, the SPR sensitivity difference for intact and treated cells could be also explained by the polyclonal nature of the antibodies used. The treatment of complex target (whole bacterial cell) could alternate active immunoreacting pairs. In the previous studies, we have shown applicability of SPR approach in conjunction with the personal bioaerosol sampler for the detection of airborne virus (Usachev et al. 2013a,b, 2014). The combined sampling and detection technology was capable of providing selective and almost real-time detection of airborne virus in a manner of minutes. The high viral concentrations in the air could be detected in
