2. unclad region was tapered rapidly to a radius such that the collecting the aerosol. The assay routine was repeated every
number of modes carried by the clad and unclad regions 5 min.
were identical. The remaining 6 cm was tapered gradually Briefly, with the laser fluorimeter on constantly, the sample
to a final radius of 100-130 µm. is pumped from the air sampler over the fibers and into
The fiber optic probes were then coated with avidin waste for 140 s followed by a 20-s wash. Air is pumped for
utilizing a thiol-terminal silane in conjunction with a 15 s to prevent dilution of the Cy5-labeled antibody, which
heterobifunctional cross-linker (2, 10). An avidin-biotin is subsequently pumped over the fibers for 80 s. The
linking system was utilized to minimize the denaturing effects fluorescent signal is recorded every second. After incubation
of direct immobilization onto the glass fiber surface on the with the labeled antibody, the fibers are rinsed with wash
capture antibodies. For immobilization of avidin, fibers buffer for 30 s to reestablish baseline. The labeled antibody
subsequently were treated in a series with (1) boiling is returned to a reservoir for reuse in the subsequent assay.
deionized water, (2) a solution of 2% (v/v) 3-mercatopro- Air Sampler Testing and Development. Various air
pyltrimethoxy silane (Fluka) in toluene, and (3) a 2 nM sampler designs based on the use of ram air to introduce the
solution of the cross-linker, N-succinimidyl 4-maleimidobu- sample into aqueous media were tested for their ability to
tyrate (Fluka) in absolute ethanol. After extensive washing collect ovalbumin released into a wind tunnel. The oval-
in deionized water, fibers were immersed in a solution of bumin was aerosolized and introduced into the center of an
avidin (Pierce) at 50 µg/mL in phosphate-buffered saline air stream flowing at 40-50 mph. Amounts of ovalbumin
(PBS) at pH 7.4. Fibers were rinsed and stored in PBS collected in the various air samplers after 5 min were
containing 0.01% sodium azide at 4 °C. measured using a standard ELISA. A standard all glass
impinger (AGI-30) equipped with a pump drawing in 12 L
Polyclonal rabbit serum specific for Bsn was generously
of air/min was used as a reference and run simultaneously
supplied by Dr. William Lee (Defense Research Establish-
with all of the experimental air samplers. Due to variations
ment, Suffield, Canada) and purified using Protein G affinity
in the total amount of ovalbumin released, values were
chromatography. Biotinylation was performed using N-
expressed as ratios of ovalbumin collected by the ram air-
hydroxysuccinimide-biotin (CalBiochem). Immediately be-
driven samplers as compared to that collected by the all
fore use, the biotinylation reagent was dissolved in dimethyl
glass impinger.
sulfoxide at 5 mg/mL. Antibody was diluted 1:1 in 0.05 M
On the basis of these data, we determined that the NRL-
sodium borate buffer, pH 9.6, containing 0.04 M sodium
built cyclone functioned better than all but one other air
chloride. The biotin solution was added at 3:1 molar excess
sampler tested and that the height of the collector could be
to antibody and allowed to react at room temperature for 45
reduced approximately 3-fold without affecting the collection
min, before purification of antibody over a Bio-Gel P-10
efficiency. The final sampler consisted of a plastic funnel as
column (Bio-Rad) equilibrated with PBS at a flow rate of 0.5
the base with an outlet at the bottom to remove fluid, a
mL/min. Purified protein peaks were identified by monitor-
plexiglass collection tube rising from the funnel (4 cm in
ing of fractions at 280 nm, and concentrations were estimated
height and 5 cm in diameter), an air intake tube 2 cm in
by adsorbance at 280 nm. Fractions were stored at 4 °C until
diameter positioned at an angle into the side of the collection
needed.
tube so as to generate the cyclone, and an air exit tube 2 cm
Avidin-coated fibers were inserted into 100-µL capillary in diameter extending out the top of the collection tube.
tubes shortened to cover the exposed length of the probe. Water was injected into the air intake tube using a syringe
Plastic T-connectors (Value Plastics) were inserted into needle bent to spray liquid into the air stream just as it entered
silicone tubing endcaps, and fluid paths were established off the collection tube.
the right angle of each T-connector (4). Each connector end
was sealed with hot glue. Avidin-coated fibers were reacted Results
with the purified biotinylated-goat anti-Bsn antibody at a
Improving Immunoassay Sensitivity. The Analyte 2000 fiber
concentration of 10 µg/mL in PBS for 1 h. For long-term
optic biosensor (Research International, Woodinville, WA)
storage, the fibers were incubated with 15 mM sodium
was selected for this project because it weighs 1.4 kg, is
phosphate containing 10 mg/mL BSA, 100 mM trehalose,
capable of performing four immunoassays simultaneously,
and 0.01% sodium azide for 30 min and then dried under a
and has an evanescent wave sensing modality that is relatively
stream of nitrogen. Four fibers were attached in series by
immune to interference from sample components (5). We
connecting with silicone tubing the distal or proximal fiber
have developed and previously described reusable sensing
T-connector of one fiber with the adjacent fiber T-connector
probes enclosed in capillary tubes for simplified introduction
(Figure 3). The dried fibers were stored at room temperature
of fluids (6). In the past, 10-min, manual sandwich immu-
until use (up to 3 months).
noassays were performed on the probes by immobilizing
Assay Protocol. Goat IgG antibody specific for Bsn was anti-Bsn antibodies directly on the fiber through a silane
labeled with fluorescent Cy5 reagent following the manu- film and heterobifunctional cross-linker (2, 10). Detection
facturers recommendations to a molar ratio of between 3:1 limits of only 106 cfu/mL were obtained using this standard
and 4:1 dye:protein (5). Briefly, protein is mixed with Cy5 methodology and were not deemed of sufficient sensitivity
labeling reagent in 0.05 M sodium borate, pH 9.6, for 1 h and for the expected aerosol field tests.
separated from unreacted dye on a 10-cm P-10 (Bio-Rad) To address the limits in detection levels, focus was placed
column equilibrated with PBS at a flow rate of 0.5 mL/min. upon moderating the detrimental effects of chemical cross-
Labeled antibody was identified visually as the first eluted linking on the antibody molecule. Detection limits observed
dye fraction. Dye/protein ratios were determined by ab- when utilizing the same anti-Bsn antibody reagent in solution
sorbance at 650 and 280 nm, respectively. by enzyme-linked immunoadsorbent assays (ELISA) indi-
The assay was performed simultaneously on four fiber cated that much lower sensitivities were possible (data not
optic probes using the Analyte 2000 (Research International) shown). Suspecting that the antibody directly cross-linked
and the automated fluidics system. Approximately 1 mL of to the glass surface was sterically hindered by excess cross-
sample collected in 15 mM sodium phosphate buffer, pH links to the surface or by concomitant adsorption, the
7.4, was pumped from the air sampler over the fiber probes. antibody was immobilized through a biotin-avidin bridge
The 1 mL represented approximately 10% of the fluid in the (Figure 1). Employing an avidin-biotin cross-linking method
air sampler so that the sampler continued to concentrate all not only limited the number of sites on the antibody used
bacteria removed from the air over the entire period it was to attach it to the surface but also had the added benefit of
2462 9 ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 32, NO. 16, 1998
3. (Alitea AB, Stockholm, Sweden) and diaphragm minipumps
(Lee Co., Westbrook, NJ). Design and assembly consider-
ations focused upon development of a reliable and robust
system that could withstand the environmental extremes
expected at Dugway along with the mechanical stresses
inherent in UAV operations. Three of the three-way valves
are cascaded to allow for four inputs using a minimal amount
of tubing, and a fourth valve is used to either recycle the
fluorescently labeled antibody or to direct sample or wash
buffer to waste (Figures 3 and 4). The peristaltic pump is
dedicated to the immunoassay reagent fluidics, and the
solenoid diaphragm pump serves to refill the cyclone sampler
after repetitive sample removal. Except for the peristaltic
pump, the entire system uses less than 1 W of power, weighs
less than 100 g, and fits on a 3 × 3 in. printed circuit board.
FIGURE 1. Schematic of immunoassay on the fiber probe. Avidin (Originally, two solenoid pumps were used to conserve
is covalently immobilized on a tapered fiber probe (1), and weight, but the peristaltic pump proved necessary for reliable
biotinylated antibody is attached to the avidin. Antigen (3.5 min) operation, especially in handling relatively dirty cyclone
and fluorescent antibody (1.5 min) added in sequence form a samples from field collections over extended time periods.)
fluorescent complex that is excited by light in the evanescent region The Analyte 2000 was controlled by manufacturer-
at the surface of the optical fiber. Fluorescent light is collected by supplied PC software (version 4.27) and is capable of
the fiber and transmitted to the photodiode.
recording a signal from each of the four fiber inputs at a rate
up to 1 reading/s. To relay data to and from the ground,
wireless modem transceivers (Freewave Technologies, Boul-
der, CO) were connected to the dedicated laptop computer
and the Analyte 2000 through the RS-232 ports and trans-
mitted between the two at 9600 Bd. Recorded signal data
were downloaded into Excel for Windows for reduction and
analysis. The airborne system was powered by two NiCad
battery packs (SR Batteries, Bellport, NY). These batteries
could run the integrated system for 2 h.
Air Sampling. The air sampling requirement presented
a unique challenge for the project. Several types of air
samplers have been developed for collecting particles from
air into aqueous reservoirs (12). Each had serious limitations
with respect to weight, power, and overall efficiency (as
measured by total particles removed in a given aerosol cloud).
The AGI, the reference standard for collecting 1-10 µm
particles from air into water, is highly efficient, but samples
air at a flow rate of only 12 L/min. The cyclone samplers
previously described intake volumes of 100-1000 L/min but
FIGURE 2. Standard curve for Bsn (Bg). Detection of various have traditionally required heavy vacuums or fans. The
concentrations of Bsn was performed using a sandwich immu- electrostatic impaction systems are efficient but require
noassay at the surface of tapered fiber probes (7-9) and the Analyte power beyond current portable batteries. To avoid these
2000. Data represent the mean ( SD of three determinations. Solid limitations, NRL built a small plastic cyclone, generally based
bars represent data using the anti-Bsn immobilized on the silane- on the design of Griffiths (13), but engineered to take
coated fiber via a biotin-avidin bridge whereas open bars represent advantage of the ram air flow from the movement of the
data obtained using anti-Bsn immobilized directly to the silane- airplane. The cyclone operates by creating a rapidly circu-
coated fiber. lating aqueous film on the inner walls of the sampler as air
enters at a speed between 40 and 50 mph. The cyclonic
further passivating the glass surface against nonspecific action serves to transfer micron-sized particles into the liquid
interactions with the antibody. The same silane chemistry from the aerosol cloud. At ram air speeds common for the
as previously described for antibody immobilization (10) was UAV platform, sampling volumes were estimated at over 100
used to attach avidin to the fiber probe. Although biotin- L air/min. Total particle collection of the NRL-designed ram
ylation ratios up to 20:1 are feasible for many antibodies, the air cyclone was determined to be 10-15-fold that of a 12
optimum molar labeling ratio for the anti-Bsn antibody was L/min AGI-30 by testing in the wind tunnel at Aberdeen
found to be less than 4 biotin molecules per antibody. When Proving Ground (14).
using the anti-Bsn antibody bound to the probe through an The remotely piloted airplanes employed for the majority
avidin-biotin bridge (11), the detection limit was reduced of the tests are from a family of model airplane kits called
to 3000 cfu/mL, nearly a 1000-fold improvement over the Telemasters, originally designed and used in Germany to
standard direct cross-linking method (Figure 2). Fiber probes string telegraph lines across valleys. Two 12 ft Telemasters
coated with antibodies in this manner could be dried and were custom built for NRL by BAI Aerosystems, Easton, MD.
rehydrated after several months without significant loss of The 12 ft Telemaster is a high wing utility aircraft with a
activity (data not shown). wingspan of 3.6 m, an empty weight of 19 kg, and a payload
System Automation. The automated fluidics prototype volume of 0.032 m3. The fuselage was slightly widened to
and controlling software code were built upon embedded allow easier access to the payload. The onboard systems
processor cards designed in house and utilized commercially include an Airtronics radio control system, a BTA wing
available microfluidic components. Several iterations re- leveling autopilot, and a Zenoah G-62 gasoline engine for
sulted in a system utilizing three-way solenoid microvalves propulsion. This aircraft was chosen because it can carry a
(Lee Co., Westbrook, NJ) with a combination of peristaltic 4.5-kg payload and is easy for the pilot to fly, an important
VOL. 32, NO. 16, 1998 / ENVIRONMENTAL SCIENCE & TECHNOLOGY 9 2463
4. FIGURE 3. Schematic of biosensor system.
characteristic for testing at night. From the point of view of tional for approximately 20 min before a positive sample
mimicking a real-life scenario, it was considered important was generated (in comparison to 5 min for two tests and 10
to be able to operate at night as well as in the daytime due min for one test). It was later determined that only a half
to the steadier wind conditions at night and the potential for pound of Bsn had been released during this test. One flight
sunlight to degrade some biological agents. To operate at was conducted when there was no Bsn release (negative
night, the pilots wore night vision goggles, and the aircraft control) and no signal was observed by the UAV.
were equipped with night vision goggle-compatible lighting Temperatures during the flights ranged from 6 to 24 °C
for the pilots to distinguish aircraft attitude. (Table 1). While low temperature combined with high
At Dugway Proving Grounds, tests were conducted from humidity (7 °C, 96% humidity) caused icing of the wings on
a small airstrip located approximately 1.6 km from the aerosol one flight, the operation of the sampler and sensor dem-
release area. The ground monitoring station was run inside onstrated no impairment. In one test flight, interference
a mobile vehicle operating approximately 100 ft from the with the radio control frequency forced the airplane into an
airstrip. The Bsn was released in varying amounts using autopilot-controlled crash on the sagebrush-covered test
either a truck-based or airplane-based crop sprayer. The range. As an indication of the ruggedness of the integrated
UAV was flown in an oval racetrack pattern, 300-500 m in design, although the airplane was damaged beyond repair,
size and 10-30 m above the ground, during nine releases. the biosensor payload was found completely intact and still
Each flight averaged 12-15 min duration and was conducted operating in the payload bay of the fuselage. The entire
at night. The Bsn aerosol was not visible in the darkness. package was removed, certified as operational, and flown
UAV flights were conducted approximately 1.6-2.0 km successfully the next evening in a second UAV.
downwind from release area.
On the first night of aerosol release testing, no Bsn was Discussion
detected in samples collected during flight. However, the The demonstration of a fully autonomous, remotely operated
Bsn was detected using an Analyte 2000 on the ground to biosensor opens the way for numerous real-world applica-
analyze eluants from wipes of the wings, proving that the tions. The integration of an air sampler with the biosensor
UAV had indeed passed through the aerosolized Bsn. With demonstrates both that the requirement of biosensor for
the intention of increasing the collection efficiency of the aqueous sample does not prevent its use for detection of
cyclone, the air intake was extended from near the fuselage airborne analytes and that the biosensor can, in theory, be
to a position forward of the wing strut. melded to a variety of automated samplers for remote
During the next four flights, positive identification of Bsn operation. Replacement of manually operated fluid ma-
in samples from the ram air cyclone by the fiber optic nipulations (such as pipetting by hand) with automated
biosensor was relayed to the ground (Figure 5). The assays, fluidics systems resulted in reduced variation within and
run in quadruplicate, showed good agreement, although there between assays (data not shown). Also, the likelihood for
was some variation in signal magnitude. The last flight test operator-induced error is eliminated. As such, the opera-
indicated a positive signal although the cyclone was opera- tional parameters defined by this integrated system are
2464 9 ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 32, NO. 16, 1998
5. FIGURE 4. Photograph of automated fluidics unit and Analyte 2000.
TABLE 1. Weather Conditions during Flights
flight no. sample identification temp (°C) humidity (%)
1-4 negative 15-20 41-43
5 positive 22 36
6 positive 18 52
7 positive 7 92
8 negative 6 80
9 positive 24 57
Acknowledgments
This work was supported by the Defense Advanced Projects
Agency and the Counterproliferation Office of the Office of
the Secretary of Defense. It builds on research supported by
the Office of Naval Research. The authors wish to thank Dr.
FIGURE 5. Data from test flights. The signal increase obtained from
William Lee (Defense Research Establishment, Suffield,
successful flights is plotted with the standard curve shown in Figure
Canada), who provided the antibody against Bsn; Dan Weber
1 for comparison. For each test, the biosensor had four probes
attached, each coated with avidin and a biotinylated rabbit anti- and Peter J. Stopa of U.S. Army Aberdeen Proving Ground,
Bsn antibody. The biosensor ran repetitive 5-min assays from take who conducted the tests of various air sampler designs in
off until landing. Sample was withdrawn from the cyclone for 3 min the Edgewood Wind Tunnel; and LTC R. Ranhofer and the
and slowly pumped over the probes. The probes were then washed staff of Dugway Proving Ground, who were responsible for
with buffer and then interrogated for the presence of Bsn by cycling preparing the airstrip and conducting the aerosol releases.
a solution of Cy5-labeled rabbit anti-Bsn (20 µg/mL) over the surface Any opinions expressed here are those of the authors and
for 2 min. The average fluorescent signal increase for the final not the U.S. Navy or Department of Defense.
cycle from the four probes ( SD after washing is plotted.
pertinent to air or water monitoring for any number of Literature Cited
environmental or industrial applications. For example, to (1) Hirshfeld, T. E.; Block, M. J. U.S. Patent No. 4,447,546, 1984.
monitor environmental remediation, intermittent ground- (2) Eigler, F. S. (Ligler misspelled by patent office); Calvert, J. M.;
water samples could be injected into an automated sensor Georger, J.; Shriver-Lake, L. C.; Bhatia, S. K.;. Bredehorst, R. U.S.
with the resulting signal relayed back to a central monitoring Patent No. 5,077,210, 1991.
(3) Anderson, G. P.; Golden, J. P. U.S. Patent No. 5,430,813, 1995.
facility. Combined with air sampling capabilities, the (4) Anderson, G. P.; Golden, J. P.; Cao, L. K.; Wijesuria, D.; Shriver-
integrated biosensor system is uniquely positioned for any Lake, L. C.; Ligler, F. S. IEEE Eng. Med. Biol. 1994, 13, 358.
number of bioanalytical challenges where full-time operator (5) Golden, J. P.; Saaski, E. W.; Shriver-Lake, L. C.; Anderson, G. P.;
presence is neither practical nor possible. Ligler, F. S. Opt. Eng. 1997, 36, 1008.
VOL. 32, NO. 16, 1998 / ENVIRONMENTAL SCIENCE & TECHNOLOGY 9 2465
6. (6) Ligler, F. S.; Golden, J. P.; Shriver-Lake, L. C.; Ogert, R. A.; (12) Lighthart, B., Mohr, A. J., Eds. Atmospheric Microbial Aerosols:
Wijesuria, D.; Anderson, G. P. ImmunoMethods 1993, 3, 122. Theory and Applications; Chapman and Hall: New York, 1994;
(7) Cao, L. K.; Anderson, G. P.; Ligler, F. S.; Ezzell, J. J. Clin. Microbiol. 397 pp.
1995, 33, 336.
(8) Tempelman, L. A.; King, K. D.; Anderson, G. P.; Ligler, F. S. Anal. (13) Upton, S. L.; Mark, D.; Douglass, E. J.; Hall, D. J.; Griffiths, W.
Biochem. 1996, 223, 50. D. J. Aerosol Sci. 1994, 25, 1493.
(9) Shriver-Lake, L. C.; Donner, B. L.; Ligler, F. S. Environ. Sci.,
Technol. 1997, 31, 837.
(10) Bhatia, S. K.; Shriver-Lake, L. C.; Prior, K. J.; Georger, J. H.; Calvert, Received for review November 11, 1997. Revised manuscript
J. M.; Bredehorst, R.; Ligler, F. S. Anal. Biochem. 1989, 178, 408. received May 14, 1998. Accepted May 26, 1998.
(11) Narang, U.; Anderson, G. P.; Ligler, F. S.; Burans, J. Biosens.
Bioelectron. 1997, 12, 937. ES970991P
2466 9 ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 32, NO. 16, 1998