APPLIED AND ENVIRONMENTAL MICROBIOLOGY,0099-2240/00/$04.0010

Oct. 2000, p. 4258–4265 Vol. 66, No. 10

Copyright © 2000, American Society for Microbiology. All Rights Reserved.

A Bead-Based Method for Multiplexed Identification andQuantitation of DNA Sequences Using Flow Cytometry

ALEXANDER SPIRO,1 MARY LOWE,1* AND DREW BROWN2†

Physics Department, Loyola College in Maryland, Baltimore, Maryland 212101, and Center for Marine Biotechnology,University of Maryland Biotechnology Institute, Baltimore, Maryland 212022

Received 11 May 2000/Accepted 24 July 2000

A new multiplexed, bead-based method which utilizes nucleic acid hybridizations on the surface of micro-scopic polystyrene spheres to identify specific sequences in heterogeneous mixtures of DNA sequences isdescribed. The method consists of three elements: beads (5.6-mm diameter) with oligomer capture probesattached to the surface, three fluorophores for multiplexed detection, and flow cytometry instrumentation. Twofluorophores are impregnated within each bead in varying amounts to create different bead types, eachassociated with a unique probe. The third fluorophore is a reporter. Following capture of fluorescent cDNAsequences from environmental samples, the beads are analyzed by flow cytometric techniques which yield asignal intensity for each capture probe proportional to the amount of target sequences in the analyte. In thisstudy, a direct hybrid capture assay was developed and evaluated with regard to sequence discrimination andquantitation of abundances. The target sequences (628 to 728 bp in length) were obtained from the 16S/23Sintergenic spacer region of microorganisms collected from polluted groundwater at the nuclear waste site inHanford, Wash. A fluorescence standard consisting of beads with a known number of fluorescent DNAmolecules on the surface was developed, and the resolution, sensitivity, and lower detection limit for measuringabundances were determined. The results were compared with those of a DNA microarray using the samesequences. The bead method exhibited far superior sequence discrimination and possesses features whichfacilitate accurate quantitation.

In environmental microbiology, there is a need to developmore rapid, sensitive, and accurate methods for detecting spe-cific DNA sequences in complex, heterogeneous mixtures ofDNA. This would enable the profiling of microbial communi-ties and quantification of changes in soil, sediment, and watermicrobiota as indicators of bioremediation. The issues in eval-uating a method include the ability to discriminate sequences,the quantitative precision and accuracy in measuring abun-dances of specific sequences, the ability to measure low con-centrations of sequences, the ease and rapidity of preparationand detection of samples, the degree of multiplexing (i.e., thenumber of sequences that can be detected in a single run), andcost.

A variety of molecular methods have been used for micro-bial profiling (13). One of these methods, oligonucleotide mi-croarrays, can handle a high degree of multiplexing, but at thistime, the slides and the equipment for printing the arrays anddetecting the signals are expensive and inconvenient to use ona routine basis. An even greater concern is that, despite effortsto construct a useful array for environmental samples, therestill is no product which meets the needs for sequence discrim-ination and quantitation of abundances. Previous oligonucle-otide microarrays for environmental samples have targetedgenes for monitoring ecotoxicity (3) or rRNA or 16S ribosomalDNA (rDNA) of nitrifying bacteria (7), Escherichia coli andVibrio proteolyticus (4), and Geobacter chapelleii (D. P. Chand-ler, D. A. Stahl, and J. F. Gaillard, DOE-NABIR PI Work-shop, abstract, 2000). These arrays used samples obtained fromculturable microorganisms. Although there are many technicalchallenges to overcome, the multiplexing capabilities of mi-

croarray technology are a powerful tool for dealing with theenormous diversity of natural populations, and for that reason,improvements in this and related technologies are being ex-plored.

In the present study, we develop a new multiplex method,related to oligonucleotide microarrays but based on micro-scopic polystyrene spheres and flow cytometry, that can iden-tify individual sequences in mixtures of DNA sequences. The“bead method” involves the following elements: microspheresbearing carboxyl groups on the surface, three fluorophores formultiplexed detection, and flow cytometry instrumentation.Two classification fluorophores (in this study, red and orangedyes) are impregnated throughout the volume of the beads invarying discrete amounts, thereby creating distinct populationsof “bead types” distinguishable by their red and orange inten-sities. The DNA assay is conducted on the surface of the beads,and a third fluorophore (in this case, green dye) is used as thereporter. In flow cytometry, the beads are directed single fileinto a thin fluid column where they are interrogated one at atime by a laser.

An oligonucleotide (“capture probe”), designed to be com-plementary to a particular target sequence, is attached to thesurface of a unique bead type. The analyte consists of a mixtureof green-labeled DNA sequences, of which some are targetedby the capture probes. By mixing different bead types in asingle tube and exposing them to the same analyte, directhybrid capture occurs between matching capture probes andsequences. Using flow cytometry instrumentation, multiplexeddetection is accomplished through measurements of the red,orange, and green emission intensities and the forward andside scatter. Detection times are typically a few seconds to afew minutes per tube.

Several DNA assays on bead surfaces have been reported.The Luminex Corporation (Austin, Tex.; personal communi-cation) suggests that the best capture probe lengths for se-

* Corresponding author. Mailing address: Physics Department, Loy-ola College in Maryland, Baltimore, MD 21210. Phone: (410) 617-2709. Fax: (410) 617-2646. E-mail: mlowe@loyola.edu.

† Present address: 719 Brinkwood Rd., Baltimore, MD 21229.

4258

on April 5, 2020 by guest

http://aem.asm

.org/D

ownloaded from

quence discrimination are 18 to 20 bases and that larger signalscan be achieved with 22- to 24-mers. Longer capture probesrange from 45 to 564 bases (8, 9, 10, 22). For multiplexed beadformats, a competitive assay, a ligation assay, and capture ofbiotinylated sequences followed by indirect labeling using flu-orescent conjugates of streptavidin (5, 8, 20) have been dem-onstrated with eukaryotic and viral systems. Target sequenceswere 150 to 462 bp (8) or ;200 bp (Luminex, personal com-munication).

In this paper, we describe the design and performance of thebead method for identifying and quantifying prokaryotic se-quences obtained from an environmental sample. As in previ-ous studies (5, 8, 20) using a multiplexed bead approach, weworked with a Becton Dickinson flow cytometer, two-colorbeads, and a similar hybridization buffer. This work differs inother ways. We developed another hybridization procedureconsisting of direct hybrid capture of single-stranded (ss), flu-orescent target DNA because of the need for high-qualitysequence discrimination and accurate quantitation of concen-trations. We also used a different fluorescence label and opti-mized the detection buffer. Longer target sequences, morerealistic for environmental studies, were tested. As a step to-ward absolute quantitation of abundances, a bead standard wasprepared with a known number of fluorescent oligonucleotideson the surface. A more refined statistical analysis procedurewas adopted to characterize the performance of the beadmethod. A more fundamental understanding of factors foroptimizing detection was achieved.

Where possible, comparative results between the beadmethod and an oligonucleotide microarray are shown. Thiswork provides a basis for developing quantitative, multiplexedbead-based assays for any set of sequences.

MATERIALS AND METHODS

Capture probe design and oligonucleotide synthesis. Target sequences werealigned using the PileUp program from the Wisconsin Package (Genetics Com-puter Group). Oligonucleotides (see Table 1) were designed with OligoTech(Oligos Etc.) to avoid stem-loops and homodimers. The sequences for the beadsare substrings of the ones used for the microarray.

The capture probes for the beads were synthesized with a 59-amino modifica-tion called “unilinker” (Operon Technologies, Inc., or Oligos Etc.). Oligonucle-otide 1234, used on the bead standard, corresponds to human locus 20 and hasa 59-unilinker, 39-fluorescein modification. These oligonucleotides were recon-stituted in water to a concentration of 200 mM.

The PCR primers, R2 (forward) and R5FS (reverse), are located within the16S and 23S rDNA regions, respectively, flanking the intergenic spacer region(ISR). Primer R5FS (Operon Technologies, Inc.) was modified at the 59 end with“fluorescein-ON” and four phosphorothioate linkages between the five terminalbases; a phosphodiester bond exists between the fluorescein and the 59-terminalbase.

PCR amplification and quantitation of the amplicon concentration. The tem-plate DNA for the PCR amplifications was plasmid DNA (pDNA) containinginserts corresponding to ISR sequences 102 (728 bp), 204 (647 bp), 401 (591 bp),and 1404 (628 bp). These sequences were obtained by cloning the ISR ofmicroorganisms (6) filtered from polluted groundwater at the U.S. Departmentof Energy site at Hanford, Wash. (D. Brown and F. Robb, unpublished data).The pDNA was purified using the Qiagen Spin Miniprep kit and resuspended inwater.

PCRs were carried out in a total volume of 29.5 ml containing 2.5 U ofAmplitaq DNA polymerase (Perkin-Elmer), 13 GeneAmp PCR buffer suppliedwith the polymerase, final concentrations of 0.68 mM (each) primer and 55 mM(each) deoxynucleoside triphosphate, and 0.5 ml of pDNA. The thermal cyclingprotocol consisted of a 3-min denaturation step at 94°C, followed by addition ofAmplitaq (hot start) and 35 cycles of denaturation (30 s at 94°C), annealing (30s at 50°C), and extension (1 min 30 s at 72°C), followed by a final extension for5 min at 72°C. The PCR products were then stored at 4°C for up to several weeks.The concentrations of the PCR products were determined from a 1% agarose gelcontaining ethidium bromide and a titration of Low DNA Mass Ladder (LifeTechnologies catalog no. 10068-013). Typical concentrations were about 40 to 70fmol of DNA per ml.

Preparation of full-length, ss-PCR products. Based on reference 14, ss-PCRproducts were prepared by adding 10 to 20 U of T7 gene 6 exonuclease (U.S.Biochemical Corp., Cleveland, Ohio) to a mix of PCR product, water (optional),

and 13 exonuclease buffer supplied with the enzyme (53 buffer contains 200mM Tris-HCl [pH 7.5], 100 mM MgCl2, and 250 mM NaCl). Incubation occurredfor 45 min at 37°C, after which no band was visible on a gel. The 59-to-39hydrolytic activity of exonuclease is inhibited by the four phosphorothioatebonds. For the microarray experiments, the exonuclease reactions were stoppedwith 1 mM EDTA, pH 8. For the bead method, no stopping was used.

Attachment of capture probes to beads. The beads (2.5 3 106 in number) wereinitially washed with water, by centrifuging them at 10,000 rpm (5,600 3 g) for 1min and removing the supernatant. The beads were resuspended by vortexingthem in 50 ml of 0.1 M MES [2-(N-morpholino)ethanesulfonic acid; Sigma], pH4.5, containing 1 nmol of the capture probe. EDC [1-ethyl-3-(3-dimethylamin-opropyl)carbodiimidehydrochloride; Pierce] was diluted in water to a concentra-tion of 10 mg/ml; 2.5 ml of the solution was added to the beads. The mix wasvortexed and incubated at room temperature in the dark for 30 min. Addition ofEDC was repeated. The beads were then washed with 1 ml of 0.02% Tween 20,followed by two washes in 500 ml of 0.1% sodium dodecyl sulfate (SDS). Afterremoval of the supernatant, the beads were resuspended in 0.1 M MES (pH 4.5)to a final concentration of 5 3 104 beads/ml and stored in the refrigerator forseveral months. The bead concentration was checked with a hemacytometer.

Beads. Typically, four kinds of carboxylated beads of the same size (5.6 6 0.06mm in diameter) were used: plain (i.e., undyed) beads (Bangs Laboratories) andthree types of colored beads, labeled 8047, 8058, and 8069 (Luminex Corp.).These colored beads contain equal amounts of orange dye (no. 80) and varyingamounts of red dye (no. 47, 58, and 69). A unique capture probe was attached toeach bead type. The bead-probes were designated plain/102, 8047/204, 8058/102,and 8069/1404, where 8047/204 means that the capture probe complementary tothe 204 PCR product is attached to bead type 8047.

Direct hybrid capture of PCR products on bead surfaces (Fig. 1b). A 1.53TMAC buffer was prepared according to the directions of Luminex Corp. (per-sonal communication) to contain 3 M tetramethylammonium chloride (Sigma),0.1% SDS, 50 mM Tris-HCl (pH 8.0) (Amresco), and 4 mM EDTA (pH 8.0). Abead-probe mix was prepared by suspending 25,000 beads of each type in 1.53TMAC buffer for a total volume of 34 ml and prewarmed at the hybridizationtemperature of 46°C. The desired volume of ss-PCR product was brought up to17 ml with water, heated at 95°C for 10 min, combined with the bead mix, andincubated at 46°C for 1 h. The beads were washed one or two times in prewarmed13 TMAC buffer, resuspended in 200 ml of 13 high-pH TMAC buffer (pH 9.8;

FIG. 1. Hybridization procedures. (a) Modified sandwich assay on the DNAmicroarray. Capture probes were attached at the 59 end to a glass slide. Amixture of 59-fluoresceinated ssDNA and 39-fluoresceinated detection probeswas pipetted over the microarray. The detection probe was designed to beadjacent to the capture probe. Amplification of the fluorescence signal wasachieved with an antifluorescein antibody, Alexa 488. (b) Direct hybrid captureon beads. Capture probes were attached at the 59 end to the bead surface. Thebeads were then combined with a mixture of 59-fluoresceinated ssDNA se-quences.

VOL. 66, 2000 BEAD-BASED METHOD FOR MULTIPLEXED IDENTIFICATION 4259

on April 5, 2020 by guest

http://aem.asm

.org/D

ownloaded from

see below), and kept at room temperature in the dark for up to 24 h beforedetection.

Flow cytometry measurements. The beads were analyzed with a Becton Dick-inson FACScan flow cytometer using Ar laser excitation at 488 nm and bloodbank saline (0.90% [wt/vol] NaCl [pH 6.0 to 7.5] at 25°C; Fisher) for the sheathfluid. The core stream was 13 high-pH TMAC buffer (3 M TMAC, 0.1% SDS,50 mM Tris buffer, 4 mM EDTA [pH 8.0]). Forward and side light scatteringenabled the separation of monomeric beads signals from other signals based onsize discrimination. Green-orange compensation was used. Typically, 100 to2,000 events per bead type were collected with a total acquisition time of 15 to180 s per tube. Cytometry data were analyzed with CellQuest software.

With each set of samples, two types of bead standards were run. The responsefunction was determined with the LinearFlow Green Flow Cytometry Low In-tensity Calibration Kit (Molecular Probes). Surface-labeled beads (see Results)were used to convert intensity values into the number of fluorescent nucleic acidsequences (FNAS) per bead.

Statistical analysis of flow cytometry data. The following quantities were usedto analyze the data: number (N) of beads of a certain type; mean signal value (ns)for beads with DNA; mean background signal (nb) for beads with no DNAhybridized to the surface; signal (5 ns 2 n b) due to DNA; standard deviation(S); coefficient of variation (CV); standard error in the mean (s 5 S/N1/2);standard error in the difference of means sd 5 (ss

2 1 sb2)1/2, and probability (p)

for the confidence interval. (Subscripts s and b refer to signal and background,respectively). The lowest detectable DNA signal was determined by nmin 5 nb 1ksd, where sd was calculated from the lowest mean signal with the condition thatns 2 nb . ksd. In this paper, p 5 68% and k 5 1 for N $ 30. The resolution inthe signal is given by dns ' kss√2.

Fluorescence imaging microscopy. Individual beads were imaged with an up-right episcopic fluorescence microscope (Nikon Labophot-2) equipped with acooled charge-coupled device (CCD) camera (First Magnitude SpecIm-3) andan Ar laser (Laser Physics 150M-006-00B). A linearly polarized laser beam (488nm) was sent into the microscope and compressed by the microscope objectivelens onto the sample plane. The fluorescence emission was selected by two glasscutting filters and a band-pass interference filter (530 nm). The beads wereresuspended in high-pH TMAC buffer and placed between two quartz coverslips.

Absorption spectrum measurements. The concentration and spectroscopicproperties of dye-labeled oligonucleotides were tested at room temperatureusing a Spectronic Genesys-5 spectrophotometer.

DNA microarrays. The alignments, PCR amplifications, quantitation of am-plicons, and preparation of ss-target sequences are described above.

(i) Oligonucleotides (Table 1). Oligonucleotides (12-, 24-, and 25-mers) weresynthesized by Molecular Tool, Inc. (now called Orchid Biocomputer), accordingto the procedure described in reference 16. Printing of seven-by-seven arraysinvolved deposition of capture probes onto the surface of mercaptosilane-coatedglass slides (Molecular Tool, Inc.). The printing and surface chemistry of theattachment process are described in reference 16.

(ii) Hybridization. The hybridization procedure is depicted in Fig. 1a, in whicha fluoresceinated detection probe is first hybridized to an ss-fluoresceinatedtarget sequence, followed by capture onto the microarray. The capture anddetection probes are adjacent to each other. The purposes of the detection probeare to reduce secondary structures near the vicinity of capture, to create a helicaltwist to the DNA which would promote hybridization, and to provide an addi-tional fluorescent center for detection.

The analyte consisted of 13 GBA salts (1 mM cetyltrimethylammonium chlo-ride, 1.5 M NaCl, 10 mM EDTA), the desired ss-PCR product volume, 1 mM(each) detection probe, and water for a total volume of 40 ml. The microarrays

were prewarmed in a moist environment for 30 to 40 min at 37°C. The analytewas initially heated to 95°C for 10 min, held at 37°C for 5 to 10 min, pipetted ontothe warm arrays, and incubated in a moist environment for 3 h at 37°C. The slideswere washed three times in 13 TNTw (10 mM Tris-HCl, pH 7.5; 150 mM NaCl;0.05% Tween 20), dried on the back, hydrated with 13 TNTw on the front, andscanned.

(iii) Fluorescence amplification. To amplify the emission signal from thearrays, a 1:500 dilution of Alexa Fluor 488 conjugate of antifluorescein rabbitpolyclonal immunoglobulin G (Molecular Probes) was prepared using a buffercontaining 13 TNTw and 1% (wt/vol) fraction V bovine serum albumin andpipetted onto the arrays. Incubation occurred for at least 30 min at roomtemperature. The slides were washed three times in 13 TNTw, dried on the back,hydrated on the front, and scanned. There were 3.3 Alexa molecules per antibodymolecule.

(iv) Detection and analysis. Each slide was scanned with a Molecular Dynam-ics 575 FluorImager using an argon laser light source (488 nm) and a 530-nmemission filter. The images were analyzed using mostly ImageQuant software(Molecular Dynamics). Additional images were acquired using a cooled CCDcamera with a Fujinon TV lens HF35A-2, 488-nm laser excitation, two barrierglass filters, and a band-pass interference filter for selecting the green fluores-cence (530 nm) of the microarray.

The statistical analysis was similar to that for the beads (see above). Smallsample statistics were used in which N 5 6 was the number of spots of each type.

RESULTS

Spectroscopy, detection buffer, and bead standard. To find abuffer suitable for the core fluid, we measured the absorptionspectra of oligonucleotide 1234 (39-fluorescein) and a water-soluble fluorescein salt (Aldrich 16,630-8) in water, blood banksaline, TNTw (pH 7.4), MES (pH 4.5), and TMAC buffer (pH8.5 and 9.8). The results were compared with a fluoresceinsolution in 0.01 M NaOH which has known spectroscopic char-acteristics and with literature data for different prototropicforms of fluorescein molecules (11). Oligonucleotide 1234 ex-hibited the dianion form of fluorescein in TMAC (pH 9.8), themonoanion form in water and in saline, and the mixture ofthese two forms in the other liquids. In TMAC and water,primer R5FS (59-fluorescein) had the same absorption spec-trum as 1234. Since the dianion form has the highest fluores-cence efficiency (11, 19), TMAC buffer (pH 9.8) was used asthe reference liquid in all of the calibration procedures and asthe core fluid for the flow cytometry measurements.

The percentage of oligonucleotide 1234 labeled with fluo-rescein was determined to be 70% based upon extinction co-efficients and optical densities at 260 and 498 nm, assuming nofree fluorescein molecules.

A bead standard was prepared by attaching oligonucleotide1234 to plain beads. Using fluorescence microscopy, the image

TABLE 1. Oligonucleotide sequencesa

Oligonucleotidename Description Sequence 59 to 39

7102 204 array capture probe SS-CT AgC gAA AAg CgA TTC CCT TTC g7103 204 array detection probe ACC CgA gAg AgA-FL7105 1404 array capture probe SS-TT CTA TAg gCT TgT gTT CAC AAg TC7106 1404 array detection probe AAg CgT TCA CAC-FL7108 401 array capture probe SS-Ag AAA TCA ACA TTC CAC AgC CAC gC7109 401 array detection probe TgC AAA TgC TCA-FL7111 102 array capture probe SS-gT TgT ggg CTT ggg TAg Agg gAC TC7112 102 array detection probe CTC CCT gTg TCC-FLR2 PCR primer, forward ggg WgA AgT CgT AAC AAgR5FS PCR primer, reverse FL-T ∧T ∧A∧g ∧C ACg TCC TTC ATC gCC1234 Standard UL-AT ggT CTT CTg gTT gCC CCC-FL2000 102 bead capture probe UL-gg gCT Tgg gTA gAg ggA CTC2001 1404 bead capture probe UL-TT CTA gAg gCT TgT gTT CAC2002 204 bead capture probe UL-CT AgC gAA AAg CgA TTg ggT

a Capture probes and PCR primer sequences for the bead method and the DNA microarray. Additional oligonucleotides, not shown here, were synthesized for testingpurposes. SS, disulfide; UL, unilinker; FL, fluorescein; ∧, phosphorothioate linkage.

4260 SPIRO ET AL. APPL. ENVIRON. MICROBIOL.

on April 5, 2020 by guest

http://aem.asm

.org/D

ownloaded from

intensities of the beads and a solution of oligonucleotide 1234in high-pH TMAC buffer were compared. The mean intensityvalue per bead was found to be equivalent to 3.25 3 105 FNAS,corresponding to 4.6 3 105 oligonucleotides attached to thebead surface. From flow cytometry measurements, this beadstandard in TMAC buffer has an emission signal 13.6-foldhigher than that in water, 15-fold higher than that in saline,and 2-fold higher than that in TNTw due to activation of thedianion form of fluorescein.

Surface distribution and spatial orientation of captureprobes. Figure 2 shows images of the surface-labeled beadstandard (Fig. 2a) and a bead with green dye molecules im-pregnated in the volume (Fig. 2b). The increase in intensityfrom the center to the perimeter of the bead in Fig. 2a corre-sponds to the increase in the emitting surface projected onto atwo-dimensional plane. In Fig. 2b, the intensity is greatest inthe center, corresponding to the largest emitting volume pro-jected onto the image plane. For Fig. 2a, there is an intensityvariation of 15% along the perimeter, possibly correspondingto slight inhomogeneities in the surface attachment. Imageanalyses also show that fluorescein molecules conjugated tooligonucleotides have a nearly random orientation on the sur-face of the beads. This is because, with linearly polarized ex-citation, a difference of intensity between the orthogonal sec-tions of the bead perimeter is expected if transient dipoles indye molecules have a predominant orientation relative to thebead surface. In fact, we observed random spatial variations inthe intensities in Fig. 2a which cannot be attributed to orien-tational phenomena. From additional polarization measure-ments, we determined that there is partial immobilization oflabeled oligonucleotides on the bead surface during the 4-nsfluorescence lifetime of the dianions (11, 19), which mostlyhave isotropic orientation.

Hybridization procedure. We compared the behavior of ss-PCR and double-stranded (ds)-PCR products in the hybridiza-tion procedure. Beads were combined with 500 fmol of ds-PCRproduct in one tube or 500 fmol of ss-PCR product in anothertube. After hybridization, the beads were analyzed by fluores-cence microscopy and flow cytometry. As shown in Fig. 3a,uniform fluorescence was observed on the bead surface for thess-sample. In contrast, the ds-sample showed random patches

of hybridization of varying size and shape (Fig. 3b). From flowcytometry measurements, these beads had a mean intensitycorresponding to 8.3 3 102 FNAS with CV ' 40% and 5.5 3103 FNAS with CV ' 15% in the ds- and ss-samples, respec-tively. Based on these data, we believe that, for quantitativeanalyses, hybridizations using ss-PCR products are advanta-geous.

Direct and cross hybridization. Sequence discrimination wasevaluated for the bead method and the DNA microarray usingthe same environmental target sequences (102, 204, and 1404)and similar capture probe sequences. For the beads, each tubecontained four types of beads and 400 fmol of a single targetsequence, an amount corresponding to near-saturation condi-tions for direct hybridization. In Fig. 4a, the direct hybrid-ization signals from matching capture probes and target se-quences were normalized to 100%. The cross-hybridizationlevel was determined for each sequence relative to the di-rect-hybridization level. The lowest cross-hybridization level(0.07%) was observed between the 1404 capture probe andsequence 204; the highest level (0.98%) appeared between the102 capture probe and sequence 204. Similar results were ob-tained with 70, 100, 150, and 310 fmol of target sequences ineach tube. In all cases, the cross hybridization was not morethan 1%.

Results for the DNA microarray are shown in Fig. 4b. Threeassays were tested: fluoresceinated target DNA only, fluores-ceinated DNA with fluoresceinated detection probe (“no-Al-exa”), and fluoresceinated DNA with fluoresceinated detec-tion probe amplified with Alexa 488. Overall, the signals fromfluoresceinated DNA were low and considered insufficient foranalysis. The Alexa samples, shown in Fig. 4b, had highersignals and lower cross-hybridization levels than the no-Alexasamples. Nevertheless, the cross-hybridization level for mi-croarrays was significantly higher than that for beads. Thesaturation for direct hybridization was approximately $900fmol of DNA per array. At 890 fmol, the lowest cross-hybrid-ization level was 0.34% between the 204 capture probe and 102sequence (compared with 0.1% for the bead method). Thehighest cross-hybridization level was 46% between the 102capture probe and the 1404 sequence (compared with 0.55 to0.62% for beads).

Lower detection limit and resolution of the concentration ofDNA. The minimum detectable amount of DNA in the analytewas determined from an analysis of the concentration curves.Two bead assays were tested. The first had a total reactionvolume of 6 ml and 12,000 beads (i.e., 3,000 beads of each

FIG. 2. Fluorescence microscopy of individual beads using linearly polarizedlaser irradiation with electric field vector E. The spatial resolution on the sampleplane was 0.4 mm. The photon density of the excitation was 2 3 1017 photons/cm2 z s, and the CCD integration time was 1 s. (a) Bead standard (5.6-mmdiameter) with 39-fluoresceinated oligonucleotides on the surface. (b) MolecularProbes calibration bead (6-mm diameter).

FIG. 3. Enhanced fluorescence microscopy images of beads with 102 targetDNA hybridized to capture probes. (a) ss target DNA prepared with exonucle-ase. (b) ds target DNA.

VOL. 66, 2000 BEAD-BASED METHOD FOR MULTIPLEXED IDENTIFICATION 4261

on April 5, 2020 by guest

http://aem.asm

.org/D

ownloaded from

type). The second had a total volume of 51 ml and 100,000beads (i.e., 25,000 beads of each type). In Fig. 5a, ns is plottedas a function of amount C of target sequence 102 hybridized toplain/102 beads. For all points, the relative uncertainties ss/nsranged from 0.8 to 4.4%. The background nb of plain beadscorresponded to 252 FNAS/bead (sb 5 11 FNAS). Using ss 514 FNAS for the lowest target concentration in the 6-ml assay,the minimum detectable signal corresponded to nmin ' 270FNAS/bead. That is, 18 DNA molecules can be detected oneach bead above background with 68% probability. The min-imum detectable amount of DNA in the tube (Cmin) was de-termined by extrapolating the concentration curve to nmin toobtain 1.3 fmol for the 6-ml assay and 13 fmol for the 51-mlassay.

We define the sensitivity Sc in the concentrations as thederivative of the signal mean with respect to the concentrationat a given point on the concentration curve: Sc 5 (dns/dC). Theresolution in DNA amount in a tube (dC) is given by dC 5dns/Sc. For the middle of the curve, we then obtain Sc ' 52FNAS/bead z fmol and dC ' 1 fmol for the 6-ml assay and Sc '13 FNAS/bead z fmol and dC ' 7 fmol for the 51-ml assay.

The bead method and the DNA microarray were compared.Quantitation of the microarray was accomplished in terms ofthe number of fluorophore molecules per square micrometeron the glass slide according to reference 1, in which the fluo-rescence intensities of the microarray spots were comparedwith a thin solution layer of the same fluorescence label ofknown concentration. The microarray results are shown in Fig.5b, in which the vertical axis represents the number of fluo-rescent labels per spot and the horizontal axis represents theamount of DNA pipetted over each array. The backgroundsignals and uncertainties were different for the no-Alexa (ss/ns ; 7 to 30%) and Alexa (ss/ns ;4 to 15%) experiments. Thenmin value increased from 1.1 3 108 to 1.8 3 108 fluorescentlabels per spot. For the no-Alexa experiment, Sc was ;1.38 3106 fluorescent labels/spot z fmol, dC was ;60 fmol (at themiddle), and Cmin was ;15 fmol. For the Alexa samples, Sc was;6.6 3 106 fluorescent labels/spot z fmol, dC was ;17 fmol,and Cmin was 6 to 10 fmol.

Competition effect in mixtures. Since the goal of the beadmethod is to study mixtures of environmental DNA sequences,we evaluated the resolution in measuring concentrations forknown mixtures of three DNA sequences, 102, 204, and 1404,using two tests. The first (Fig. 6a) consisted of seven samples(tubes) where the concentrations of 102 and 1404 were varied,keeping 204 constant, and the total concentration was constantat 400 fmol. The second (Fig. 6b) consisted of five sampleswhere the concentration of 204 was increased, holding 102 and1404 constant. The total reaction volume in each tube was 51ml with 75,000 beads (25,000 of each type). The intersamplevariation in the hybridization levels was obtained for the se-quences held at a constant concentration and was found to bedifferent for the two tests. In Fig. 6a, the CV was 9% forsequence 204. However, in Fig. 6b, the mean intensity de-creased monotonically for 102 and 1404. The CV was no morethan 15% across all five tubes. There appears to be a compe-tition effect in mixtures in which the measured signal dependsupon whatever else is in the sample. The value of 15% repre-sents the intersample resolution in concentrations if competi-tion effects are disregarded.

DISCUSSION

Low levels of cross hybridization are important for quanti-fying abundances of microorganisms in environmental samplescontaining many unknown sequences. For the bead method,the cross-hybridization level never exceeded 1% of the direct-hybridization signal. This corresponds to 10 to 30 fluorescein-ated DNA molecules per bead, a value comparable to theminimum number of sequences that could be detected in ourexperiments. Thus, one can expect the actual cross-hybridiza-tion level to be even less. These results are far superior to theperformance of the oligonucleotide microarray.

Prior studies using a multiplexed bead assay for DNA alsodemonstrate excellent sequence discrimination. These studiesinclude an HLA tissue typing assay (5); a viral load assay inwhich six sequences associated with human immunodeficiencyvirus, hepatitis C virus, and herpes simplex virus were detectedat the femtomole level (20); and single nucleotide polymor-phism detection of nine single nucleotide polymorphism mark-

FIG. 4. Direct and cross hybridization. The analyte consisted of a singletarget sequence. The data for matching target and capture probes were normal-ized to 100%. (a) Bead method. The 102 capture probe was attached to two beadtypes, plain and 8058. (b) Oligonucleotide microarray. Amplification of fluores-cence with Alexa 488 was used.

4262 SPIRO ET AL. APPL. ENVIRON. MICROBIOL.

on April 5, 2020 by guest

http://aem.asm

.org/D

ownloaded from

ers located near the ApoE locus on chromosome 19 (8). Basedon these results and ours, we believe that the bead methodshould be able to distinguish other sets of sequences useful forenvironmental studies, including 16S rDNA sequences andbiodegradation alleles. Our preliminary experiments using 16SrDNA sequences indicate excellent cross-hybridization resultsbut a weaker signal in comparison with ISR sequences.

There are several possible reasons for the high-quality se-quence discrimination. All of the hybridizations were con-ducted in TMAC, which roughly equalizes the stability of theA:T and G:C base pairs, thereby reducing hybridization biasbased on sequence variations (21). In addition, thermal tests ofthe beads with capture probes indicated that the beads arestable at 95°C for short time intervals, and so the hybridizationprocedure involved 95 and 46°C steps, which are higher thanwhat could be used with the microarray. We did not conducthybridization tests using more conventional hybridization buff-ers; it is possible that if the target sequences are differentenough, then these buffers may be adequate (8).

The quantitation in the bead experiments was accomplishedin terms of FNAS units. MESF units (the number of moleculesof equivalent soluble fluorochromes), widely used in flow cy-tometry, do not indicate the actual number of sequences butonly how intense the dye-labeled sample (bead or biologicalcell) is relative to the same dye solution containing free flu-orophore molecules (17, 18). The problem is that, when deal-ing with labeled DNA, such commercial standards are of lim-ited use because the fluorescent properties of dyes can besignificantly changed upon conjugation with biopolymers andare sensitive to pH (2, 11, 12, 15, 19). In this work, a beadstandard was prepared with a known number of fluorescein-ated oligonucleotides on the surface; this is a spectrallymatched standard, responsive to the surrounding fluid, thatenabled us to make a transition from relative fluorescenceintensities to FNAS units. We believe this is a convenient andcomprehensive procedure to directly compare and evaluate theresults of different bead experiments and different types offluorescence detection.

Analysis of Fig. 5a and 6 indicates that ,0.1% of the totalamount of DNA in the analyte was actually used for fluores-cence detection. This occurred even when the bead surface wasnot fully occupied by target molecules and suggests that thesensitivity and lower detection limit of the bead method can beimproved by optimizing the hybridization conditions. How-ever, even at the current level, in which a minimum of severalfemtomoles of target DNA is needed in the tube, we are stillable to detect DNA without additional fluorescence amplifi-cation (e.g., multiply labeled target molecules and fluorescentantibodies). This is an advantage for accurate quantitation,where it is necessary to know the uncertainty associated witheach preparation step.

Further study of intersample statistics with DNA mixtures isnecessary. There may be uncontrolled variability in the prep-aration conditions (temperature, volume, number of beads,etc.). However, at best, since the intrasample uncertainty was 1to 4%, we believe this is the minimum CV value for inter-sample statistics. We speculate that the competition effectwhich changed the intersample statistics may be due to a de-crease in the diffusion rate at higher concentrations compli-cated by intermolecular interactions.

There have been many technical improvements with DNAmicroarrays, and further improvements in both the beadmethod and microarrays will continue to be made. Thus, anunequivocal judgment on the relative performance of thesetwo methods is difficult to make. However, given our data, withregard to the lower detection limit, we did not notice a signif-

FIG. 5. Dependence of hybridization on target amount (sequence 102) forthe bead method and the DNA microarray. (a) Two bead experiments using a6-ml (E) and a 51-ml (h) total volume in each tube. (b) Results from twomicroarray experiments, one using no fluorescence amplification (no-Alexa) (h)and the other using Alexa (E). For both graphs, the mean background value(horizontal dashed line) and the lower limit of detection nmin are shown. The68% confidence intervals are indicated by plus signs; in panel a, only the largestintervals are shown.

VOL. 66, 2000 BEAD-BASED METHOD FOR MULTIPLEXED IDENTIFICATION 4263

on April 5, 2020 by guest

http://aem.asm

.org/D

ownloaded from

icant difference between the bead method and the microarray.With regard to other considerations, including surface chem-istry, hybridization protocols, detection instrument, quantita-tion, ease of developing a useful assay, and cost, we believe thebead method has advantages for environmental applications.At this time, the surface chemistry on beads appears to bemore stable, enabling higher temperatures and simpler hybrid-ization protocols to be used while still achieving excellent se-quence discrimination. The hybridization reactions on beads arefaster than those on microarrays; this is most likely due to thethree-dimensional nature of the bead system, which allows mixingand close proximity between the beads and target molecules. Fordeveloping useful assays, the bead method is far more convenient;modifying the set of sequences to be studied is easily accom-plished by attaching new oligonucleotides to the beads andpipetting different beads together. Arrays, however, need to bereprinted, requiring specialized equipment and more time.

The bead method also has several features making it pref-erable to microarrays for quantitation. First, because hundredsor thousands of beads of each type are examined, good statis-tics are easy to achieve; on arrays, comparable statistics have tobe accomplished by either many replicate spots on high-densityarrays or very large spots. Second, for beads, our data showexcellent spatial homogeneity in the attachment of the captureprobes to the bead surface and also in the hybridization oftarget sequences to the capture probes. This is due to thespherical symmetry of the beads, which enables uniform con-ditions for the surrounding DNA. This level of homogeneitycannot be surpassed by microarrays. In fact, for large spots,Bogdanov et al. (1) showed that strong attachment in homo-geneities are induced by hydrodynamic redistribution of theDNA concentration inside the solution drop on hydrophobic

glass surfaces. Third, the random orientation of the fluoro-phores on the bead surface produces adequate conditions forquantitative comparison with solutions. In contrast, the fluoro-phores on microarrays may have a predominant orientation(1), making the arrays sensitive to the geometry of the opticalexcitation and its polarization.

With regard to cost and the ability to set up similar beadexperiments in any lab, there are several considerations. Whilethis paper concentrates on three sequences, we do not antici-pate difficulties in dealing with a larger number of captureprobes. Currently, 64 red-orange bead types that are suitablefor flow cytometers using an argon laser light source are com-mercially available (Luminex Corp.). Custom beads can also beordered. Flow cytometers, of the type used in this study, arequite common; thus, it may not be necessary for the researcherto buy specialized detection instrumentation. However, exist-ing software for quantitative analysis of the data is not opti-mized for real-time multiplexed detection but is effective forslow, careful postprocessing.

Finally, we wish to note that beads can be used to preparesequences for subsequent steps. By using a cell sorter, a searchfor new genes in environmental samples may be expedited.

ACKNOWLEDGMENTS

We acknowledge Frank Robb at COMB and the staff at MolecularTool, Inc. (Yu-Hui Rogers, Miriam Donaldson, Valery L. Bogdanov,and Michael Boyce-Jacino), for the target sequences, DNA microarraywork, and hosting M.L.; the staff at Luminex Corporation for sharingtheir protocols and experience; and Gina Hamlin Green and MarkSoloski at the Johns Hopkins University for use of the flow cytometer.

We acknowledge DOE NABIR grant DE-FG02-99ER62868. Pre-liminary work was partially funded by DOE DE-FG07-96ER62320 andNSF CTS-9253633.

FIG. 6. Dependence of hybridization on the composition of the analyte. The target sequences were 102 (E), 204 (h), and 1404 (3). (a) All tubes contained 400fmol of DNA in total. The concentration of sequence 204 was kept constant at 100 fmol for tubes 1 to 7. The amount of 102 increased from 0 to 300 fmol, while theamount of 1404 decreased from 300 to 0 fmol in decrements of 50 fmol. (b) The total number of DNA molecules was varied from tube to tube. The amounts of 102and 1404 were kept constant at 100 and 200 fmol, respectively. The concentration of 204 was 0 fmol in tube 1, 25 fmol in tube 2, 50 fmol in tube 3, 100 fmol in tube4, and 200 fmol in tube 5.

4264 SPIRO ET AL. APPL. ENVIRON. MICROBIOL.

on April 5, 2020 by guest

http://aem.asm

.org/D

ownloaded from

REFERENCES

1. Bogdanov, V. L., Y.-H. Rogers, and M. Boyce-Jacino. 1997. Fluorescentimaging and quantitation of solid support bound nucleic acids. Proc. SPIE2985:129–137.

2. Der-Ballan, G. P., N. Kameda, and G. L. Rowley. 1988. Fluorescein labelingof Fab9 while preserving single thiol. Anal. Biochem. 173:59–63.

3. Doktycz, M. J., and K. L. Beattie. 1997. Genosensors and model hybridiza-tion studies, p. 205–225. In A. J. Beugelsdiik (ed.), Automation technologiesfor genome characterization. John Wiley & Sons, Inc., New York, N.Y.

4. Eggers, M. D., W. J. Balch, L. G. Mendoza, R. Gangadharan, A. K. Mallik,M. G. McMahon, M. E. Hogan, D. Xaio, T. R. Powdrill, B. Iverson, G. E. Fox,R. C. Willson, K. I. Maillard, J. L. Siefert, and N. Singh. 1997. Advancedapproach to simultaneous monitoring of multiple bacteria in space. In Pro-ceedings of the 27th International Conference on Environmental Systems.SAE Technical Paper Series 972422. Society of Automotive Engineers, Inc.,Warrendale, Pa.

5. Fulton, R. J., R. L. McDade, P. L. Smith, L. J. Kienker, and J. R. Kettman.1997. Advanced multiplexed analysis with the FlowMetrix system. Clin.Chem. 43:1749–1756.

6. Gurtler, V., and V. A. Stanisich. 1996. New approaches to typing and iden-tification of bacteria using the 16S-23s rDNA spacer region. Microbiology142:3–16.

7. Guschin, D. Y., B. K. Mobarry, D. Proudnikov, D. A. Stahl, B. E. Rittmann,and A. D. Mirzabekov. 1997. Oligonucleotide microchips as genosensors fordeterminative and environmental studies in microbiology. Appl. Environ.Microbiol. 63:2397–2402.

8. Iannone, M. A., J. D. Taylor, J. Chen, M.-S. Li, P. Rivers, K. A. Slentz-Kesler, and M. P. Weiner. 2000. Multiplexed single nucleotide polymor-phism genotyping by oligonucleotide ligation and flow cytometry. Cytometry39:131–140.

9. Imai, T., Y. Sumi, M. Hatakeyama, K. Fujimoto, H. Kawaguchi, N. Ha-yashida, K. Shiozaki, K. Terada, H. Yajima, and H. Handa. 1996. Selectiveisolation of DNA or RNA using single-stranded DNA affinity latex particles.J. Colloid Interface Sci. 177:245–249.

10. Jacobsen, C. S. 1995. Microscale detection of specific bacterial DNA in soilwith a magnetic capture-hybridization and PCR amplification assay. Appl.Environ. Microbiol. 61:3347–3352.

11. Klonis, N., and W. H. Sawyer. 1996. Spectral properties of the prototropicforms of fluorescein in aqueous solution. J. Fluorescence 6:147–157.

12. Kumke, M. U., G. Li, and L. B. McGown. 1995. Hybridization of fluorescein-labeled DNA oligomers detected by fluorescence anisotropy with proteinbinding enhancement. Anal. Chem. 67:3945–3951.

13. Madsen, E. L. 1998. Epistemology of environmental microbiology. Environ.Sci. Technol. 32:429–439.

14. Nikiforov, T. T., R. B. Rendle, M. L. Kotewicz, and Y.-H. Rogers. 1994. Theuse of phosphorothioate primers and exonuclease hydrolysis for the prepa-ration of single-stranded PCR products and their detection by solid-phasehybridization. PCR Methods Appl. 3:285–291.

15. Nunnally, B. K., H. He, L.-C. Li, S. A. Tucker, and L. B. McGown. 1997.Characterization of visible dyes for four-decay fluorescence detection inDNA sequencing. Anal. Chem. 69:2392–2396.

16. Rogers, Y.-H., P. Jiang-Baucom, Z.-J. Huang, V. Bogdanov, S. Anderson,and M. T. Boyce-Jacino. 1999. Immobilization of oligonucleotides onto aglass support via disulfide bonds: a method for preparation of DNA microar-rays. Anal. Biochem. 266:23–30.

17. Schwartz, A., and E. Fernandez-Repollet. 1993. Development of clinicalstandards for flow cytometry. Ann. N. Y. Acad. Sci. 677:28–39.

18. Schwartz, A., G. E. Marti, J. W. Gratama, and E. Fernandez-Repollet. 1998.Standardizing flow cytometry: a classification system of fluorescence stan-dards used for flow cytometry. Flow Cytometry 33:106–114.

19. Sjoback, R., J. Nygren, and M. Kubista. 1995. Absorption and fluorescenceproperties of fluorescein. Spectrochim. Acta A51:L7–L21.

20. Smith, P. L., C. R. WalkerPeach, R. J. Fulton, and D. B. DuBois. 1998. Arapid, sensitive, multiplexed assay for detection of viral nucleic acids usingthe FlowMetrix system. Clin. Chem. 44:2054–2056.

21. Wood, I. W., J. Gitschier, L. A. Lasky, and R. M. Lawn. 1985. Base compo-sition-independent hybridization in tetramethylammonium chloride: amethod for oligonucleotide screening of highly complex gene libraries. Proc.Natl. Acad. Sci. USA 82:1585–1588.

22. Zammatteo, N., I. Alexandre, I. Ernest, L. Le, F. Brancart, and J. Remacle.1997. Comparison between microwell and bead supports for the detection ofhuman cytomegalovirus amplicons by sandwich hybridization. Anal. Bio-chem. 253:180–189.

VOL. 66, 2000 BEAD-BASED METHOD FOR MULTIPLEXED IDENTIFICATION 4265

on April 5, 2020 by guest

http://aem.asm

.org/D

ownloaded from