Analytical Biochemistry 401 (2010) 271–279

Contents lists available at ScienceDirect

Analytical Biochemistry

journal homepage: www.elsevier .com/locate /yabio

Sensitive detection of multiplex toxins using antibody microarray

Wei Lian a, Donghai Wu b, Daniel V. Lim c, Shouguang Jin a,*

a Department of Molecular Genetics and Microbiology, University of Florida College of Medicine, Gainesville, FL 32610, USAb Key Laboratory of Regenerative Biology, Guangzhou Institute of Biomedicine and Health, Chinese Academy of Sciences, Guangzhou, Guangdong 510663, Chinac Department of Cell Biology, Microbiology, and Molecular Biology, University of South Florida, Tampa, FL 33620, USA

a r t i c l e i n f o

Article history:Received 9 January 2010Received in revised form 3 February 2010Accepted 25 February 2010Available online 10 March 2010

Keywords:Antibody microarrayDetectionRicinCholera toxinSEBNanoparticle

0003-2697/$ - see front matter � 2010 Elsevier Inc. Adoi:10.1016/j.ab.2010.02.040

* Corresponding author. Fax: +1 352 273 8905.E-mail address: sjin@mgm.ufl.edu (S. Jin).

1 Abbreviations used: PCR, polymerase chain reaimmunosorbent assay; Ab, antibody; NRL, Naval Reseacoupled device; NP, nanoparticle; FLISA, fluorescence-IgG, immunoglobulin G; CT, cholera toxin; SEB, staphytrimethoxysilylpropyldiethylenetriamine; RuBpy, tris(2um(II) hexahydrate; TEOS, tetraethylorthosilicate; NHEDC, 1-ethyl-3-(3-dimethylaminopropyl)carbodiimihydroxysuccinimide; mAb, monoclonal antibody; LC,phosphate-buffered saline; BSA, bovine serum albuminTween 20; HRP, horseradish peroxidase; pAb, polyclonaPAGE, sodium dodecyl sulfate–polyacrylamide gel elecidene fluoride; LOD, limit of detection.

a b s t r a c t

Using a newly developed fluorescent nanoparticle (NP) that gives rise to a high-intensity and stable fluo-rescent light, a sensitive antibody (Ab) microarray assay system has been developed for specific detectionof bioterrorism agents, as exemplified by ricin, cholera toxin (CT), and staphylococcal enterotoxin B (SEB).The Ab microarray uses a sandwich format that consists of capture Abs, analytes (toxins), biotinylateddetection Abs, and avidin-conjugated NP. In all three cases, polyclonal Abs (pAbs) displayed superiorityover monoclonal antibodies (mAbs) in capturing toxins on microarray slides even when the pAbs andmAbs had similar affinity as determined by enzyme-linked immunosorbent assay (ELISA). The detectionsystem was successfully used to detect toxins spiked in milk, apple cider, and blood samples. We wereable to detect ricin at 100 pg/ml in buffer and at 1 ng/ml in spiked apple cider or milk, whereas CT andSEB were detected at 10 pg/ml in buffer and 100 pg/ml in spiked apple cider or milk. High specificitieswere also demonstrated in the detection of mixed toxin samples with similar sensitivities. The matrixeffect of blood samples on the detection of mixed toxins seems to be minimal when the toxin concentra-tion is at or above 100 ng/ml. The current study highlights the significant role of pAb and NP in increasingselectivity and sensitivity of toxin detection in a microarray format.

� 2010 Elsevier Inc. All rights reserved.

To combat bioterrorism, great efforts have been put forth to de-velop new methods for rapid, sensitive, and accurate detection ofbioterrorism agents. To date, the most sensitive methods remainto be the polymerase chain reaction (PCR)1 and enzyme-linkedimmunosorbent assay (ELISA). The PCR-related methods are more sui-ted for detection of organisms, such as bacteria and viruses, fromwhich nucleic acids can be extracted for specific amplification. TheELISA-based methods are more robust for detection of toxins, whichare often proteins in nature. In comparison with the PCR method,the ELISA-based detection is less sensitive to the matrix effect and pre-sumably gives fewer false-positive results. In addition, the automatedaccessory fluidics of biosensors makes the detection fast and easy to

ll rights reserved.

ction; ELISA, enzyme-linkedrch Laboratory; CCD, charge-

linked immunosorbent assay;lococcal enterotoxin B; DETA,,20-bipyridyl)dichlororutheni-4OH, ammonium hydroxide;

de hydrochloride; NHS, N-liquid chromatography; PBS,; PBST, PBS containing 0.05%l antibody; Ag, antigen; SDS–trophoresis; PVDF, polyvinyl-

handle. The current study focuses on the detection of environmentaland bacterial toxins by using an antibody (Ab) microarray.

A large number of biosensors have recently been developed forthe purpose of real-time detection of pathogens and toxic agents inthe fields and clinical laboratories [1–6]. Among them, the arraybiosensor designed by the Naval Research Laboratory (NRL) is themost well known in the field [7–13]. The central part of the NRLbiosensor is an ELISA-like immunoassay conducted on glass slidesthat also serve as the evanescent waveguide. A diode laser thatgives rise to an evanescent wave excites the fluorescence-labeledanalyte, and a charge-coupled device (CCD) camera collects theemission light. This array biosensor has been used to detect com-plex samples of a variety of pathogens and toxins. The limits ofdetection were shown to be as low as 103 cfu/ml for bacteria and0.1 ng/ml for toxins [9,11].

Driven by genomic and proteomic studies, protein microarrayhas emerged as a powerful new technology. Although not aswidely used as DNA microarray, protein microarray has alreadybeen applied to clinical diagnosis [14–17]. For example, the proteinarrays commercialized by Whatman/Schleicher & Schuell, havebeen used to detect cytokines and other biomarkers in human ser-um (http://www.schleicher-schuell.com). The array system takes amultiplex format and is able to detect cytokines to the picogram-per-milliliter (pg/ml) level. Protein microarray has also been usedto study protein–protein interactions, protein modifications, pro-tein profiling, and enzymatic reactions [18–21].

272 Sensitive detection of multiplex toxins / W. Lian et al. / Anal. Biochem. 401 (2010) 271–279

Previously, we reported the applications of a nanoparticle (NP)to a variety of labeling techniques, including fluorescence-linkedimmunosorbent assay (FLISA), immunocytochemistry, immunohis-tochemistry, DNA microarray, and protein microarray [22]. Using asandwich microarray, we were able to specifically detect humanand mouse immunoglobulin Gs (IgGs) to 1 pg/ml. In the currentstudy, we used the NP-labeled microarray to detect the bioterror-ism agents ricin, cholera toxin (CT), and staphylococcal enterotoxinB (SEB) in a multiplexed fashion. With mixed and spiked samples,we were able to detect toxins with high specificity in addition tosensitivity that is superior to any reported biosensors or micro-arrays for toxins.

Materials and methods

Materials

Trimethoxysilylpropyldiethylenetriamine (DETA) was pur-chased from United Chemical Technologies (Bristol, PA, USA).Tris(2,20-bipyridyl)dichlororuthenium(II) hexahydrate (RuBpy),tetraethylorthosilicate (TEOS), Triton X-100, n-hexanol, cyclohex-ane, ammonium hydroxide (NH4OH, 28–30 wt%), 1-ethyl-3-(3-dimethylaminopropyl)carbodiimide hydrochloride (EDC), and N-hydroxysuccinimide (NHS) were purchased from Sigma–Aldrich(St. Louis, MO, USA). Avidin was purchased from Molecular Probes(Eugene, OR, USA).

Ricin (A chain) was purchased from Sigma–Aldrich. CT was pur-chased from EMD Biosciences (La Jolla, CA, USA). SEB was pur-chased from Toxin Technology (Sarasota, FL, USA). Monoclonalantibodies (mAbs) against ricin were purchased from ResearchDiagnostics (Flanders, NJ, USA). Rabbit anti-ricin Abs were ob-tained from U.S. Biological (Swampscott, MA, USA) and Sigma–Al-drich. mAbs against CT and rabbit anti-CT Abs were purchasedfrom Biodesign International (Saco, ME, USA). mAbs against SEBwere purchased from Anogen (Mississauga, Ontario, Canada). Rab-bit anti-SEB and sheep anti-SEB were purchased from ToxinTechnology.

The biotinylation reagent biotin–LC (liquid chromatography)–NHS was obtained from Pierce (Rockland, IL, USA). Streptavidin-la-beled Cy3 was obtained from Amersham Biosciences (ArlingtonHeights, IL, USA). Microarray Super Epoxy substrate was obtainedfrom Erie Scientific (Portsmouth, NH, USA). A MicroCaster Hand-held Arrayer was purchased from Whatman/Schleicher & Shuell(Dassel, Germany).

NP synthesis and surface modification

The NP was synthesized by using a microemulsion method asdescribed previously [22]. Briefly, the microemulsion was formedby mixing cyclohexane (7.5 ml), n-hexanol (1.8 ml), Triton X-100(1.77 ml), 20 mM RuBpy dye in water (0.48 ml), and TEOS(0.1 ml). After mixing for 20 min, 60 ll of NH4OH was added to ini-tiate the polymerization. The reaction was allowed to continue for24 h. When the polymerization was complete, an equal volume ofacetone was added and vortexed to break the microemulsion state.The solidified silica NP was collected by centrifugation at 3000g for10 min and washed three times with 95% ethanol. Betweenwashes, the NP was dispersed by vortex and sonication. The result-ing NPs are uniform in shape with an average diameter of 70 nm, asdescribed previously [22]. The surface modification of NP was per-formed with DETA, EDC, and NHS chemistry, as described previ-ously [22]. Avidin was conjugated to NP through an NHS-functional group and the excess NHS ester was quenched by add-ing hydroxyamine and Tris–HCl (pH 7.5), as described previously[22].

Sandwich FLISA

Microtiter plates were coated with 100 ll per well of captureAbs (10 lg/ml) in phosphate-buffered saline (PBS). The plates werekept at room temperature for 2 h or at 4 �C overnight, followed byblocking with 2% bovine serum albumin (BSA) for 1–2 h. Samplescontaining analytes or controls were added to each well and incu-bated at room temperature for 1–2 h, followed by three washeswith PBST (PBS containing 0.05% Tween 20). The bound analyteswere detected by probing with biotinylated Abs (10 lg/ml) specificto the toxins for 1–2 h at room temperature, followed by labelingwith avidin-conjugated NP in 0.1% BSA/PBS for 30 min. Afterwashes, 0.1 ml of PBS was added into each well and the fluores-cence intensity was measured using a Tecan Safire plate readerwith excitation/emission wavelengths of 430/590 nm.

ELISA and determination of Ab titers

To determine Ab titers, 100 ll of toxins (1 lg/ml) in PBS wasadded to each well of a 96-well plate (Nunc MaxiSorp immuno-plate), incubated for 2 h at room temperature or overnight at4 �C, and washed three times with PBST. Abs specific to the toxinswere diluted serially and added into the coated plate. After incuba-tion for 1–2 h at room temperature, the plate was probed with anexcess amount of horseradish peroxidase (HRP)-conjugated sec-ondary Ab, followed by color development [23]. A titration curvewas plotted for each Ab. Based on the midpoints of the titrationcurves, the relative titer for each Ab was determined.

Microarray analysis of toxins

Capture Abs were diluted to 0.5 mg/ml in PBS containing 40%glycerol and printed on Super Epoxy microarray slides using a man-ual MicroCaster microarrayer (Whatman/Schleicher & Schuell). Toperform multiplexed assays on the same slide, multiple blocks wereprinted as subarrays. A hydrophobic Pap pen was used to draw aboundary around each block within which the applied sample wascontained. The arrayed Super Epoxy substrate was good for use forup to 1 month when stored at room temperature. Before applyinganalytes, the array was inverted face down and blocked with 2%BSA in PBS for 1–2 h. Thereafter, excess capture Abs were washedaway with the blocking buffer by shaking. Dilutions of analytes in0.1% BSA or the spiked samples were applied and incubated for 2 hat room temperature or overnight at 4 �C with gentle agitation. Ex-cess toxins were removed by three washes with PBST, each for5 min with shaking. Then biotinylated detection Abs, diluted to1 lg/ml with 0.1% BSA/PBS, were added to the corresponding blocksand incubated for 1–2 h at room temperature. The arrays werewashed with PBST and blocked briefly with 1% BSA/PBS before prob-ing with avidin–NP. The avidin–NP was diluted to 100 lg/ml withPBST containing 0.1% BSA and probed the array for 30 min. The la-beled slide was washed with PBST three times (5 min each), rinsedwith cold water briefly, and dried by either centrifugation or blow-ing with the compressed nitrogen gas. The slides were scanned usinga GenePix 4000B microarray scanner (Axon Instruments, Foster City,CA, USA), and the data were processed using GenePix Pro 3.0 soft-ware (Axon Instruments).

Preparation of spiked apple cider and milk

Buffered apple cider was prepared by adding 1/10 volume of10� PBS stock solution and 1/20 volume of 2% BSA and adjustedto pH 7.4. Serial dilutions of toxins were made in buffered apple ci-der as well as in buffer solution (0.1% BSA/PBS, pH 7.4) for compar-ison. A volume of 100 ll was applied to the microarray.

Fig. 1. Comparison of fluorescence signals derived from Cy3 and NP in the detection of CT and SEB by sandwich FLISA. Here a 96-well plate coated with mAbs against CT orSEB was incubated with the respective toxins at the indicated concentrations. The plate was then probed with the respective detection Abs that are biotinylated together withavidin-conjugated NP or Cy3. A.U., arbitrary units.

Sensitive detection of multiplex toxins / W. Lian et al. / Anal. Biochem. 401 (2010) 271–279 273

Milk (2% reduced fat) purchased from a local grocery store wasused to spike with either single or mixed toxins. For comparisonpurposes, the same serial dilutions were made with either milkor buffer (0.1% BSA/PBS) and subjected to the microarray analysis.

Preparation of spiked serum samples

Serum samples were collected from three healthy volunteers. Atoxin mixture was prepared in PBS buffer that contained ricin, CT,and SEB, each at a concentration of 1 lg/ml. The toxin solution wasthen serially diluted at a 1:10 ratio with the serum from each of thethree individuals. As controls, 1 lg/ml BSA was serially dilutedwith the sera in parallel. Four dilutions were made from 0.1 to100 ng/ml of each toxin and were subjected to microarraydetections.

Safety considerations

All work was performed in a licensed biosafety level 2 labora-tory. The toxins were handled according to the Guidelines for Work

Fig. 2. Detection of CT and ricin by Western blot. Molecular weight standards (M), CT,probed separately with mAb or pAb against each of the two toxins.

with Toxins of Biological Origin and the Biodefense Safety ProgramTechnical Safety Requirement (see http://www.ehs.ufl.edu/Bio/ forthe latter).

Results and discussion

Fluorescent NP renders higher detection sensitivity

A silica NP encapsulating the inorganic fluorescent dye RuBpywas previously shown to be highly fluorescent and resistant tophotobleaching [22]. We compared fluorescence signals derivedfrom NP and Cy3 in a sandwich FLISA, which is identical to an ELISAexcept that it uses fluorescent labeling instead of enzymatic reac-tion. A microtiter plate was coated with mAbs against ricin, CT,or SEB as capture Abs, followed by incubation with the respectivetoxins at various concentrations. The plate was then probed withthe respective detection Abs that are biotinylated together withavidin-conjugated NP or Cy3. As shown in Fig. 1, the assay labeledwith NP gives a signal approximately 2 orders of magnitude stron-ger compared with the traditional labeling by Cy3, demonstrating

and ricin (R) were separated on SDS–PAGE, transferred onto PVDF membrane, and

Table 1Antibody titers.

Monoclonal antibody Polyclonal antibody

Ricin 1:400 1:1600CT 1:10,000 1:10,000SEB 1:800 1:10,000

Fig. 3. Comparison of mAb and pAb as capture Abs. Microarray slides werepatterned as follows: left and right columns with mAb and pAb, respectively, upperand lower rows with anti-CT and anti-SEB, respectively. Each slide was probed withone analyte (CT or SEB) at four different concentrations and the correspondingbiotinylated detection Ab and avidin–NP.

Fig. 4. Dose–response curves for parallel detection of toxins. (A) Multiplexed microarrayand pAb SEB. Each slide was probed with four different concentrations of ricin, CT, or SEBCorrelation between fluorescence intensity and toxin concentration. Dashed lines show

274 Sensitive detection of multiplex toxins / W. Lian et al. / Anal. Biochem. 401 (2010) 271–279

that the NP is a much better labeling agent than traditionalfluorophores.

Detection of toxins in microarray format

Our original design for microarray was to use mAb as captureAb and to use polyclonal antibody (pAb) as detection Ab. The ratio-nale was based on the prediction that mAbs should have higherspecificity for antigen (Ag) binding and, thus, could be used as cap-ture Ab to achieve high selectivity in capturing target analytes. Forthe detection Ab, pAb would amplify the detection signal by bind-ing at multiple sites on the analytes. This design is also practical gi-ven that mAbs are much higher in cost than pAbs. For the detectionAb, a relatively large amount of Ab is needed for biotinylation aswell as for probing the microarray chips. On the other hand, forthe capture Ab, only a small volume of concentrated Ab is neededfor printing microarray slides.

As capture Abs, mAbs against ricin, CT, and SEB were printed onglass slides that were coated with epoxy group. After blocking with2% BSA/PBS and washing away the excess Abs, various amounts ofricin, CT, and SEB were applied to the multiplexed arrays. Follow-ing incubation and washes, the arrays were probed with therespective pAb that was biotinylated and fluorescence signal wasgenerated by incubation with avidin-conjugated NP. With numer-ous trials, detection of CT was reasonable; however, detection of ri-cin was problematic due to cross-reaction with CT, whereasdetection sensitivity for SEB was very poor. We were interestedin understanding the cause of such variations.

To test whether the observed cross-reactivity between ricin andCT in the microarray format can be attributed to the quality of the

s were patterned with biotinylated human IgG (positive control), pAb ricin, pAb CT,, followed by detecting with detection Abs and fluorescence labeling with NP. (B–D)the linear dynamic ranges. A.U., arbitrary units.

Sensitive detection of multiplex toxins / W. Lian et al. / Anal. Biochem. 401 (2010) 271–279 275

mAbs, Western blot analysis was performed. Ricin and CT wereseparated on sodium dodecyl sulfate–polyacrylamide gel electro-phoresis (SDS–PAGE), transferred onto polyvinylidene fluoride(PVDF) membrane, and probed separately with mAb or pAb againsteach of the three toxins. As shown in Fig. 2, all four Abs recognizedcorresponding toxins without cross-reactivity. Therefore, thecross-reactivity observed on microarray is not likely caused bythe nonspecificity of the Ab itself.

Fig. 4 (cont

Next, we speculated that the intrinsic affinity of the Ab maydetermine the outcome. The affinities of the Abs were evaluatedby measuring their titers by conventional ELISA. As described inMaterials and methods, a Nunc MaxiSorp immunoplate was coatedwith 1 lg/ml toxin (ricin, CT, or SEB) and incubated with corre-sponding Abs serially diluted from an identical starting proteinconcentration. The relative titers were determined as the mid-points of the titration curves. As shown in Table 1, the relative titer

inued)

276 Sensitive detection of multiplex toxins / W. Lian et al. / Anal. Biochem. 401 (2010) 271–279

of mAb against CT is much higher than that of mAb against ricin orSEB, explaining the better outcomes of detecting CT over ricin orSEB in the initial experiments. This suggests that an important ele-ment for the success of immunoassays is the formation of a stableAg–Ab complex. Because the stability of the Ag–Ab complex isdetermined by the affinity of Ag–Ab binding as well as by thevalency of the binding, it can be predicted that pAbs, with similarspecificity, will have a better Ag-capturing capability than mAbsdue to the formation of more stable complex with their Ags bybinding to multiple epitopes.

Comparison of mAb and pAb for Ag capture

Using the microarray format, we compared pAb and mAb forAg-capturing abilities. Taking advantage of the fact that the titersof mAb and pAb against CT are comparable, the detection limitsof the microarrays were determined using each of the two Abs ascapture Ab and using the same pAb as detection Ab. As the resultsin Fig. 3 show, a 10-fold higher detection sensitivity was achievedwhen pAb was used as the capture Ab than when mAb was used,indicating that with similar affinity (titer) by ELISA, pAb has a

Fig. 5. Comparisons of toxins spiked in matrices and buffer. (A) Serially diluted CT, SEB, aBecause of different detection sensitivities, only the points within linear ranges are plrepresent averages of five fluorescence spots. A.U., arbitrary units.

superior capture function over mAb. Consistent with this, higherdetection sensitivities for ricin and SEB were also achieved in themicroarray format when pAb was used as capture Ab than whenmAb was used even when the capture pAbs were diluted to thesame titers as the capture mAbs, demonstrating the superiorityof pAb over mAb as capture Ab in the microarray. The multivalencyof the Ag binding capacity likely contributes to this, rendering amuch more stable Ag–Ab complex.

Using pAbs as capture Abs resulted in increased fluorescenceintensity and, therefore, also increased the signal-to-noise ratio.The increase of signal-to-noise ratio is crucial to the improvementof detection sensitivity and reduction of cross-reactivity. It is note-worthy that the pAbs against CT and SEB detected correspondingtoxins at the lowest concentration tested without obvious cross-reactivity (Fig. 3).

The major hindrance to Ab microarray development is the costof mAb production. Moreover, it is estimated that only 5–20% ofcommercial mAbs are useful for microarray, possibly due to lowaffinity [24]. Based on our experimental data, we propose usingaffinity-purified pAb or a pool of mAbs, if available, for any targetmolecule bearing more than one epitope because binding at

nd ricin in apple cider (AC) or buffer were subjected to detection by the microarray.otted. (B) Serially diluted toxins in milk compared with those in PBS buffer. Data

Sensitive detection of multiplex toxins / W. Lian et al. / Anal. Biochem. 401 (2010) 271–279 277

multiple sites can dramatically increase the stability of the Ag–Abcomplexes. Furthermore, the same pAb or pooled mAb can be usedas both capture and detection Abs, whereas a pair of mAbs recog-nizing two different epitopes are needed for capture and detectionAbs, respectively, in sandwich assays.

Multiplex and parallel detection of toxins

Our ultimate goal was to make multiplexed microarrays for thedetection of multiple bioterrorism agents at the same time. In thisstudy, we chose three toxins (ricin, CT, and SEB) as examples to testthe assay system. A new microarray was generated, as shown inFig. 4A. Biotinylated detection Abs were printed on the top rows aspositive controls to normalize data from slide to slide, followed bypAbs against ricin, CT, or SEB in the next three consecutive rows. Se-rial dilutions of each analyte were probed and visualized by therespective detection Ab (biotinylated) and avidin–NP. The correla-tions between fluorescence intensity and concentration of each ana-lyte are shown in Fig. 4B–D. A limit of detection (LOD) was set foreach toxin to generate clearly visible fluorescent spots on scannedimages that correlated to fluorescent intensity of 400 arbitrary units.The LOD for ricin was 100 pg/ml (P = 0.008758 vs. blank) (Fig. 4B)with a linear dynamic range between 1 and 100 ng/ml(R2 = 0.9927), the LOD for CT was 10 pg/ml (P = 0.011493) with a lin-ear dynamic range between 100 pg/ml and 1 ng/ml (R2 = 0.9997)(Fig. 4C), and the LOD for SEB was 10 pg/ml (P = 0.000466) with a lin-ear dynamic range between 10 pg/ml and 1 ng/ml (R2 = 0.9428)(Fig. 4D). These detection limits are lower than those for any previ-ously reported immunoassays. For instance, Goldman and cowork-ers reported the use of quantum dots and Abs to detect a similarset of toxins in a sandwich ELISA format [25]. When detected indi-vidually, detection limits were 3, 10, and 30 ng/ml for SEB, CT, andricin, respectively. They further reported a multiplexed fluoroimmu-noassay, performing sandwich immunoassays for the detection oftoxins simultaneously in single wells of a microtiter plate; however,no LOD was determined in that study [25].

Detection of toxins spiked into apple cider and milk

To test the practical use of the detection system, spiked applecider and milk were used as samples. In each experiment, paralleldilutions of the analytes were made in either buffer or the

Fig. 6. Simultaneous detection of toxins by detection cocktail. (A) Microarray slides weremixture. (B) Serially diluted mixtures of ricin, CT, and SEB were probed, followed by de

matrices. As shown in Fig. 5A, within their linear dynamic ranges,CT, SEB, and ricin diluted in apple cider have similar dose–responsecurves as in the PBS buffer, although the fluorescent signals weredecreased by approximately 50% and the LODs were increased byapproximately 1 order of magnitude. The LODs for CT and SEB wereincreased from 10 to 100 pg/ml, whereas the LOD for ricin was in-creased from 100 pg/ml to 1 ng/ml. Clearly, to obtain the samefluorescent signal as CT, it requires 10-fold higher SEB and 100times more ricin, consistent with the sensitivities observed earlier.The matrix effects of apple cider to ELISA as well as PCR have beenreported, and in a less robust detection the matrix effect couldmake the signal undetectable [26,27]. Because we had strong sig-nals at the beginning, sensitive detection in the matrix can stillbe achieved. Similar results from spiked milk are shown inFig. 5B except that the milk’s matrix effect on CT was more dra-matic than on the other two toxins. GM1 ganglioside is known tobe the natural receptor for CT, and the presence of GM1 in milkhas been well documented [28]; thus, it is possible that the pres-ence of GM1 in milk interfered with the binding by the Abs. Clearly,the matrix effects of different food samples differ on different ana-lytes due to either the endogenous Ab or other nonspecific bindingeffects. Therefore, for the purpose of quantification, a standardcurve for each matrix is required for accurate determination of tox-in contaminations in each food type.

Simultaneous detection of toxins

All of the detections described above used only one detectionAb for a specific toxin such that three slides are needed to detectthree different toxins. If a cocktail of Abs can be used withoutcross-reactivity, we should be able to detect multiple toxins in asingle microarray format. To test the feasibility, ricin (10 ng/ml),CT (10 ng/ml), and SEB (10 ng/ml) were probed individually as wellas in a mixture. The detection Ab cocktail was made by mixing1 lg/ml each of biotinylated anti-ricin, anti-CT, and anti-SEB Abs.The cocktail was used in place of the individual detection Abs suchthat one slide was used instead of three slides. As shown in Fig. 6A,single toxins can be detected without cross-reactivity; thus, themixture of the three toxins can be detected simultaneously. Whenserially diluted mixtures were probed by the same method, as theresults in Fig. 6B show, the detection limits for each toxin were thesame as in parallel assays of detecting single toxins, again

patterned as before. Ricin, CT, and SEB (10 ng/ml) were probed separately and in atection with antibody cocktail and fluorescent NP.

278 Sensitive detection of multiplex toxins / W. Lian et al. / Anal. Biochem. 401 (2010) 271–279

demonstrating no significant cross-reactivity among the threedetection Abs. Therefore, simultaneous detection of ricin, CT, andSEB was achieved.

Simultaneous detection of spiked serum samples

Taking advantage of the simultaneous detection format, mix-tures of ricin, CT, and SEB were spiked into serum samples and ap-plied to the microarray. Serum samples were collected from threehealthy volunteers. A mixture solution of toxins containing ricin,CT, and SEB, each at a concentration of 1 lg/ml, was made. The tox-in solution was then serially diluted at a 1:10 ratio in serum sam-ples from the three individuals, whereas BSA was used as anegative control. Four dilutions were made from 0.1 to 100 ng/mlof each toxin and subjected to microarray detections.

Results shown in Fig. 7 demonstrate the serum matrix effects onricin, CT, and SEB. In the detection of ricin (Fig. 7A), the fluores-cence signals from the three serum samples were decreased 5- to10-fold compared with the BSA control, and the LOD was also in-

Fig. 7. Simultaneous detection of spiked serum samples. Mixtures of ricin, CT, andSEB were serially diluted in PBS/BSA or serum from three different individuals,followed by microarray detection. Dose–response curves for ricin (A), CT (B), andSEB (C) are plotted. A.U., arbitrary units.

creased approximately 10-fold. In the detection of CT and SEB(Fig. 7B and C, respectively), the fluorescent signals from the serumsamples dropped more sharply compared with those of ricin detec-tion, and there were larger variations among the three individuals.At the concentration of 10 ng/ml, the fluorescence intensities had a100-fold decrease for the serum 3 sample when compared with theBSA control.

It has been reported that it is difficult to detect SEB spiked inhuman blood, urine, or saliva [7]. Our experimental results suggestthat serum has an inhibitory effect, possibly due to some of itscomponents binding to the analytes. In cases of strong inhibition,there might be a specific Ag–Ab interaction, likely due to priorexposure to the same epitope. Interestingly, the inhibitory effectof the serum becomes significant only when the toxin concentra-tions are at or below 10 ng/ml, whereas the detection is not af-fected when the toxin concentrations are at or above 100 ng/ml,implicating neutralization of all of the toxin binding Ab moleculesin the serum by the excess amount of toxins.

Conclusions

We have developed an NP-mediated Ab microarray system forthe detection of bioterrorism agents exemplified by ricin, CT, andSEB toxins. High sensitivity, specificity, and reproducibility wereachieved. The major hindrance in protein microarray developmentis the lack of efficient capture and detection agents. mAb was con-sidered as the method of choice; however, we found that substitut-ing mAb with highly purified pAb, even though having similar titerby ELISA, could dramatically improve the microarray performance.A likely explanation is that pAb enhances the Ag capture by bind-ing multiple sites on the analyte. The microarray format allows amultiplexed, high-throughput, and high-parallel assay. Further-more, the miniature feature permits low consumption of both sam-ple and reagents, reducing the amounts of biohazard waste as wellas the costs to conduct the assay. It should be pointed out that themicroarray scanner that we used in the current study is designedfor Cy3 and Cy5 and, thus, is not optimal for the detection of NP.Once a proper detector with optimal excitation/emission wave-lengths for NP becomes available, the detection signals should in-crease by 20 times from the current level based on our previousestimate [22].

Acknowledgments

We thank H.V. Baker for allowing us to use the microarray scan-ner in acquiring fluorescence intensity data. This work was sup-ported in part by the UF Opportunity Fund (to S.J.) and theNational Basic Research Program of China (to D.W., 2010CB945500).

References

[1] S. Kwakye, V.N. Goral, A.J. Baeumner, Electrochemical microfluidic biosensorfor nucleic acid detection with integrated minipotentiostat, Biosens.Bioelectron. 21 (2006) 2217–2223.

[2] D. Dell’Atti, S. Tombelli, M. Minunni, M. Mascini, Detection of clinicallyrelevant point mutations by a novel piezoelectric biosensor, Biosens.Bioelectron. 21 (2006) 1876–1879.

[3] L. Stoica, R. Ludwig, D. Haltrich, L. Gorton, Third-generation biosensor forlactose based on newly discovered cellobiose dehydrogenase, Anal. Chem. 78(2006) 393–398.

[4] G. Liu, Y. Lin, Biosensor based on self-assembling acetylcholinesterase oncarbon nanotubes for flow injection/amperometric detection oforganophosphate pesticides and nerve agents, Anal. Chem. 78 (2006) 835–843.

[5] J.A. Hansen, J. Wang, A.N. Kawde, Y. Xiang, K.V. Gothelf, G. Collins, Quantum-dot/aptamer-based ultrasensitive multi-analyte electrochemical biosensor, J.Am. Chem. Soc. 128 (2006) 2228–2229.

[6] S.A. Soper, K. Brown, A. Ellington, B. Frazier, G. Garcia-Manero, V. Gau, S.I.Gutman, D.F. Hayes, B. Korte, J.L. Landers, D. Larson, F. Ligler, A. Majumdar, M.Mascini, D. Nolte, Z. Rosenzweig, J. Wang, D. Wilson, Point-of-care biosensor

Sensitive detection of multiplex toxins / W. Lian et al. / Anal. Biochem. 401 (2010) 271–279 279

systems for cancer diagnostics/prognostics, Biosens. Bioelectron. 21 (2006)1932–1942.

[7] F.S. Ligler, K.E. Sapsford, J.P. Golden, L.C. Shriver-Lake, C.R. Taitt, M.A. Dyer, S.Barone, C.J. Myatt, The array biosensor: portable, automated systems, Anal. Sci.23 (2007) 5–10.

[8] K.E. Sapsford, Y.S. Shubin, J.B. Delehanty, J.P. Golden, C.R. Taitt, L.C. Shriver-Lake, F.S. Ligler, Fluorescence-based array biosensors for detection ofbiohazards, J. Appl. Microbiol. 96 (2004) 47–58.

[9] C.R. Taitt, Y.S. Shubin, R. Angel, F.S. Ligler, Detection of Salmonella entericaserovar Typhimurium by using a rapid, array-based immunosensor, Appl.Environ. Microbiol. 70 (2004) 152–158.

[10] C.R. Taitt, J.P. Golden, Y.S. Shubin, L.C. Shriver-Lake, K.E. Sapsford, A. Rasooly,F.S. Ligler, A portable array biosensor for detecting multiple analytes incomplex samples, Microb. Ecol. 47 (2004) 175–185.

[11] K.E. Sapsford, C.R. Taitt, N. Loo, F.S. Ligler, Biosensor detection of botulinumtoxoid A and staphylococcal enterotoxin B in food, Appl. Environ. Microbiol. 71(2005) 5590–5592.

[12] M.C. Moreno-Bondi, C.R. Taitt, L.C. Shriver-Lake, F.S. Ligler, Multiplexedmeasurement of serum antibodies using an array biosensor, Biosens.Bioelectron. 21 (2006) 1880–1886.

[13] M.M. Ngundi, C.R. Taitt, F.S. Ligler, Simultaneous determination of kineticparameters for the binding of cholera toxin to immobilized sialic acid andmonoclonal antibody using an array biosensor, Biosens. Bioelectron. 22 (2006)124–130.

[14] W. Zhang, J. Huang, M.F. Zhou, L.Y. Chen, Y.P. Ding, H.J. Cao, Y.Y. Geng,S.Q. Wang, Protein chip for detection of different HCV antibodies:preparation, quality control, and clinical evaluation, Mol. Diagn. 9 (2005)81–87.

[15] F. De Ceuninck, E. Marcheteau, S. Berger, A. Caliez, V. Dumont, M. Raes, P.Anract, G. Leclerc, J.A. Boutin, G. Ferry, Assessment of some tools for thecharacterization of the human osteoarthritic cartilage proteome, J. Biomol.Tech. 16 (2005) 256–265.

[16] C. Belluco, E. Mammano, E. Petricoin, L. Prevedello, V. Calvert, L.Liotta, D. Nitti, M. Lise, Kinase substrate protein microarray analysisof human colon cancer and hepatic metastasis, Clin. Chim. Acta 357(2005) 180–183.

[17] Y. Li, R.J. Schutte, A. Abu-Shakra, W.M. Reichert, Protein array method forassessing in vitro biomaterial-induced cytokine expression, Biomaterials 26(2005) 1081–1085.

[18] M. Letarte, D. Voulgaraki, D. Hatherley, M. Foster-Cuevas, N.J. Saunders, A.N.Barclay, Analysis of leukocyte membrane protein interactions using proteinmicroarrays, BMC Biochem. 6 (2005) 2.

[19] D.S. Gembitsky, K. Lawlor, A. Jacovina, M. Yaneva, P. Tempst, A prototypeantibody microarray platform to monitor changes in protein tyrosinephosphorylation, Mol. Cell. Proteomics 3 (2004) 1102–1118.

[20] M.B. Mintz, R. Sowers, K.M. Brown, S.C. Hilmer, B. Mazza, A.G. Huvos, P.A.Meyers, B. Lafleur, W.S. McDonough, M.M. Henry, K.E. Ramsey, C.R. Antonescu,W. Chen, J.H. Healey, A. Daluski, M.E. Berens, T.J. Macdonald, R. Gorlick, D.A.Stephan, An expression signature classifies chemotherapy-resistant pediatricosteosarcoma, Cancer Res. 65 (2005) 1748–1754.

[21] B.T. Houseman, J.H. Huh, S.J. Kron, M. Mrksich, Peptide chips for thequantitative evaluation of protein kinase activity, Nat. Biotechnol. 20 (2002)270–274.

[22] W. Lian, S.A. Litherland, H. Badrane, W. Tan, D. Wu, H.V. Baker, P.A. Gulig, D.V.Lim, S. Jin, Ultrasensitive detection of biomolecules with fluorescent dye-doped nanoparticles, Anal. Biochem. 334 (2004) 135–144.

[23] E. Harlow, D. Lane, Antibodies: A Laboratory Manual, Cold Spring HarborLaboratory Press, Cold Spring Harbor, NY, 1988.

[24] V. Sakanyan, High-throughput and multiplexed protein array technology:protein–DNA and protein–protein interactions, J. Chromatogr. B 815 (2005)77–95.

[25] E.R. Goldman, A.R. Clapp, G.P. Anderson, H.T. Uyeda, J.M. Mauro, I.L. Medintz,H. Mattoussi, Multiplexed toxin analysis using four colors of quantum dotfluororeagents, Anal. Chem. 76 (2004) 684–688.

[26] C.F. Hsu, T.Y. Tsai, T.M. Pan, Use of the duplex TaqMan PCR system fordetection of Shiga-like toxin-producing Escherichia coli O157, J. Clin. Microbiol.43 (2005) 2668–2673.

[27] S. Park, R.A. Durst, Modified immunoliposome sandwich assay for the detectionof Escherichia coli O157:H7 in apple cider, J. Food Prot. 67 (2004) 1568–1573.

[28] J. Holmgren, A.M. Svennerholm, C. Ahren, Nonimmunoglobulin fraction ofhuman milk inhibits bacterial adhesion (hemagglutination) and enterotoxinbinding of Escherichia coli and Vibrio cholerae, Infect. Immun. 33 (1981) 136–141.