Biosensors & Bioelectronics 16 (2001) 399 408 www.elsevier.

com/locate/bios

Development of an evanescent-eld bre optic sensor for Escherichia coli O157:H7

A.P. Ferreira a,*, M.M. Werneck b, R.M. Ribeiro b
a

Department of Sanitation and En6ironmental Health, National School for Public Health, Oswaldo Cruz Foundation, Rio de Janeiro, Brazil b Biomedical Engineering Program, Federal Uni6ersity of Rio de Janeiro, PO Box 68564, 21945 -970, Rio de Janeiro, Brazil Received 6 March 2000; received in revised form 14 March 2001; accepted 12 April 2001

Abstract An intensity-modulated bre optic sensor was developed for Escherichia coli O157:H7. The interaction between the whole natural bacteria and the guided lightwave was carried out by means of evanescent-eld coupling. A correlation between optical response and the current number of bacteria was achieved. The device sensitivity had been calibrated for initial number of bacteria (N0) from 10800. The sensor sensitivity was 0.016 ( 9 0.001) dB/h/N0. The sensing mechanism starts together with the log phase leading the present sensor response to be ve to ten times faster than conventional bacteriological techniques. 2001 Elsevier Science B.V. All rights reserved.
Keywords: Fibre optic sensor; Evanescent eld; Escherichia coli O157:H7

1. Introduction Escherichia coli has been recognized as a commonplace microorganism found in the intestinal tract of human and other warm-blooded animals. E. coli typically colonizes the infant is gastrointestinal tract within hours of life. Usually it remains harmlessly conned to the intestinal lumen, only becoming pathogenic in the debilitated or immunosuppressed host, or when gastrointestinal barriers are violated. Therefore, E. coli has served as an indicator for faecal contamination in water and food. Enterohemorrhagic Escherichia coli O157:H7 is an emerging pathogen; which is the cause of foodborne illness, the major cause of serious outbreaks and sporadic cases of hemorrhagic colitis and hemolytic-uremic syndrome (Pickering et al., 1994; Siegler, 1995; Nataro and Kaper, 1998). The major impact of bacterial induced diarrhoea disease is on children under the age of 10 and on elderly people living in less-developed countries, in which bacterial diarrhoea diseases remain a signicant public-health problem. It represents also the
* Corresponding author. Tel.: + 55-21-598-2746; fax: +55-21-2707352. E-mail address: aldo@ensp.ocruz.br (A.P. Ferreira).

greatest challenge for both physician and scientic researches. E. coli O157:H7 is the cause of acute and persistent diarrhoea disease and ongoing morbidity that has malnutrition as one of its major contributors (Huilan et al., 1991). The level of biohazard is high due to the extremely low dose (10 microorganisms) required for infection (Nataro and Kaper, 1998). The greatest drawback for pathogen-detection techniques is the lack of reliable and sensitive means for measuring their presence in real time. Evanescent-wavecoupling sensor technology is a suitable technique for microorganism measurement in its natural form that may be applied to the biosensor development taking advantage of the current understanding of the whole cell architecture of the microbial (Watts et al., 1994; Hutchinson, 1995; Bousse, 1996; Muller et al., 1997; Nath et al., 1997). The biological and medical areas such as living cells and bacteria detection remain a challenging technique. The advantages of bre optic sensing are well known and have been reported in the literature (Smoczynski et al., 1993; Ligler et al., 1993; Karube and Nakanishi, 1994). In the last few years bre optic sensors have been increasingly employed in biological sensing because of their electrical isolation, electromagnetic-interference immunity, compactness, lightweight, sensitivity, in-line

0956-5663/01/$ - see front matter 2001 Elsevier Science B.V. All rights reserved. PII: S 0 9 5 6 - 5 6 6 3 ( 0 1 ) 0 0 1 4 9 - X

400

A.P. Ferreira et al. / Biosensors & Bioelectronics 16 (2001) 399408

implementation by means of evanescent-eld coupling, potential low cost and biological compatibility. The optical sensor has revealed to present a faster response than conventional techniques of clinical laboratory analysis. One of the major advantages of using the coupling with the guided lightwave evanescent eld is the development of in-line bre optic devices. The bre is not interrupted and the light is not removed from the bre. The signal processing takes place inside the optical bre thus characterizing an intrinsic sensor. A bre optic biological sensor has been designed in our laboratory and was described elsewhere (Ferreira et al., 1999a,b). In this work, using the same measurement principle, the sensor has been employed to quantify E. coli O157:H7. This paper describes for the rst time, to the best of our knowledge, experimental results on sensitivity calibration of an intensity-modulated bre optic evanescent sensor for the E. coli O157:H7 as a whole bacterium. The complete bacteria growth phases were optically sensed and time resolved with reproducibility. Electron and optical microscopy were also employed for a detailed understanding of the interaction dynamics between the bacteria and the optical bre probe.

population N(t) at the log phase can be derived from Eq. (2): N(t)=No 2t/g (3)

As more and more bacteria are competing for food and nutrients the booming growth stops and the number of bacteria stabilizes in the stationary phase. At the same time dying and dividing the organisms are at equilibrium. Death is due to reduced nutrients, pH changes, toxic waste and reduced oxygen. In some cases cells do not die but they are not multiplying. Eventually, when the death rate exceeds the growth rate, the cells enter the death phase. The population is dying in a geometric fashion so there is more death than new cells, and the population may be entirely destroyed.

2.2. The sensing mechanism of the sensor

The sensing mechanism of the sensor described here relies upon the attenuation of a guided lightwave by means of evanescent-eld coupling that takes place at the sensitive bre. As the bacteria grow in the neighbourhood of the sensitive bre, the guided lightwave is changed in intensity. The culture medium is adherent to the cladding of the almost unclad sensitive bre. Normally, light travelling inside the bre experiences total internal reection when light beams strike the interface between the bre core and the cladding. This occurs when the angle of incidence is greater than the critical angle. Thus, a propagating optical wave through the optical bre is bounded in its core, but allowed to interact with the outside medium by evanescent-eld coupling when this eld is exposed (Rahnavardy et al., 1997). The evanescent eld of the guided lightwave penetrates beyond the core surface and exponentially decreases in its magnitude an order of a wavelength away from the core/cladreecting interface. The evanescent-wave penetration depth (dp) is calculated by: dp = u , 2y(n sin2q n 2 )1/2 cl
2 

2. Principles

2.1. The bacteria growth e6olution

As reported from the literature (Pelczar et al., 1993), the bacterium growth evolution follows a well-known pattern that comprises four phases, as is briey described below. The LAG phase begins once the bacteria are isolated in a suitable biological culture medium. Growth is very slow at rst (little or no division), while the organism is adjusting its metabolism to the food and nutrients in its new environment. The log (or logarithmic) phase starts once the metabolic machinery is running and the bacteria start multiplying exponentially, doubling in number (binary ssion) at every few minutes. The time required for the bacteria population to double after the lag phase is called generation time g: g= t . n (1)

(4)

Where n is the number of generations that occur in the period of time t (in min). A more practical formula may be derived from the denition stated above and gives: g= t . 3.3(log10N log10N0) (2)

Where, N is the current number of bacteria at time t and N0 is the initial number of bacteria. The current

Where dp is the distance by which the magnitude of the electric eld decays of l/e to its intensity at the core/clad interface, q is the internal incident-ray angle with the normal to the core/cladding interface, u is the vacuum wavelength of the light beam and nco and ncl are the core and cladding index of refraction, respectively. The evanescent-eld coupling interaction mechanism will cause optical attenuation of the guide lightwave when any material (gas, liquid or solid) gets into contact with the surface of the etched bre clad. The loss of power is proportional to the intrinsic bulk absorption and scattering (Rayleigh and Mie). Thus, as a light ray progresses down the multimode bre by bouncing off

A.P. Ferreira et al. / Biosensors & Bioelectronics 16 (2001) 399408

401

the core-cladding interface, it loses power at each bounce. While the loss per bounce is small, the accumulated loss after a large number of bounces can be signicant even fbr a modest length of the probe bre. When a light ray travelling in a medium of a particular refractive index impinges on the boundary of another slightly attenuation medium of lower refractive index, it suffers the following approximate fractional loss of power on refraction: 4(q/q 2) c
q 2 q 2 c

3. Experimental The sensor described here may be envisaged as comprising by three parts: The optical circuitry, the evanescent probe bre (sensitive bre) and the biological culture medium.

3.1. The optical circuitry set-up

Fig. 1 shows a schematic drawing of the biosensor set-up. The optical source is a 3-mW CW GaAlAs laser (AsGa Microeletronica, Brasil) with graded-index mul timode bre pigtail, emitting at 840 nm. The output bre was spliced with a bi-directional optical bre coupler (CPqD Telebras, Brasil). Half of the emitted optical power propagates over the probe bre (sensing element). The light modulated by the growing bacteria exits the bre and is detected by the photodiode PD2 providing the electrical signal Samplein. The other half of the optical power is detected by the photodiode PD1 from which an electrical reference signal Refin is generated. Both electrical signals are amplied and measured by a twochannel calibrated optical-power meter (Graseby Optronics, USA). An AID board controlled by the LabVIEW software (National Instruments Corporation) digitises both output signals from the opticalpower meter (Sampleout. and Refout ) by means of its IEEE-488 interface. The AID board collects 800 points per min of both Sampleout and Refout recording and storing the respective averages for a period of 24 h per measurement cycle.

T=

 

uh . 4yncl

(5)

Where T is the fractional loss of power, qc is the critical angle of two media, q is incidence angle at the boundary between the media (measured with respect to the plane of the interface, rather than its normal), u is the optical wavelength, h is the bulk-attenuation coefcient of the culture medium at the region of contact, and ncl is the (real part of the) refractive index of the clad at the boundary (Snyder and Love, 1983). Thus, the fractional power reected back into the original medium is (1 T) and the fraction of the original power remaining after N reections is (1 T)N. In this formula q B qc and should not be too close to it for the approximation to be valid. This means that it applies to rays that would have been well guided in the absence of loss. Furthermore the tighter such guidance (i.e. the smaller q/qc), the lower the loss on reection. The optical attenuation can be generated by two interrelated physical mechanisms: variation of the clad refractive index ncl as well as of its absorption coefcient h. When the bacteria are allowed to grow around the optical bre, both effects can occur: 1. In the lag phase enzymes are released because of the bacteria metabolism, causing the change of the index of refraction. Although this issue is not the aim of this paper, the interdependence between absorption coefcient and index of refraction has been experimentally shown and is theoretically described by the Kramers-Kronig Eqs. (Nussenzveig, 1972). 2. Due to the existence of an increasing number of bacteria in contact with the bre during the log phase, the medium becomes more opaque as time goes by. Consequently the intrinsic absorption also changes. Both effects may change the mechanism of interaction between the evanescent eld and the outside material medium, coupling out different amounts of the bre-guided lightwave. Therefore, the measurement of the light power at the exit of the bre shows indirectly the bacteria growth. As explained in more detail in Section 4, the decreased optical power is mainly related to the number of bacteria present in the volume occupied by the evanescent eld around the bre.

3.2. The e6anescent optical probe bre

In order to expose the evanescent lightwave eld, 20 cm of a graded-index multimode optical bre (62.5/125 mm) was clad-stripped by chemical etching. The etching was performed by means of hydrouoric acid solution (38%) during 11 min in order to leave between 0.5 to 1 mm of clad over the core of the bre. After this time, the chemical reaction was stopped by immersion in demonised water and then in phosphate-buffered saline (PBS) with pH 7.4 for 15 min, in order to remove all remains of water from the probe. The exact etching

Fig. 1. Optical circuitry of the sensor.

402

A.P. Ferreira et al. / Biosensors & Bioelectronics 16 (2001) 399408

time was determined by monitoring diameters in a previous measurement using a calibrated optical microscope. This method has been detailed described by Muller and co-workers (Muller et al., 1998). After the chemical etching the bre was wound into a single loop, with a total sensitive surface area of approximately 40.0 mm2, in which the evanescent eld can be accessed. The probe bre was put over the culture medium in which the bacteria grew, selectively.

3.3. Selecti6e culture medium for E. coli O157:H7

All of the described experiments in this paper were performed using Escherichia coli O157:H7 CDC EDL933 (ATCC 43894), which was supplied by the Centre for Disease Control and Prevention (CDC/USA). In order to promote the E. coli O157:H7 growth, the selective culture medium employed was MacConkey Sorbitol Agar (SMAC) Difco 0079 17 7 (Difco Laboratories, Brasil) at an incubation temperature of 35 C. After optical measurements with the sensor, the bacteria underwent microbiological identity tests in order to ensure that only that specic microorganism had grown on each Petri plate. The identity tests were made in accordance with Farmer and Davis (1985). Although biochemical characteristics associated with the great majority of E. coli O157:H7 serotypes are well known, there are not much data available on their identication. It must be stressed that about 75 to 94% of E. coli strains quickly ferment D-sorbitol, while the O157:H7 strain does not (Thompson et al., 1990). It is also referred to this pathogen its inability to produce b-glucuronidase which hydrolyses-4-methylumbelliferyl-D-glucuronide. This does not happen with other serbtypes of E. coli. Another parameter is that more than 90% of E. coli O157:H7 strains produce one or two unique biochemical prole numbers on a MicroScan conventional gram-negative identication panel (Baxter Diagnostics, Inc., California) and other D-sorbitol negatives were not detected by this technique (Abbott et al., 1994). Finally, E. coli is a facultative anaerobe, which means that it can survive in the presence or absence of oxygen, it is a gram-negative bacterium and is not fastidious in its nutritional requirements.

for 24 h. At the end of this period, the purity of the material was checked and the Petri plates were stored at 4 C for further dilution. For obtaining a dilution with N0 = 10, 20, 30, 40, 50, 60, 70, 80 and 90 microorganisms, were used 100 ml of PBS, pH 7.4, whereas for N0 = 100, 200, 400 and 800 the volume used was 500 ml of PBS, pH 7.4. For each sample the cells were counted using the Coulter Counter (Beckman, USA), which allows an accuracy of 91%. Finally, the probe bre was rested upon the Petri plates with the culture as described above, the dilution with a known number of microorganisms was poured onto the sensitive area of the probe and the hardware and software were started. In all cases the SMAC was supplemented with glycerol at 0.2%, which allows the maintenance of the residual humidity and provides a better interaction between the microorganisms and the sensitive area. This was done in order to avoid the culture going dry during the tests and thus inserting another variable into the process.

3.5. Taking pictures of the process

For a direct observation of the microorganisms interaction with the probe bre, photographs were taken employing the scanning electron microscope and the optical microscope.

3.5.1. Scanning electron microscopy Micrographs were taken at 316 min (N0 = 400) and 410 min (N0 = 20), corresponding to the elapsed time after which the sensor reaches half of the optical-signal drop. Cells were xed with 2% (v/v) glutaraldehyde in PBS, pH 7.4, for 3 h at 4 C. After washing 3 times with PBS, pH 7.4, 4% (v/v) of osmium tetroxide was added to each sample during 1 h at 4 C. After dehydration in a graded series of ethanol dilutions, cells were subjected to a critical point drying with CO2 (Samdri-780A, Tousimis Research Corporation). Samples were covered with ag 0 Id thin lm by an ion sputtering (JFC-1 100 Ion Sputter, JEOL) and examined under a digital scanning microscope (Carl Zeiss DSM94OA). Photographs were taken with Agfa APX 100 black and white lm. 3.5.2. Optical microscopy The optical microscopy pictures were taken after 398 min (N0 = 100), corresponding to the stationary optical signal. Cells were xed with 2% (v/v) glutaraldehyde in PBS, pH 7.4, for 3 h at 4 C and then washed 3 times with PBS, pH 7.4. The images in differential interference contrast were obtained with the Zeiss Axioplan 2 microscope (Carl Zeiss) and photographed with a Polaroid MC2OO (Carl Zeiss) using Kodak Tmax 100 lm.

3.4. Bacteriological preparations for the sensor calibration

In order to calibrate the sensor for its sensitivity, several tests were performed using different N0. E. coli O157:H7, available lyophilised in ampoules, was restored with 1.0 ml of PBS, pH 7.4, inoculated in several Petri plates with SMAC and incubated at 35 C

A.P. Ferreira et al. / Biosensors & Bioelectronics 16 (2001) 399408

403

Fig. 3. Optical phase-contrast micrograph of the E. coli 0 1 57:H7 after Iout(t) reached the stationary state. Two levels of magnication are shown on 3(a b).

Fig. 2. Scanning electron micrograph of E. coli O157:H7 (log phase) that are in physical contact with the probe bre. Two levels of magnication are shown.

4. Results and discussion

4.1. Presentation of the experimental results

Fig. 2 and Fig. 3 display micrographs of the probe bre showing the physical contact with the E. coli O157:H7 growing around it. Fig. 2 shows the micrograph taken by means of the scanning electron microscopy on two levels of magnication. It displays the bre-bacteria physical contact during the Iout(t) drop (log phase) as shown in Fig. 4 and Fig. 5. Fig. 2(a) shows an island-like cluster of E. coli O157:H7 on top of the probe bre. Notice that the optical bre, as evaluated by the scale shown at the left of the picture, is only about 60 mm in diameter, which is approximately the same as the one of the core. Fig. 2(b) shows with a greater spatial resolution an as-like amorphous structure of the E. coli O157:H7 cluster in which the bacteria are stacked as foam. Fig. 3 shows a micrograph taken from a phase-contrast optical microscopy with two levels of magnication. It displays the bre-bacteria physical contact, after Iout(t) has reached the stationary phase, which are in

physical contact with the probe bre. The optical micrograph shows an almost physical isolation of the E. coli O157:H7 cluster (on top of the probe bre) from the others colonies. Because of the short range of the evanescent eld, the former were optically probed while the later were not. In order to characterize the sensitivity of the sensor, several measurements were carried out, each one during a 24 h interval (1440 min), employing thirteen different values of N0. These values were varied from 10800 E. coli O157:H7 bacteria samples. Fig. 4 shows the plot of

Fig. 4. Optical response Iout(t) of the sensor parameterised by the initial number N0 =0, 10, 80 and 800 of E. coli O157:H7.

404

A.P. Ferreira et al. / Biosensors & Bioelectronics 16 (2001) 399408

Fig. 5. Line 1 and line 2: The optical response Iout(t) of the sensor for two similar tests with N0 =200, referred to the left vertical axis (the superimposed straight lines are averaging of the data used to calibrate the sensor). Line 3: The estimated number of bacteria, as predicted by Eq. (3), referred to the right axis.

the temporal optical response Iout(t) (in arbitrary units) of three of such measurements (N0 =10, 80 and 800) plus the blank test (N0 =0). Fig. 5 shows two separated tests for N0 =400 (lines 1 and 2) and the theoretical current number of bacteria N(t) representing the population of the colony along its life evolution, as previewed by Eq. (3) (line 3). The superimposed straight lines are eye-guide lines representing an averaging of the experimental plot used to calibrate the sensor (see Section 4.6). The theoretical plot also shown at Fig. 5 is the representation of Eq. (3). As explained on the text (see Sections 4.2 and 4.6) the superimposed straight lines represent a mathematical idealization that helps on the sensor calibration. The arrows guide the reader for the correct vertical axis corresponding the two plots. As seen from the two tests performed with N0 = 200 shown in Fig. 5 it is possible to infer that the results of the optical measurements were quite reproducible. The good reproducibility of the optical response of the sensor is guaranteed by the systematic procedure that was employed for the biological culture as well the sensing probe bre insertion.

null angular coefcient was tted with an average of 2709 4 min or approximately 4.5 h, meaning a repeatability of $ 1.5%. DtLAG may be attributed to the time range that E coli 0 1 57:H7 spent in its lag phase, despite their initial number N0. Since enzymes are usually released by E. coli O157:H7 along the lag phase, the average refractive index of the culture medium ncm is likely to vary with time. As described in Section 2, the variation of ncm may cause changes in Iout(t), as stated by Eq. (5). However, the measurements shown in Fig. 6 suggest that the sensor, while probing the lag phase, is insensitive to this effect. A possible explanation for this is that the refractive index changed very little and the variation of the optical response was buried into the noise. It is possible to observe in Fig. 2 that the optical signals present a slight slope during this phase. Therefore it possible to infer that the technique used in this work is

4.2. Analysis of the sensor response Iout(t) in the lag phase of the culture
The plots displayed in Fig. 4 show different phases of Iout(t) for any used N0. In the rst phase, it may be observed a DC level Iout(t) = ILAG with an almost similar time range. The ILAG level is assigned to the sensor response in which the E. coli O157:H7 remains in its lag phase during ZtLAG time delay. Fig. 6 shows a plot of DtLAG against N0 for all measurements. The linear relationship with an almost

Fig. 6. Time delay in the lag phase DtLAG (min) against the initial number N0 of E. coli O157:H7.

A.P. Ferreira et al. / Biosensors & Bioelectronics 16 (2001) 399408

405

more sensitive to the bacteria division (growth) than to the variation of the index of refraction of the culture medium itself.

4.3. Analyses of the sensor response Iout(t) in the log phase

Following the lag phase, E. coli O157:H7 starts the log phase in which the size growth and replication take place. The curves displayed in Fig. 4 show that just after approximately 270 min the sensor output signal Iout(t) begins to drop gradually. The linear decreasing of Iout(t) may be attributed to the sensor tracking of the log phase evolution when the optical attenuation, as sensed by the evanescent eld, monotonically increases. It should be carelully examined from the electron scanning micrograph shown in Fig. 2 that the colony does not cover the lull surface of the probe bre. If the attenuation coefcient h and the refractive index ncm of the culture medium change in opposite directions, as the bacteria grow, the critical angle qc increases. As a consequence, shown by Eq. (5), the effect of increasing the attenuation coefcient is partially offset by improved optical guidance, for rays of a given incidence angle, and the nal effect on the optical power output is unnoticeable. However if the attenuation coefcient h and the refractive index ncm of the culture medium change in the same direction, as bacteria grow, the critical angle qc decreases. Both effects are now coherently summing up and the fractional loss T is reinforced, for rays of a given incidence angle, and the nal effect is a decrease of Iout(t). By extrapolation from the lag phase, it is unlikely that the enzymes produced by the E. coli O157:H7, even in the log phase, may affect the sensor Iout(t) for all N0 employed. In this way, along the log phase, only the attenuation coefcient a suffered appreciable changes, thus affecting Iout(t). In other words, the sensor described here seems to be insensitive to the average refractive index changes, if it changes at all. Only intrinsic absorption as well Rayleigh and Mie scattering affect the optical signal Iout(t).

instant of time, a fraction of E. coli O157:H7 stops its growth and replication, thus reaching its own stationary phase. In this way, the smaller derivative of Iout(t) is assigned to the coexistence of E. coli O157:H7 clusters over the probe bre, at both the log and the stationary phase thus affecting the optical response of the sensor. It is also noticeable in Fig. 4, for all measurements performed with different N0 (until N0 = 800), that ISTAT is always smaller than the previous one. The sensor response Iout(t) has shown an optical-dynamic reserve in the log phase that makes it clear that E. coli O157:H7 really reached the stationary phase. Thus, the ISTAT level (lower than ILAG) may be assigned to the stationary phase of the E. coli O1 57:H7 evolution, when the attenuation, as sensed by the evanescent eld, remains unchanged. After the clad was stripped from the probe bre, a surface area of approximately 40.0 mm2 has been achieved, and the evanescent eld was exposed. On the other hand, it should be carelully examined from the optical micrograph shown by Fig. 3(a) and (b), that E. coli O157:H7 at the stationary phase were able to cover almost all the surface area of the probe bre.

4.5. Analyses of the sensor response Iout(t) in the death phase

For a 24 h monitoring, the optical sensor response Iout(t) did not show any remarkably change after the stationary phase was reached. Although the E. coli O157:H7 starts its death phase some time after the end of the log phase, the sensor described here does not seem to optically discriminate the stationary phase from the death phase.

4.6. Calibration of the sensor

In Fig. 5 three eye guidelines were drawn for each Iout(t) regime of the sensor and E. coli O157:H7 phase. Along with the lag phase, an averaged unitary normalized straight line was outlined. Just at the beginning of the log phase, Iout(t) suddenly dropped with a sharp derivative. It was smoothed by a non-null angular-coefcient straight line. A similar procedure was performed at the stationary phase extrapolated from t= . But the steady state straight line outlined had an ISTAT DC level lower than the ILAG DC level. Two crossing points among the three straight lines may be viewed in Fig. 5. The rst one x the boundary between the lag and the log phase and the second one, analogously, x the boundary between the log and the stationary phase. The ordinate difference between the two cross points provides the optical attenuation DIout (in dB) for each N0. Similarly, the abscissa difference between the two

4.4. Analyses of the sensor response Iout(t) in the stationary phase

All plots displayed in Fig. 4 show that after the end of the Iout(t) drop, the sensor response reaches another DC level (ISTAT). Following the log phase, the E. coli 0157:117 stops their size growth and replications, thus reaching the stationary phase in which Iout(t) = ISTAT. A careful examination of Fig. 4 and Fig. 5 shows that, in the so-assigned log phase Iout(t) turn to a lower derivative, a little before reaching the stationary phase, it is likely that between the purely log and stationary phase, a transition region takes place. From some

406

A.P. Ferreira et al. / Biosensors & Bioelectronics 16 (2001) 399408

Fig. 7. Sensor calibration curve. It is shown the time derivative iLOG of Iout(t) in the log phase against N0. A linear dependence was tted.

cross points provides the time width DtLOG (in h) of the log phase for each N0. The indirectly measured (from Fig. 5) time derivative iLOG of Iout(t) in the log phase varies with N0 and may be calculated from: iLOG(N0) DIout . DtLOG (6)

Fig. 7 shows a plot of the sensor sensitivity bLOG (dB/h) at the log phase when the initial number of bacteria is changed from N0 =10 N0 =400. Although the available data is not enough to draw statistical conclusions it is possible to t the data with a straight line with a regression coefcient of 0.985 and standard deviation of 0.351. This gives us a provisional calibration curve as shown in Fig. 7. By employing the slope of optical response at the log phase, i.e. the rate of increase in bacterial numbers, the angular coefcient was calculated to be DiLOG/DN0 = (0.016 i 0.001) dB/h/N0. It means that for each E. coli O157:H7 bacterium inoculated upon the Petri plate the speed of the sensor response at the log phase increases by 0.016 dB/h. For N0 =800 the angular coefcient is given as iLOG(800)$ 6 dB/h. Therefore, it is possible to extract from the output signal Ii(t), the N0 by measuring the angular coefcient iLOG of the log phase. The N0 is directly related to the degree of contamination of the sample, when applying the system in vivo.

4.7. General analyses

It has been reported that bre optic evanescent-eld coupling presents some drawbacks when used as a biological sensor (Ramsden, 1997). In other biological processes as well, some bio-components may interfere in the detection of the true target. However, as was already explained at Section 3.3, the biological procedures herein employed were not only intrinsically highly selective, but also absolute. This explains why no inter-

ference on the detection of E. coli 0 1 57:H7 was observed. Although not shown here, it was also carried out a measurement in which none E. coli Ol 57:H7 were present in the culture medium negative control or blank measurement. In this case, the output signal Iout (t) did not change even after 24 h of monitoring, and the noise pattern observed was similar to the one presented in Fig. 4 (with the presence of E. coli O157:H7). In other words, none microorganism (except E. coli O157:H7) could grow. It is difcult to infer the causes of the noise superimposed in the output signal seen in Fig. 4 particularly because it is an ultra-low-frequency noise. Nevertheless it is possible to point out possible mechanisms that could cause it: 1. Thermal and acoustic instabilities, affecting the optical circuitry in every point, perturb the lightwave before it reaches the multimode bre optic 3dB coupler. This may cause deviations of the coupler ratio from the expected value of 3dB. The deviations may be specically caused by the mode coupling and polarization-dependent loss (PDL) effects (Renner, 1998). 2. Electronic noise, due to instabilities of the detection and amplication circuitry. The sensor response relies on the interaction between the lightwave and the bacteria causing the optical attenuation at the probe bre. The optical interaction arises from the evanescent-eld coupling since there is a physical contact between the bre and the bacteria, as shown in Fig. 2 and Fig. 3. The physical contact between the SiO2 (probe bre) and the whole E. coli O157:H7 occurs because the bacteria are allowed to grow around the probe bre. This mechanism greatly simplies the construction of the sensor, when compared with 6ther processes, for instance the silanization technique employed by Weetall (1993) for a true biosensor. A bacteria colony may grow indenitely if an innite amount of nutrients would be available, as well as if other conditions, such as temperature, pH etc., would be optimised. However, using culture medium in a Petri plate, it is obvious that there is a strong limitation and the biological medium constrains the bacteria growth in an available time range. In this study, preliminary experiments have shown that E. coli O157:H7 is able to grow for up to about 72 h, using similar Petri plates and culture medium as those employed for the optical measurements shown in Fig. 4. Therefore, the question that arises is: why, since all conditions remains the same, E. coli O157:H7 reached the stationary phase in an average of 5 h (see Fig. 4), when this should happen only in about 72 h? This behaviour may be explained in the following way:

A.P. Ferreira et al. / Biosensors & Bioelectronics 16 (2001) 399408

407

It is known that E. coli O157:H7, as well as some other bacteria, do not grow inside the bulk of the culture medium, but only at the surface. However, a careful examination of Fig. 2(b) suggests that E. coli O157:H7 grows and stacks one over each other, up to 3 layers, generating voids like foam. The as-like amorphous structure of E. coli O157:H7 cluster means that the lightwave evanescent eld would not be attenuated (absorbed and scattered) in the voids or in places where the stacked bacteria are \ \1 mm far away from the surface of the probe bre. The cells of E. coli O157:H7 that happen to be all over the Petri plate, without physical contact with the probe bre, go on growing for 72 h because there are enough nutrients. However, these optically isolated bacteria do not affect the sensor response Iout(t), since they are out of reach of the sensing mechanism, the evanescent eld, that extends only about one or slightly over a, wavelength from the core-clad interface. On the other hand, the bacteria that touched the bre during the inoculation, together with those that happen to grow over and around the bre, are within the evanescent eld, and are sensed and monitored. However, since the probe bre was rested upon the surface of the biological medium prior to the optical measurements, the lightweight probe bre was not completely buried inside the gel-like culture medium, because of the supercial tension and the nutrients in excess slowly slide down from the top of the probe bre. Also due to the surface tension, only a thin lm of culture medium will remain touching the bre. Therefore, the bacteria that happen to be isolated in the probe bre will grow and reproduce until the exhaustion of this limited amount of nutrients. The micrograph shown in Fig. 2(a), taken while the sensor reached half of the optical-signal drop, show the bacteria clusters isolated along the centre of the bre width. This is the initial level of the culture, but due to evaporation the level drops leaving the clusters behind. The calibration of the sensor sensitivity with N0, as shown by Fig. 7, features a linear dependence of iLOG( = DIout/DtLOG) from N0 =10 until N0 =400. For N0 = 800 a deviation was observed from the linear dependence, which suggests a possible saturation of the sensor response.

2. The use of a longer probe bre (the actual measures 20 cm) will attenuate more light and therefore will present a higher sensitivity for the same bacteria concentration; 3. At present, it is possible to detect only one microorganism per Petri plate at a time. One approach to overcome this limitation would be the miniaturization of the probe bre such as that an array of them might be employed. Several different microorganisms may be selectively sensed, for instance, by multiplexing different wavelengths into a single optical bre bus (wavelength-division multiplexing WDM).

5. Conclusions An evanescent-eld and intensity-modulated bre optic sensor for detection and monitoring of E. coli O157:H7 has been described and calibrated. These bacteria have been successlully detected and quantied in its natural form, 5 10 times faster than conventional bacteriological techniques. With this system it is possible to conclude which bacteria have been inoculated and its concentration. The development of this highly sensitive and selective probe for real-time pathogen detection has improved biological sensing. With minor modications the method can be used to test for food contamination as well as for clinical essays and environmental monitoring. Further investigations are under way in our laboratories and will be object of luture publications.

Acknowledgements We would like to thank the Escola Nacional de Saude Publica of the Fundacao Oswaldo Cruz (ENSP/ FIOCRUZ) for the bacteriological support on this work; the Foundation for the Research Support of the State of Rio de Janeiro (FAPERJ) and the Brazilian National Research Centre (CNPq/PADCT) for nancial support. We are also in dept with Mr. Jose Roberto da Rocha Bernardo, MSc, and Dr. Ulisses Garcia Casado Lins, from the Federal University of Rio de Janeiro, for their assistance in electron scanning and optical microscopy, respectively.

4.8. Possible impro6ements of the sensor

In order to increase the sensitivity and the time derivative (speed at the log phase) of the sensor in its presently basic conguration, some simple improvements are suggested: 1. Optimisation of the wavelength for a better sensitivity. For instance, by using a longer wavelength we would have a larger sensitivity volume because the evanescent eld would also be larger;

References
Abbott, S.L., Hanson, D.F., Felland, T.D., Connell, S., Shum, A.H., Janda, J.M., 1994. Escherichia coli O157:H7 generates a unique biochemical prole on MicroScan conventional gram-negative identication panels. J. Clin. Micro. 32 (3), 823 824. Bousse, L., 1996. Whole cell biosensors. Sensors & Actuators B34, 270 275.

408

A.P. Ferreira et al. / Biosensors & Bioelectronics 16 (2001) 399408 Nussenzveig, H.M., 1972. Causality and Dispersion Relations. Academic Press, New York. Pelczar, M.J., Chan, E.C.S., Krieg, N.R., Edward, D.D., 1993. Microbiology: Concepts and Applications. McGraw-Hill Publishing Company, USA. Pickering, L.K., Obrig, T.G., Stapleton, F.B., 1994. Hemolytic-uremic syndrome and enterohemorragic Escherichia coli. Pediat. Infect. Dis. J. 13, 459 476. Rahnavardy, K., Arya, V, Wang, A., Weiss, J.M., 1997. Investigation and application of the frustated-total-internal-reection phenomenon in optical bres. Appl. Opt. 36 (10), 2183 2187. Ramsden, J.J., 1997. Optical Biosensors. J. Molec. Recog. 10, 109 120. Renner, H., 1998. Polarization characteristics of optical waveguides with separable symmetric refractive-index proles. J. Opt. Soc. Am. A 15, 1401 1410. Siegler, R.L., 1995. The hemolytic uremic syndrome. Pediat. Clin. North America 42, 1505 1529. Smoczynski, C., Costello, M., Bliss, M., Bharani, N.K., 1993. Increased detection of anaerobic bacteria with the use of bre optic light. Clin. Infect. Dis. 16 (4), 322 324. Snyder, A.W., Love, J.D., 1983. Optical Waveguide Theory. Chapman and Hall, London. Thompson, J.S., Hodge, D.S., Borczyk, A.A., 1990. Rapid biochemical test to identify verocytoxin-positive strains of Escherichia coli serotype O157. J. Clin. Micro. 28, 2165 2168. Watts, H.J., Lowe, C.R., Pollard-Knight, D.V., 1994. Optical biosensor for monitoring microbial cells. Analyt. Chem. 66, 2465 2470. Weetall, H.H., 1993. Preparation of immobilized proteins covalently coupled trough silane coupling agents to inorganic supports. App. Biochem. Biotech. 41, 157 188.

Farmer, J.J., Davis, B.R., 1985. H7 antiserum-sorbitol fermentation medium: a single screening method for detecting Escherichia coli O157:H7 associated with hemorrhagic colitis. J. Clin. Micro. 22, 620625. Ferreira, A.P., Werneck, M.M., Ribeiro, R.M., 1999a. Pathogen detection using evanescent wave breoptic biosensor. In: 18th Congress of the International Commission for Optics, Proceedings of The International Society of Optical Engineering (SPIE) 3749, pp. 395 396. Ferreira, A.P., Werneck, M.M, Ribeiro, R.M., 1999b. Aerobiological pathogen detection by evanescent wave bre optic sensor. Biotech. Tech. 13 (7), 447 452. Huilan, S., Zhen, L.G., Mathan, MM., 1991. Etiology of acute diarrhoea among children in developing countries: a multicentre study in ve cduntries. Bull. World Health Org. 69, 549 555. Hutchinson, A.M., 1995. Evanescent wave biosensors. Molec. Biotech. 3, 47 54. Karube, I., Nakanishi, K., 1994. Microbial biosensors process and environmental control, IEEE Eng. Med. Biol., 364 374. Ligler, F.S., Golden, J.P., Shriver-Lake, L.C., Ogert, R.A., Wijesuria, D., Anderson, G.P., 1993. Fibre-optic biosensor for the detection of hazardous materials. Immunomethods 3, 122 127. Muller, C., Hitzman, B., Schubert, F., Scheper, T., 1997. Optical chemo and biosensors for use in clinical applications. Sensors & Actuators B40, 71 77. Muller, M., Squier, J., Wilson, K.R., Brakenhoff, G.J., 1998. 3D microscopy of transparent objects using third-harmonic generation. J. Micros. 191 (3), 266 274. Nataro, J.P, Kaper, J.B., 1998. Diarrheagenic Escherichia coli. Clin. Microb. Rev. 11 (1), 142 201. Nath, N., Jain, S.R., Anand, S., 1997. Evanescent wave bre optic sensor for detection of L. dono6ani specic antibodies in sera of kala azar patients. Bios. Bioeletron. 12 (6), 491 498.