A Generic Particle-Based Nonradioactive Homogeneous Multiplex Method for High-Throughput Screening...

12
A Generic Particle-Based Nonradioactive Homogeneous Multiplex Method for High-Throughput Screening Using Microvolume Fluorimetry Christine Martens, 1 Alice Bakker, Ana Rodriguez, Richard Bruce Mortensen, and Ronald W. Barrett Affymax Research Institute, 4001 Miranda Avenue, Palo Alto, California 94304 Received December 8, 1998 We have developed a novel fluorescence-based ho- mogeneous binding assay for high-throughput screen- ing of chemical compounds. In this assay, a Cy5- or Cy5.5-labeled ligand binds to receptor immobilized on a particle, either a bead or a cell. The resulting local- ized signal can be detected by a modified microvolume fluorimeter (MVF). When a molecule which competes with the labeled ligand is present, the localized fluo- rescence on cells or beads is reduced. Image process- ing software enumerates events and analyzes fluores- cence intensity. We describe MVF assays for the IL-1 and IL-5 receptors. Using synthetic peptides with a range of affinities for the IL-1 receptor, we obtained IC 50 data consistent with those determined by radioli- gand binding assays. Because the image processing software can discriminate among events with differ- ent diameters, we were able to develop a multiplex assay, in which the IL-1R and IL-5R assays were car- ried out in the same well with each receptor immobi- lized on a different size of bead. IC 50 values generated in the multiplex assay for ligands specific to each re- ceptor were comparable to those determined indepen- dently. Finally, similar IC 50 values were obtained in a 16-ml volume in an 864-well plate. This homogeneous, nonradioactive, miniaturizable, and multiplex-capa- ble assay holds much promise for screening of combi- natorial libraries and compound collections. © 1999 Academic Press Recent advances in the identification of human dis- ease-related genes have revealed many new molecules which can be investigated as possible targets for phar- maceutical agents. At the same time, the development of combinatorial chemistry has provided new technol- ogies for preparing unprecedented numbers of poten- tial drug molecules to be screened. As the number of molecules increases, novel assay formats must be de- veloped to accommodate the demand for rapid, minia- turized assays of tiny amounts of chemical compounds which result from combinatorial synthetic methods. To take full advantage of the rapid pace of target identification, it is desirable to develop simple assay formats which utilize generic reagents and which can be adapted to new targets with a minimum of develop- ment effort. The demands of large-scale robotics make nonradioactive assays very attractive, and the trend toward miniaturization places a very high value on homogeneous formats to eliminate wash or separation steps. Ideally, high signals and low background noise should allow very short read times for each well. A volumetric capillary cytometry system (1) has been developed by Biometric Imaging, Inc. (Mountain View, CA) for enumeration of fluorescently labeled blood cells, and this instrument has been modified to detect fluorescent tags on cells or beads in a 96-well or higher density plate format (2). Using this microvolume fluo- rimeter (MVF 2 ), we have developed a sensitive homo- geneous equilibrium binding assay to detect the bind- ing of a fluorescent ligand to its receptor, which is immobilized on beads or expressed on cells. The fluo- rescent reagents are inexpensive and can be used as tags on generic detection reagents, such as streptavi- din, or can be coupled to custom ligands using standard 1 To whom correspondence should be addressed. Fax: 650-424- 9860. 2 Abbreviations used: MVF, microvolume fluorimeter; bn-, biotin- conjugated; BSA, bovine serum albumin; NHS, N-hydroxysuccinim- ide; ECD, extracellular domain; MBP–IL-1RA, fusion protein of E. coli maltose-binding protein with interleukin-1 receptor antagonist; Pl-PLC, phosphoinositol-dependent phospholipase-C. 20 0003-2697/99 $30.00 Copyright © 1999 by Academic Press All rights of reproduction in any form reserved. Analytical Biochemistry 273, 20 –31 (1999) Article ID abio.1999.4184, available online at http://www.idealibrary.com on

Transcript of A Generic Particle-Based Nonradioactive Homogeneous Multiplex Method for High-Throughput Screening...

Page 1: A Generic Particle-Based Nonradioactive Homogeneous Multiplex Method for High-Throughput Screening Using Microvolume Fluorimetry

AMU

CaA

R

miCaiflwricarIgsearlicd1nbnA

ew

9

2

Analytical Biochemistry 273, 20–31 (1999)Article ID abio.1999.4184, available online at http://www.idealibrary.com on

Generic Particle-Based Nonradioactive Homogeneousultiplex Method for High-Throughput Screening

sing Microvolume Fluorimetry

hristine Martens,1 Alice Bakker, Ana Rodriguez, Richard Bruce Mortensen,nd Ronald W. Barrettffymax Research Institute, 4001 Miranda Avenue, Palo Alto, California 94304

eceived December 8, 1998

mootmvtw

ifbmnthss

dCcfldrgiirtd

ci

We have developed a novel fluorescence-based ho-ogeneous binding assay for high-throughput screen-

ng of chemical compounds. In this assay, a Cy5- ory5.5-labeled ligand binds to receptor immobilized onparticle, either a bead or a cell. The resulting local-

zed signal can be detected by a modified microvolumeuorimeter (MVF). When a molecule which competesith the labeled ligand is present, the localized fluo-

escence on cells or beads is reduced. Image process-ng software enumerates events and analyzes fluores-ence intensity. We describe MVF assays for the IL-1nd IL-5 receptors. Using synthetic peptides with aange of affinities for the IL-1 receptor, we obtainedC50 data consistent with those determined by radioli-and binding assays. Because the image processingoftware can discriminate among events with differ-nt diameters, we were able to develop a multiplexssay, in which the IL-1R and IL-5R assays were car-ied out in the same well with each receptor immobi-ized on a different size of bead. IC50 values generatedn the multiplex assay for ligands specific to each re-eptor were comparable to those determined indepen-ently. Finally, similar IC50 values were obtained in a6-ml volume in an 864-well plate. This homogeneous,onradioactive, miniaturizable, and multiplex-capa-le assay holds much promise for screening of combi-atorial libraries and compound collections. © 1999

cademic Press

Recent advances in the identification of human dis-ase-related genes have revealed many new moleculeshich can be investigated as possible targets for phar-

1

To whom correspondence should be addressed. Fax: 650-424-860.

cP

0

aceutical agents. At the same time, the developmentf combinatorial chemistry has provided new technol-gies for preparing unprecedented numbers of poten-ial drug molecules to be screened. As the number ofolecules increases, novel assay formats must be de-

eloped to accommodate the demand for rapid, minia-urized assays of tiny amounts of chemical compoundshich result from combinatorial synthetic methods.To take full advantage of the rapid pace of target

dentification, it is desirable to develop simple assayormats which utilize generic reagents and which cane adapted to new targets with a minimum of develop-ent effort. The demands of large-scale robotics make

onradioactive assays very attractive, and the trendoward miniaturization places a very high value onomogeneous formats to eliminate wash or separationteps. Ideally, high signals and low background noisehould allow very short read times for each well.A volumetric capillary cytometry system (1) has been

eveloped by Biometric Imaging, Inc. (Mountain View,A) for enumeration of fluorescently labeled bloodells, and this instrument has been modified to detectuorescent tags on cells or beads in a 96-well or higherensity plate format (2). Using this microvolume fluo-imeter (MVF2), we have developed a sensitive homo-eneous equilibrium binding assay to detect the bind-ng of a fluorescent ligand to its receptor, which ismmobilized on beads or expressed on cells. The fluo-escent reagents are inexpensive and can be used asags on generic detection reagents, such as streptavi-in, or can be coupled to custom ligands using standard

2 Abbreviations used: MVF, microvolume fluorimeter; bn-, biotin-onjugated; BSA, bovine serum albumin; NHS, N-hydroxysuccinim-de; ECD, extracellular domain; MBP–IL-1RA, fusion protein of E.

oli maltose-binding protein with interleukin-1 receptor antagonist;l-PLC, phosphoinositol-dependent phospholipase-C.

0003-2697/99 $30.00Copyright © 1999 by Academic Press

All rights of reproduction in any form reserved.

Page 2: A Generic Particle-Based Nonradioactive Homogeneous Multiplex Method for High-Throughput Screening Using Microvolume Fluorimetry

cracevflmawrbgmm

M

bToApNLw

satPbspt1ft(aslwatauraanwtss

sluwwgtr

cbSio

ar17(fbwi0iambaicPss

abfitbwo2sbaawttbMs

21PARTICLE-BASED HOMOGENEOUS BINDING ASSAY

hemistries. The assay format can be adapted to a wideange of analyses, including receptor binding, enzymessays, immunoassays, and nucleic acid binding, andan be miniaturized with little additional developmentffort. In this paper we describe the use of the micro-olume fluorimeter to measure binding of two differentuorescent ligands to the IL-1 receptor and to deter-ine IC50 values for a series of peptides with a range of

ffinities. The results are compared to those obtainedith other assay methods. A second assay, for the IL-5

eceptor, is also described, and we demonstrate thatoth receptors can be assayed independently in a sin-le well. The ability to perform multiplex assays in ainiaturized format is a powerful advantage of thisethod for very high-throughput screening assays.

ATERIALS AND METHODS

Chemical reagents. Streptavidin, bovine serum al-umin, biotinylated bovine serum albumin (bn-BSA),ween 20, buffer components, Sepharose G-25, andther chemical reagents were purchased from Sigma–ldrich Chemical Co. (St. Louis, MO). Biotin–NHS wasurchased from Molecular Probes (Eugene, OR). Cy5-HS and Cy5.5-NHS were purchased from Amershamife Science (Philadelphia, PA). Tissue culture mediaere purchased from GIBCO/BRL (Grand Island, NY).Biochemical reagents. The receptors used in these

tudies were soluble extracellular domain (ECD) prep-rations which were prepared as described (3). Briefly,he IL-1 and IL-5 receptor ECDs were expressed in aI-glycan-linked form in Chinese hamster ovary fibro-lasts (3, 4). Soluble ECD was released from the cellurface by digestion with phosphoinositol-dependenthospholipase C. In some cases the receptor was par-ially purified by affinity chromatography on MBP-IL-RA (for the IL-1R) or on MAb179 (for the IL-5R). Ausion protein of Escherichia coli maltose-binding pro-ein (MBP) with the IL-1 receptor antagonist proteinMBP-IL-1RA, see below) (5, 6) was expressed in E. colind purified by affinity chromatography on an amylo-e–Sepharose column. This material was used as un-abeled competitor in competition binding assays, oras labeled on free amino groups with Cy5-NHS esterccording to the manufacturer’s directions. After reac-ion with Cy5-NHS, the labeled MBP–IL-1RA was sep-rated from unreacted dye on a Sepharose G-25 col-mn. Fractions were tested for their ability to bind toeceptor-coated beads. Active fractions were poolednd used at the dilution determined to be optimal byssay. Purified human IL-5 was a gift from M. McKin-on (Glaxo-Wellcome, Stevenage, UK) and was labeledith Cy5.5-NHS according to manufacturer’s direc-

ions. Labeled IL-5 was purified and tested as de-cribed for MBP–IL-1RA. IL-1-binding peptides were

ynthesized on an Applied Biosystems peptide synthe- l

izer using Fmoc-protected amino acids (SynPep, Dub-in, CA). Peptide sequences are shown in Table 1. Forse as a tracer, the 15-amino-acid peptide, AF11733,as labeled at the amino terminus with Cy5-NHS andas purified on a reverse-phase HPLC column using aradient of 20–40% acetonitrile in water plus 0.1%rifluoroacetic acid. The structure of the active mate-ial was confirmed by mass spectrometry.An IgG fraction of monoclonal antibody MAb179 was

onjugated to biotin–NHS using a 10-fold molar ratio ofiotin:IgG. Conjugated material was purified over aepharose G-25 column before use. Radioligand bind-

ng assays for the IL-1-binding peptides were carriedut as described (4).Beads. The 10-mm beads used in these studies weremonodisperse, custom-synthesized macroporous sty-

ene/divinylbenzene copolymer, functionalized with a,12-diaminododecane linker (7). The 25-, 45-, and5-mm beads were purchased from Polysciences, Inc.Warrington, PA). The 78.1-mm beads were purchasedrom Bangs Laboratories, Inc. (Fishers, IN). The 10-mmeads were prepared for coating by an initial ethanolash. To coat beads, 5 3 106 beads were washed twice

n 1 ml of phosphate-buffered saline (PBS) containing.05% NaN3 and were resuspended in 1 ml. Beads werencubated sequentially with each of the following re-gents: 50 mg/ml of bn-BSA in PBS–NaN3, 1 h; 50g/ml of streptavidin in PBS–NaN3, 1 hp 50 mg/mln-MAb179, 1 h; and 2 mg/ml purified receptor ECD ordilution of PI-PLC harvested receptor (3), 1 h. Each

ncubation was carried out at room temperature withonstant mixing and was followed by two washes withBS–NaN3 plus 1% BSA. Receptor-coated beads weretored in PBS–NaN3 plus 1% BSA at 4°C until use theame day.Assay setup. The 96-well plates with clear bottom

nd black sidewalls were purchased from Costar (Cam-ridge, MA). For the 96-well plate format, assays had anal volume of 100 ml. Two assay designs were used. Inhe preimmobilized receptor format, IL-1R-coatedeads were prepared as described above. After the finalash step beads were resuspended at a concentrationf 1 3 105 beads per ml in PBS/1% BSA/0.05% Tween-0/0.05% NaN3. For competition assays, a 1:3 dilutioneries of the competing peptides was prepared in thisuffer. Cy5-MBP–IL-1RA or Cy5AF11733 solution wasdded to a final concentration of 3.75 nM. Finally, anliquot of the bead suspension containing 2500 beadsas added to each well. Plates were incubated at room

emperature in the dark for 2 h and were then read onhe microvolume fluorimeter. In the second format,eads were coated with bn-BSA, streptavidin, and bn-Ab179 as described above and then washed and re-

uspended as described for receptor-coated beads. Di-

ution series of peptides were prepared and fluorescent
Page 3: A Generic Particle-Based Nonradioactive Homogeneous Multiplex Method for High-Throughput Screening Using Microvolume Fluorimetry

ltw2isiamar

ww8ts

siflsdHaasd1omEwfeftsfiPstFiodespodsdee

ca

R

l1ttowacwt1bitlebarcstsC5uow

ttawswcmairbsdn

1iug

22 MARTENS ET AL.

igands were added as described above. Receptor washen added to a final concentration of 1.25 nM. Wellsere incubated for 30 min, at which time an aliquot of500 beads was added to each well. Following a furtherncubation for 2 h, plates were read on the MVF in-trument. This assay format is termed “postincubationmmobilization.” For cell-based assays, 2500 cells weredded in place of receptor-coated beads in the preim-obilized method, and no detergent was present in the

ssay buffer. All other reagents and incubation timesemained the same.

For 864-well assays, all reagents and concentrationsere as described above, but assays were carried outith 400 beads in a final assay volume of 16 ml. The64-well plates were custom-manufactured by Polyfil-ronics (Rockland, MA) with clear bottom and blackidewalls. Well size was 2.5 3 2.5 3 6 mm.Instrumentation. The microvolume fluorimetry in-

trument (Biometric Imaging, Inc.) was a modified cap-llary fluorimeter (1, 2) which is designed to detectuorescence from Cy5 and Cy5.5. In the modified in-trument, the focal point of the bottom of each well isetermined by reflectance from the plate bottom. Ae/Ne laser beam illuminates the scan area at 633 nm,nd imaging of the well is achieved by moving the stagedistance of 1 mm in 250 steps while the laser raster-

cans, by a galvo-scanning mirror, in the perpendicularirection a distance of 1 mm in 250 steps. Thus the-mm2 scan area is imaged in an array of 62,500 pixelsf 4 mm2 each. Scan time for a 96-well plate is about 4in, and about 28 min for an entire 864-well plate.mitted light is collected in two photomultiplier tubesith bandpass filters for Cy5 (645–680 nm, PMT0) or

or Cy5.5 (680–730 nm, PMT1), and digitized. For thexperiments reported here, default settings were usedor laser power and PMTs. Assays were set up to con-ain 50 to 100 analyzable events (beads or cells) in eachcan. Each well resulted in a 245K interleaved datale, which was split into two image files, one for eachMT, using a Java application (P. Egli, Affymax Re-earch Institute). These two files were imported intohe IPLab image processing program (Scanalytics, Inc.,airfax, VA). In IPLab, images were segmented to

dentify events which conformed to the characteristicsf signal intensity, roundness, and area which wereefined for each experiment. Within each well, everyvent was assigned a series of values that corre-ponded to these characteristics (i.e., sum and mean ofixel counts, area, perimeter, and standard deviationf the radius as a measure of roundness), and theseata were averaged for all events in a well. Using thisoftware, irregular events which might be caused byebris can be eliminated, and events of different diam-ters can be analyzed independently. Mean values for

ach well were extracted and exported to Excel (Mi- p

rosoft, Redmond, WA) for further data manipulationsnd IC50 calculations.

ESULTS

Detection of binding of fluorescent ligand to immobi-ized receptor. Initial studies were carried out with0-mm diameter beads. Beads were coated with a mul-ilayer matrix resulting in a high-density array of ac-ive receptor, as described under Materials and Meth-ds. This process is a minor modification of a procedurehich has been characterized extensively and used forvariety of different assay formats (3). These IL-1R-

oated beads were incubated in wells of a 96-well plateith Cy5-conjugated MBP–IL-1RA fusion protein in

he absence or presence of excess unlabeled MBP–IL-RA and read in the MVF. Beads were brightly stainedy the Cy5-MBP–IL-1RA reagent; signals were dimin-shed to undetectable in the presence of a high concen-ration of unlabeled ligand (Figs. 1A and 1B). Cy5.5-abeled IL-5 did not stain the IL-1R-coated beads, andxcess IL-5 did not reduce staining of IL-1R-coatedeads by Cy5-MBP–IL-1RA (Figs. 1C and 1D). A sep-rate population of beads was coated with the IL-5eceptor, which bears the same MAb179 epitope at itsarboxy terminus (3). These beads were brightlytained when incubated with a Cy5.5-labeled prepara-ion of IL-5, and excess unlabeled IL-5 blocked thatignal (Figs. 1E and 1F). IL-5R beads did not bind they5-MBP–IL-1RA, and Cy5.5–IL-5 staining of the IL-R-coated beads was not changed in the presence ofnlabeled IL-1RA (Figs. 1G and 1H). Thus the bindingf fluorescently labeled MBP–IL-1RA and IL-5 to beadsas saturable and receptor dependent.Experiments were carried out to identify the concen-

ration of soluble receptor used to coat the beads andhe concentration of fluorescent tracer that resulted inn optimal signal-to-background ratio. The conditionshich gave the largest signal-to-noise ratios with the

mallest amount of receptor and fluorescent ligandere used for later experiments. Because absolute con-

entrations of receptor and ligand were not deter-ined, this empirical method was used to optimize

ssay conditions for each receptor–ligand pair. Exper-ments in which the unpurified PI-PLC harvest of theeceptor extracellular domain was used to coat theeads, instead of purified ECD, showed that equivalentignals could be generated without interference fromebris or other proteins present in the harvest (dataot shown).As an alternative to the Cy5-conjugated MBP–IL-

RA fusion protein, we tested whether an IL-1R-bind-ng peptide could be directly conjugated with Cy5 andsed as the tracer molecule. High-affinity peptide li-ands had been identified for the IL-1 receptor using

hage display libraries (4). One peptide in this family,
Page 4: A Generic Particle-Based Nonradioactive Homogeneous Multiplex Method for High-Throughput Screening Using Microvolume Fluorimetry

FIG

.1.

Flu

ores

cen

tli

gan

dsbi

nd

spec

ifica

lly

toth

eir

cogn

ate

rece

ptor

sim

mob

iliz

edon

10-m

mbe

ads.

(A–D

)B

eads

coat

edw

ith

IL-1

RE

CD

are

brig

htl

yst

ain

edw

hen

incu

bate

dw

ith

Cy5

-MB

P–I

L-1

RA

(A)

but

not

wit

hC

y5.5

–IL

-5(C

);si

gnal

sar

ein

hib

ited

byad

diti

onof

MB

P-I

L-1

RA

(B)

but

not

IL-5

(D).

(E–H

)B

eads

coat

edw

ith

IL-5

RE

CD

are

brig

htl

yst

ain

edw

hen

incu

bate

dw

ith

Cy5

.5–I

L-5

(E)

but

not

wit

hC

y5-M

BP

–IL

-1R

A(G

);si

gnal

sar

ein

hib

ited

byad

diti

onof

IL-5

(F)

but

not

MB

P–I

L-1

RA

(H).

Alt

hou

ghco

mpe

ted

(dar

k)be

ads

are

diffi

cult

todi

scer

nin

the

imag

es,t

her

ear

ebe

twee

n80

and

120

bead

sin

each

fiel

d.

23

Page 5: A Generic Particle-Based Nonradioactive Homogeneous Multiplex Method for High-Throughput Screening Using Microvolume Fluorimetry

AtbwNAtrcCaste

pifMBEwIfmgctvbi

utcn

iebobbtiaaMstlmT1wtbbIbtw

MA

A

A

A

A

a

b

d alig

24 MARTENS ET AL.

F11733 (Table 1), which had high binding affinity forhe receptor (IC50 of approx 6 nM in a radioligandinding assay [S. Yanofsky, personal communication]),as labeled at the N-terminal phenylalanine using theHS-ester of Cy5 and purified by HPLC. This Cy5–F11733 conjugate brightly stained beads coated with

he IL-1 receptor but not beads coated with the IL-5eceptor. The IC50 determined for MBP–IL-1RA in aompetition binding assay was the same whether they5–peptide ligand or the Cy5-MBP–IL-1-RA was useds the tracer molecule (data not shown). Because theynthetic peptide was relatively inexpensive and easyo conjugate and purify, it was used for subsequentxperiments.IC50s of several peptides of range of affinities. A

anel of closely related IL-1-binding peptides with var-ous affinities (4) was used to assess the MVF assayormat. Competition assays were carried out in the

VF as follows: 10-mm beads were coated with bn-SA, streptavidin, and bn-MAb179 as described above.ach of five IL-1-binding peptides, plus MBP–IL-1RA,as serially diluted in triplicate assay wells; purified

L-1R ECD at 1.25 nM was added to each well, and wasollowed by Cy5–AF11733 at a concentration deter-ined empirically to result in high signal-to-back-

round ratios. After a 30-min incubation, antibody-oated, washed beads were added and allowed to bindo the IL-1R. Assay wells were read on the MVF. IC50

alues agreed well with those obtained by radioligandinding assays (Fig. 2). Fluorescence intensity of beads

TAB

Comparison of IC50 Values Determined Using Fo

Compound structurea RLBb

Preimmobilized receptorc

Expt 1 Expt 2 Expt 3 Mean

BP-IL1RA 1.0 1.4 1.9 0.8 1.F11733FEWPTGYYQJYALPL 6.2 3.8 1.9 1.1 2.F11710FEWTPGYWQPYALP-LDS-NH2 17 14 7.5 6.1 9.

F11864Ac-[1-Nal]-pYQJYALPL 110 230 230 160 210F12409ptK-bzS-pY-QJYALPL 460 2000 2800 1700 2200F11929DNTAWYERFLLQ-NH2 450 2700 2000 2100 2200

Note. Numbers indicate IC50 values in nanomolar, determined anssay conditions (see text). RLB, radioligand binding assay IC50 (nM

a Abbreviations: Ac-, acetylated amino terminus; 1-Nal, 1-naphthyenzylserine; -NH2, amidated carboxy terminus.

b Radioligand binding assay in 24-well (compounds under 10 nM)c Receptor-coated, washed beads added simultaneously with ligand Epitope-specific antibody-coated beads added after incubation of

n the wells containing the highest concentrations of c

nlabeled competitor were equal to or even lower thanhe surrounding fluid containing the unbound fluores-ent ligand, thus demonstrating the very low level ofonspecific ligand binding to the coated beads.Comparison of assay formats. The kinetics of bind-

ng may differ depending on whether the receptor mol-cule is immobilized at high density on a bead surfaceefore exposure to the ligand, in which case bindingccurs on a solid phase, or free, monovalent receptorinds to ligand in solution and is later captured on aead. We investigated different formats for assays inhe MVF using the IL-1 system. In the postincubationmmobilization format, receptor, Cy5-labeled ligand,nd competitor compounds were mixed together andllowed to interact for 2 h at room temperature.Ab179-coated beads (prepared as the bn-BSA/

treptavidin/bn-MAb179 matrix) were then added tohe mixture, and images were obtained. Thus receptor–igand binding occurred in solution phase, with the

ixture sampled later by the antibody-coated beads.he mean total (uncompeted) signal intensity was5,892 6 295 counts per event when 50 to 70 eventsere analyzed in each of three wells in this format. In

he preimmobilized receptor format, receptor-coatedeads were prepared using the bn-BSA/streptavidin/n-MAb179 sandwich followed by incubation with theL-1R ECD. After binding of the receptor to the anti-ody, beads were washed to remove any excess recep-or. These receptor-coated beads were added to assayells containing Cy5-MBP–IL-1RA and dilutions of

1

at Variations of Microvolume Fluorimetry Assay

Postincubation immobilizationd Cells

Expt 1 Expt 2 Expt 3 Mean Expt 1 Expt 2 Expt 3 Mean

0.9 1.0 0.7 0.8 0.5 0.7 0.9 0.7

2.4 2.6 2.3 2.4 0.1 0.2 0.4 0.2

18 9.7 12 13 2.6 1.3 16 6.7

120 68 120 100 290 240 580 370

2300 3000 2900 2700 1500 1700 1200 1500

2300 1600 1700 1900 2300 2900 4600 3300

hown for three replicate experiments using each of three different

anine; pY, phosphotyrosine; J, azetidine; ptK, phthaloyllysine; bzS,

in 96-well plates (compounds over 10 nM) (4).nd tracer.and, tracer, and receptor.

LE

rm

4

2

1

d s).lal

or

ompetitor peptides, and images were acquired after

Page 6: A Generic Particle-Based Nonradioactive Homogeneous Multiplex Method for High-Throughput Screening Using Microvolume Fluorimetry

2ictuewp1wsI1oabdt

ScdrwMfasTi

tiga

Fc

25PARTICLE-BASED HOMOGENEOUS BINDING ASSAY

h incubation. This format is similar to a solid-phasemmunoassay in which binding occurs on a surfaceoated with high-density receptor. Mean total signal inhis format was 8411 6 275 counts per event for thencompeted wells. In a third assay format, CHO cellsxpressing the IL-1R ECD in the PI-glycan-linked formere used instead of receptor-coated beads. Cells (2500er well) were added to wells containing Cy5-MBP–IL-RA and dilutions of competitor peptides, and imagesere acquired after 2 h incubation. Cells gave a mean

ignal of 25,781 6 857 counts per event. Table 1 showsC50 values determined for five peptides and MBP–IL-RA in each of the three assay formats. The rank orderf IC50 was the same in all three assay formats; thebsolute values varied only slightly. The two bead-ased assays showed excellent reproducibility bothay-to-day (Table 1) and within an assay, as judged by

IG. 2. Competition binding curves for IL-1R-binding peptides meounts. Peptide structures are shown in Table 1.

he standard errors determined for the triplicate wells. d

omewhat more variability of IC50s was observed in theell-based assay format, presumably because of minoraily variations in the condition of the cells. However,esults obtained from cell-based assays agreed veryell with the results of radioligand binding assays andVF assays performed with beads. In all three assay

ormats, the peptides which were found to be of lowffinity in the radioligand binding assay were mea-ured as having an even higher IC50 in the MVF assay.hus all three assay formats can be used in the MVF

nstrument with comparable results.

Development of multiplex binding assay. Becausehe MVF instrument detects an image of each particlen an assay well, the assay can be adapted to distin-uish beads of several different sizes in a multiplexssay. To test the instrument’s capabilities, beads with

red in the MVF assay. Data on y axis are total, uncorrected pixel

asu

iameters of 25, 45, and 75 mm were coated with IL-1R

Page 7: A Generic Particle-Based Nonradioactive Homogeneous Multiplex Method for High-Throughput Screening Using Microvolume Fluorimetry

aCwfmtw1atfdm

asM

obeari1wrcaCaitb

F AFe t th

26 MARTENS ET AL.

s described for 10-mm beads, and were incubated withy5-AF11733. For each of these beads, specific signalsere obtained with the Cy5–ligand and were competed

ully with unlabeled MBP–IL-1RA (Fig. 3). The instru-ent was readily able to distinguish stained beads of

hese four different sizes. Competition binding assaysere carried out using IL-1R immobilized on beads of0, 25, 45, and 75 mm, using Cy5–AF11733 as tracernd unlabeled AF11733 as competitor. For each beadype, the IC50s of MBP–IL-1RA and of AF11733 wereound to be approximately 1 nM. These experimentsemonstrated that the apparent binding affinity deter-ined in the MVF assay was independent of bead size.A further increase in assay throughput can be

chieved by testing each compound or sample pool oneveral targets in a single assay mixture. Because the

IG. 3. Detection of IL-1R-coated beads of various sizes with Cy5–ach panel. There are between 80 and 120 beads in each field excep

VF can distinguish events based on fluorescence and a

n size, a multiplex assay may be designed by immo-ilizing different receptors on beads of different diam-ters. Theoretically, the assay can be carried out bydding several immobilized receptors with their fluo-escent ligands to wells containing test compounds. Wemmobilized IL-1R on 78.1-mm beads and IL-5R on0-mm beads using the bnBSA/SA/bnMAb179 sand-ich procedure described above. Beads were washed to

emove any unbound receptor or MAb179. Aliquotsontaining about 2500 beads of each preparation weredded to wells containing an appropriate dilution ofy-5–AF11733 and Cy5.5–IL-5. To one set of wells wasdded buffer only; another set of wells received a block-ng concentration (1 mM) of unlabeled MBP–IL-1RA; ahird received 1 mM IL-5; and a fourth received bothlocking reagents. After incubation, these wells were

11733, a synthetic IL-1-binding peptide. Bead sizes are indicated ine 75 mm beads, where there are about 10 to 15.

nalyzed on the MVF. Beads of 10 and 78.1-mm were

Page 8: A Generic Particle-Based Nonradioactive Homogeneous Multiplex Method for High-Throughput Screening Using Microvolume Fluorimetry

rwl7wwAgb

aIr

Irwtasatbtob

F( 5–I1 ad

27PARTICLE-BASED HOMOGENEOUS BINDING ASSAY

eadily detectable and distinguishable when stainedith both Cy-conjugated ligands (Fig. 4A). Adding un-

abeled MBP–IL-1RA blocked the staining of the8.1-mm beads without affecting the 10-mm beads,hile unlabeled IL-5 blocked the 10-mm bead signalsithout affecting the 78.1-mm beads (Figs. 4B and 4C).ll staining was abolished when both unlabeled li-ands were added (Fig. 4D). Thus two independentinding assays can be carried out in a single well.To characterize the multiplex assay in greater detail,competition assay was carried out to measure the

C s of AF11733 and IL-5 in the presence of both

IG. 4. Multiplex binding assay measured in the MVF. IL-1R was iA) Well contained both bead preparations, Cy5–AF11733, and Cy5.

mM unlabeled AF11733 added. (D) As in A, with both competitors

50

eceptors and both fluorescent ligands. IL-1R and c

L-5R were immobilized on 78.1- and 10-mm beads,espectively, as described above. Four sets of assayells were prepared; all contained both bead prepara-

ions, Cy5–AF11733, and Cy5–IL-5. To one set of wellsdilution series of unlabeled IL-5 was added; another

et of wells received dilutions of unlabeled AF11733;nd a third set received both competitors. Addition ofhe IL-5 competitor reduced signals on the 10-mmeads in a dose-dependent manner, without affectinghe signals of the 78.1-mm beads. Conversely, additionf the IL-1 competitor inhibited signals of the 78.1-mmeads without affecting the 10-mm beads. When both

obilized on 78.1-mm beads; IL-5R was immobilized on 10-mm beads.L5. (B) As in A, with 1 mM unlabeled IL-5 added. (C) As in A, withded.

mm

ompetitors were present, signals were reduced on

Page 9: A Generic Particle-Based Nonradioactive Homogeneous Multiplex Method for High-Throughput Screening Using Microvolume Fluorimetry

b(pI1ccmt

Tobeigao

hwwestcuoaqstIrwm

F1fc 78.u ).

28 MARTENS ET AL.

oth bead populations. Competition curves of the dataFig. 5) showed that each assay behaved inde-endently; the same IC50 values were calculated for theL-5 competitor regardless of the presence of the IL-R-coated beads and Cy5–AF11733, and for the IL-1ompetitor regardless of the presence of the IL-5 assayomponents. Thus this homogeneous binding assayay be carried out in a multiplex format with at least

wo different assays being carried out in the same well.Adaptation of the MVF assay to miniaturized wells.

he MVF instrument reads the fluorescence intensityf individual particles to determine extent of ligandinding, rather than the average signal intensity of anntire assay well. The area of the well surface which isnterrogated by the current instrument is 1 mm2 re-ardless of the size or volume of the well. Thus thessay may be scaled directly to smaller volumes with-

IG. 5. Competition binding curves generated in multiplex assay0-mm beads, Cy5–AF11733, and Cy5.5-IL-5. (A, B) Addition of IL-5rom 10-mm beads (IL-5R, ■). (C, D) addition of AF11733 inhibitsompetitors decrease signals on both bead populations (■, 10-mm; F,ncompeted wells (100%) and wells containing 1 mM competitor (0%

ut altering the concentration of each component. We i

ave carried out competition binding assays in 864-ell plates which were manufactured with black side-alls and an optical-quality clear bottom. For thesexperiments, the IL-1 assay system was used, with theame panel of IL-1-binding peptides used as wereested in the 96-well format described above. All assayomponents and concentrations were identical to thosesed in the 96-well assay. Volumes were scaled to 0.16f the amounts used in the 96-well plate for a finalssay volume of 16 ml. Incubation times and data ac-uisition and analysis were as described above. Table 2hows a comparison of the IC50 values determined inhe 96- and 864-well formats. The results show that theC50s retain the same rank order as in both MVF andadioligand binding assays conducted in 96-well plates,ith only small differences in the absolute IC50s deter-ined. With additional optimization of bead dispens-

at. All wells contained IL-1R-coated 78.1-mm beads, IL-5R-coateds not affect signals of 78.1-mm beads (IL-1R, F) but inhibits signals

nals from 78.1-mm (F) but not from 10-mm beads (■). (E, F) Both1-mm beads). Data on y axis are percentage binding, calculated from

formdoesig

ng, the MVF assay can be carried out in volumes as

Page 10: A Generic Particle-Based Nonradioactive Homogeneous Multiplex Method for High-Throughput Screening Using Microvolume Fluorimetry

lr

D

nwnBptptanltmaonnslCtspabcqr9sserp

flpmdsltathshouocuo

ctbprmcbStaItbtRpf

MAAAAA

p wea

29PARTICLE-BASED HOMOGENEOUS BINDING ASSAY

ow as 5 ml in this 864-well format, further conservingeagents and test compounds.

ISCUSSION

There is a crucial need for nonradioactive homoge-eous high-throughput screening assays to be usedith the tremendous expansion of chemical entitiesow being generated through combinatorial chemistry.inding assays have been used for many years withrogressively smaller assay volumes, increased sensi-ivity, and reduced manipulations, but further im-rovements in these assays are needed to exploit fullyhe new sources of potential drug molecules. Bindingssays have proven to be a productive way to identifyovel receptor ligands, and the characteristics of radio-

igand binding assays are well-understood. As high-hroughput screens move toward higher-density for-ats, the difficulties of performing radioligand binding

ssays become more prominent. The microvolume flu-rimetry assay described here offers an excellent alter-ative for screening. In this nonradioactive, homoge-eous assay format, receptor-coated particles aretained when fluorescent ligand binds. Ligands can beabeled with readily available fluorescent tags, such asy5 and Cy5.5, using well-understood chemistries. The

ags are relatively inexpensive and easy to obtain, soeveral different conditions can be tested during therocess development required to produce a biologicallyctive labeled ligand. Because the assay is particle-ased, miniaturization is straightforward, with nohange in the concentration of assay components re-uired. Individual beads can be considered to be sepa-ate events, and once an assay has been set up in a6-well plate, it can be reduced in volume simply bycaling down the volumes of each component. Eitheroluble receptor preparations on beads or whole cellsxpressing the receptor of interest can be used as theeceptor-coated particle. No unusual assay buffer com-

TAB

Peptide IC50 Values Determin

Compound

864-well plate a

Expt 1 Expt 2

BP-IL-1RA 0.4 0.5F11733 0.19 0.44F11710 1.9 1.5F11864 130 110F11929 540 530F12409 510 900

Note. Peptide IC50 values are reported in nanomolar. In three repliceptide concentration. Assay volume was 16 ml. The same area of thessay.

onents are required. a

Similar particle-based assays can be set up usingow cytometry to detect binding events (8, 9). Multi-lexing is possible using the flow cytometer’s ability toeasure particle sizes and to detect combinations of

ifferent fluorophores in a given particle. In compari-on to the MVF, however, flow cytometers require fararger sample volumes, analysis times are much longerhan the MVF’s ;2 s per well, and sample handlingnd data analysis are more cumbersome compared tohe plate-reading format of the MVF. Flow cytometeristograms depicting fluorescence intensity of gated,tained cells generally show a 5- to 10-fold (or evenigher) variability in signal within a clonal populationf cells; because of this, several thousand events aresually measured for good statistics. In the MVF, webserved much smaller standard deviations in pixelounts, 15 to 25% over the whole population of individ-al beads in a given well, with well-to-well differencesf only 2 to 3%.The scintillation proximity assay (SPA) has become a

ommon homogeneous binding assay format for high-hroughput screening of pharmaceutical compounds,ut miniaturizing to and beyond the 384-well formatoses significant challenges for each new assay. Theelative concentrations of receptor and radioligandust be kept as low as possible to avoid high nonspe-

ific binding and to maintain good assay sensitivity,ut low concentrations lead to low specific binding.cintillation counting times of SPA signals in minia-urized assays become prohibitively time-consuming iflarge number of wells are read one or a few at a time.

n some cases, specialized ultrasensitive imaging sys-ems can be used to detect light output from SPAeads, but there are limits imposed by the intensity ofhe signal generated and background luminescence.egardless of the assay volume, radioactive ligandsose significant problems for large-scale screening ef-orts in both robotic dispensing and waste disposal. In

2

in an 864-Well MVF Assay

ys

96-well RLBxpt 3 Mean

0.35 0.42 1.2 1.00.45 0.36 1.2 1.91.7 1.7 2.2 17

94 111 210 11080 483 2200 45050 687 2200 460

experiments in 864-well plates, nine assay wells were tested for eachll was read on 864-well and 96-well plates. RLB, radioligand binding

LE

ed

ssa

E

36

ate

ddition, expensive commercial or custom-labeled trac-

Page 11: A Generic Particle-Based Nonradioactive Homogeneous Multiplex Method for High-Throughput Screening Using Microvolume Fluorimetry

emrlat

tbdubwaeamlSecnsartccwlricpco

gtbcesarcbtsaamtbmr

tcbnlstIbIbtumiioh

asc9Twm1

atmmbarbstbiwSlsaeeFitcmwa

30 MARTENS ET AL.

rs must be used. Finally, it is not possible to runultiple assays in a single well using the SPA or other

adioligand binding assay formats. These limitationsed us to investigate other ways to carry out bindingssays which could be conducted in multiplex, minia-urized modes.

Fluorescence energy transfer employing long-life-ime fluorophores (10, 11) can be used to carry out someinding assays, and has the advantage of being nonra-ioactive and homogeneous. This technique is oftensed for determining intermolecular distances in ainding event because the energy transfer occursithin a well-defined distance, and has been adapteds a screening protocol for some targets (10–13). How-ver, for high-throughput screening, energy transferssays have several drawbacks. First, the fixed inter-olecular distance required for energy transfer may

imit use of this method for some types of interactions.econd, each assay requires development of two differ-nt labeled assay components. The labeling reagentsan be expensive and not widely available. While ge-eric reagents such as streptavidin may be used inome cases, many assays require custom labeled re-gents. The high cost of europium or other lanthanideeagents can limit the amount of work that can be doneo optimize labeling and assay conditions. In someases, the physicochemical nature of the lanthanideonjugates require the presence of unusual buffers,hich may be a poor environment for some receptor–

igand pairs. Time-resolved fluorescence detectors cur-ently read 96- and 384-well plates; further miniatur-zation will require major modification of existingommercial instruments. Finally, miniaturized multi-lex assays using this technique will require signifi-ant time and investment in development of novel flu-rophores as well as advanced instrumentation.Three different MVF assay designs were investi-

ated in this study. Allowing the receptor–ligand in-eraction to occur before adding the antibody-coatedeads to the mixture allows free association and disso-iation to occur in a truly homogeneous, monovalentnvironment in solution. The receptor population isampled randomly by the MAb179-coated beads addedfter binding has come to, or near, equilibrium. Wheneceptor is immobilized on beads first, there is no re-eptor left in solution which is not immobilized to theead. The assay kinetics may be somewhat differenthan in assays in which receptor and ligand interact inolution before immobilizing on beads. For multiplexssays, it is necessary to add the precoated beads to thessay, and the design of high-throughput screens isade easier when all assay components are added at

he same time. While the addition of beads after incu-ation may be theoretically more accurate for deter-ining IC values of competitor compounds, the data

50

eported here demonstrate that all three formats c

ested gave similar results. Cells usually express re-eptors at a density lower than is achieved by theead-coating scheme we used, but sufficiently high sig-als can be obtained by adjusting the concentration of

abeled ligand in the assay. In our hands, cells wereomewhat more variable in size and staining intensityhan were beads, and we observed greater variation inC50 values determined on different days, but cell-ased and bead-based assay results were comparable.n addition, cells were more irregular in shape, andecause more debris was present in the cell prepara-ions, it was in some cases difficult to distinguish gen-ine events from the debris. With these caveats, cellsay be very useful when no source of soluble receptor

s available or if no nonblocking antibody is available tommobilize the receptor on beads, since all three meth-ds gave equivalent results in the assays reportedere.The MVF assay is readily miniaturized because the

rea of the plate which is interrogated remains theame regardless of assay volume. The same reagentoncentrations can be used in a 100-ml volume in a6-well plate or in a 5-ml volume in a 1536-well plate.hus an assay can be optimized in the standard 96-ell plate, with the simplicity of liquid handling andanipulation, and then quickly transferred to 864- or

536-well plates for high-throughput screening.Because the MVF instrument can discriminate

mong different sizes of beads, several different recep-ors can be assayed independently in a single well toinimize demand for assay compound and to maxi-ize throughput. As demonstrated here, assays may

e carried out on at least four different bead sizes inddition to using cells. In the duplex assay shown,eceptors on the two bead sizes bound to ligands la-eled with different fluorophores, so that large andmall beads were separately detected by the two pho-omultiplier detectors. The assay capacity can be dou-led by using beads of two different diameters arendependently coated with two different receptors,ith ligands for each coupled to distinct fluorophores.everal receptor-coated bead preparations and their

igands could be mixed to create an assay in which eachize of bead would represent two targets distinguish-ble by the PMTs. Using the four different bead diam-ters we tested would make it possible to carry outight assays in a single well with the current MVF.uture improvements in instrumentation are likely to

ncrease the number of wavelengths available for exci-ation and detection, which will make it possible toarry out many more assays in a single well. Sensitiveultiplex binding assays in the miniaturized formatill allow efficient and effective use of the smallmounts of compound generated by combinatorial

hemical synthesis technologies.
Page 12: A Generic Particle-Based Nonradioactive Homogeneous Multiplex Method for High-Throughput Screening Using Microvolume Fluorimetry

bm(cuiclshbTaatrbbrnwa

tswfoMsatp

iscpgFeaac

A

R

PapdhrfIp(s

R

1

11

1

111

1

31PARTICLE-BASED HOMOGENEOUS BINDING ASSAY

The use of spatial arrays of immobilized receptors haseen proposed as an alternative means for carrying outiniaturized binding assays. Microarrays of antibodies

14, 15) or other binding molecules, such as avidin (16),an be “printed” onto plastic sheets or into assay platessing inkjet microspotting technology, and ligand bind-

ng can be interrogated with a fluorescence microscope, aharge-coupled device camera, or an imaging instrumentike the MVF. While the technology appears to workuccessfully for immunoassays (14, 16) or nucleic acidybridization assays (17), adaptation to ligand–receptorinding assays is likely to present significant challenges.he manufacturing process for these arrays is complex,nd receptors are more likely than antibodies to losectivity when they are dried onto a surface. Quality con-rol may be more difficult for spatial arrays than foreceptor-coated beads, which can easily be prepared inulk and tested in advance for activity. Mixtures of beadsearing different receptors can be handled as a singleeagent using standard pipetting devices. Incorporating aew receptor into the assay is straightforward, especiallyith the use of epitope tags and high-affinity nonblockingntibody for immobilization.Many other fluorescence-based assays may be adapted

o the MVF format. For example, a fluorescent peptideubstrate for a protease could be immobilized on a bead;hen cleaved, the fluorescent group would be removed

rom the particle, resulting in a signal decrease. This typef assay could be constructed for a variety of enzymes.ost antibody-binding assays, including colorimetric as-

ays that are currently carried out as ELISAs, could bedapted to this homogeneous method. Assays for pro-ein–protein interactions and DNA hybridization are alsoossible with this instrument.The use of microvolume fluorimetry promises signif-

cant improvements in the efficiency of biomolecularcreening of large combinatorial libraries as well asonventional sources of potential therapeutic com-ounds. The many advantages of this technology sug-est that it will be used for a wide variety of assays.urther instrumentation developments are likely toxpand the utility of the MVF for binding and enzymessays. With its unique ability to carry out multiplexssays, it is anticipated that MVF will replace manyurrent high-throughput screening technologies.

CKNOWLEDGMENTS

We thank Emily Tate, Andrea Dobbs, Yanhua Feng, and Mirnaadic for cell culture and preparation of receptor harvests; Dr.

alaniappan Balasubramanian and Cecilia Moszgai for preparationnd characterization of synthetic peptides; Dr. Stephen Yanofsky forroviding unpublished data on IL-1-binding peptides and for helpfuliscussions; Dr. Bruce P. England for purified IL5, IL-5 receptor andelpful discussions; Kevin Yin and Liang Tang for mass spectromet-ic analysis of the Cy5-labeled IL-1-binding peptide; Erik Whitehornor cloning and expression of the PI-glycan-linked human IL-1 andL-5 receptors; Haim Kedar and Tom Pratt for assistance with com-uter software and hardware; and Ed Goldberg and Phil GoodingBiometric Imaging, Inc., Mountain View, CA), for assistance withoftware and instrumentation.

EFERENCES

1. Dietz, L. J., Dubrow, R. S., Manian, B. S., and Sizto, N. L. (1996)Cytometry 23, 177–186.

2. Shulakoff, E. (1998) Gen. Eng. News 18, 18.3. Whitehorn, E., Tate, E., Yanofsky, S., Kochersperger, L. M.,

Davis, A., Mortensen, R. B., Yonkovich, S., Bell, K., Dower,W. J., and Barrett, R. W. (1995) Bio/Technology 13, 1215–1219.

4. Yanofsky, S. D., Baldwin, D. N., Butler, J. H., Holden, F. R.,Jacobs, J. W., Balasubramanian, P., Chinn, J. P., Cwirla, S. E.,Peters-Bhatt, E., Whitehorn, E. A., Tate, E. H., Akeson, A.,Bowlin, T. L., Dower, W. J., and Barrett, R. W. (1996) Proc. Natl.Acad. Sci. USA 93, 7381–7386.

5. Hannum, C., Wilcox, C., Arend, W., Joslin, F., Dripps, D., Heim-dal, P., Armes, L., Sommer, A., Eisenberg, S., and Thompson, R.(1990) Nature 343, 336–340.

6. Eisenberg, S., Evans, R., Arend, W., Verderber, E., Brewer, M.,Hannum, C., and Thompson, R. (1990) Nature 343, 341–346.

7. Needels, M. C., Jones, D. G., Tate, E. H., Heinkel, G. L., Koch-ersperger, L. M., Dower, W. J., Barrett, R. W., and Gallop, M. A.(1993) Proc. Natl. Acad. Sci. USA 90, 10700–10704.

8. Fulton, R., McDade, R., Smith, P., Kienker, L., and Kettman, J.,Jr. (1997) Clin. Chem. 43, 1749–1756.

9. Oliver, K., Kettman, J., and Fulton, R. (1998) Clin. Chem. 44,2057–2060.

0. Soini, E., Hemmila, I., and Dahlen, P. (1990) Ann. Biol. Clin. 48,567–571.

1. Mathis, G. (1995) Clin. Chem. 41, 1391–1397.2. Lovgren, T., Merioo, L., Mitrunen, K., Makinen, M., Makela, M.,

Blomberg, K., Palenius, T., and Pettersson, K. (1996) Clin.Chem. 42, 1196–1201.

3. Stenroos, K., Hurskainen, P., Eriksson, S., Hemmila, I.,Blomberg, K., and Lindqvist, C. (1998) Cytokine 10, 495–499.

4. Ekins, R., and Chu, F. (1992) Ann. Biol. Clin. 50, 337–353.5. Ekins, R. (1998) Clin. Chem. 44, 2015–2030.6. Silzel, J., Cercek, B., Dodson, C., Tsay, T., and Obremski, R.

(1998) Clin. Chem. 44, 2036–2043.7. Eggers, M., Hogan, M., Reich, R., Lamture, J., Ehrlich, D., Hol-

lis, M., Kosicki, B., Powdrill, T., Beattie, K., Smith, S., et al.(1994) Biotechniques 17, 516–525.