Single-bead, Single-molecule,

Single-cell Fluorescence
Technologies for Drug Screening and Target Validation
MARTIN HINTERSTEINER AND MANFRED AUER
Novartis Institutes for BioMedical Research, Innovative Screening Technologies,
Vienna, Austria

According to many current reports, the pharmaceutical business will hit a wall over the next few
years. The generic competition is expected to wipe out a double-digit billion-dollar amount from
top companies’ annual sales between 2007 and 2012 (Wall Street Journal, online, December 6, 2007).
The industry’s science engine has stalled, new blockbusters are lacking, and patent expirations
are a big problem. Also, the U.S. Food and Drug Administration is pulling back on approvals,
requesting larger safety studies. Among the different approaches taken throughout the industry to
improve productivity and to reduce the attrition rate of compounds in the drug discovery process,
an extended application of quantitative biology and biophysical methods is ranked very high.
Fluorescence spectroscopy and imaging represented the main detection technologies for assays
and screening methods in recent years. Today, label-free detection methods, such as isothermal
titration calorimetry, differential scanning calorimetry, tandem mass spectrometry (MSn ), light
scattering, or interferometry, start to provide viable alternative readouts for physicochemical
characterization of leads and hit list triaging. However, the multidimensional nature of fluorescence
along with its high sensitivity and single-molecule resolution remains an unparalleled source of
molecular parameters to extract all different kinds of information on molecules and ligand–protein
complexes in solution. Although fluorescence-based methods are currently applied throughout the
different stages of the industrial drug discovery process, they are usually applied in an unconnected
way. We have developed a fully integrated hit and lead discovery process combining bead-based
synthesis and screening methods with confocal fluorescence microspectroscopy. The primary on-
bead screening process provides fluorescent ligands that after a multistep characterization process
ultimately leads to fully mechanistically characterized cellularly validated binders and inhibitors of
target protein interactions. The unlabeled small-molecular inhibitors represent chemical starting
points in drug discovery and target validation.

Key words: high-throughput screening; affinity selection; chemical biology; bead-based screening;
one-bead-one-compound; confocal spectroscopy; drug discovery

Introduction a set of sequential steps, ranging from target identifi-

Current Challenges in Drug Discovery
and Strategies to Reduce Attrition
Rates of Compounds from Research
to Market Placement
Since the early days of serendipitous drug discover-
ies and soil screening, constant scientific and techno-
logical advances have changed the process by which
new medicines are generated. Today this process is
highly formalized, standardized, and categorized into
Address for correspondence: Univ. Doz. Dr. Manfred Auer, Novartis In-
stitutes for BioMedical Research Vienna, Innovative Screening Technolo-
gies, Brunner Straße 59, A-1230 Vienna, Austria. Voice: +43-1-80166-
257; fax: +43-1-80166-593.
manfred.auer@novartis.com
cation and assay development to clinical testing and
legal approval. Nevertheless, it currently takes approx-
imately 12 years for a company to bring a new drug
to the market in a difficult, long, and costly process.
From the 2500 to approximately 10,000 hits identi-
fied through high-throughput screening (HTS) in the
discovery phase, approximately 250 compounds en-
ter the preclinical phase and approximately 10 com-
pounds enter clinical testing in phase 1 to determine
safety and dosage. Efficacy and side effect evalua-
tion in phase 2 followed by testing for adverse re-
actions occurring after long-term use in phase 3
would then typically leave one drug candidate to be
approved by the U.S. Food and Drug Administra-
tion. The requirements that a successful drug candi-
date must fulfill are high. They include a substantial

Ann. N.Y. Acad. Sci. 1130: 1–11 (2008).

C 2008 New York Academy of Sciences.
doi: 10.1196/annals.1430.055 1
2 Annals of the New York Academy of Sciences
medical need, a clear rationale for application in
disease indications, bioavailability, appropriate phar-
macokinetics and pharmacodynamics, efficacy and
safety, a therapeutic window, a synthetic route that can
be operated on a large scale, a competitive advantage
over current and emerging therapies, biomarkers, and
a clear mechanistic proof of concept, just to name a few.
After the advances in assay development and au-
tomation technologies during the 1990s, more than
200,000 compounds can be tested in so-called HTS
factories per day. With compound archives of large
pharmaceutical companies containing more than
1,000,000 compounds, speed in HTS is not a criterion
in lead discovery anymore. What has not lived up to
the expectations is the gain in productivity of this drug
discovery process and the success rate measured in
terms of new drug approvals. Despite ever-increasing
research and development investments, the new molec-
ular entities, that is, the number of new U.S. Food and
Drug Administration drug approvals per year, has been
decreasing since 1996.1 The high attrition rates dur-
ing later stages of clinical compound testing have also
contributed to the widely lamented productivity crises
in drug discovery.2–5 To bring more and better new
molecular entities in a shorter time and with less in-
vestment to the market, companies focus strongly on
the reduction of attrition rates during later stages of
drug development. The strategy to achieve this goal is
to produce higher-quality lead compounds for promis-
ing targets at the beginning of the process by setting
predefined criteria for lead optimization and by plan-
ning the clinical trials as early as possible.
Within the highest-ranked strategic elements for im-
proving drug discovery productivity are the following:
• Functionalizing the human genome by novel
molecular pathway approaches, by introducing
new in vivo genetic models, and by expanded
genome mining.6
• Increasing the understanding of disease-
underlying mechanisms to choose valid drug tar-
gets as starting points. Translational biology and
translational medicine arose as new scientific dis-
ciplines and are indicative for the integration of
clinical science into the early drug discovery pro-
cess. Proof-of-concept trials and patient stratifi-
cation are equally important.
• Finding ways to expand the chemical universe by
increasing library diversity and improvements in
structural biology and modeling.
• Finally, there is a need for more quantitative bi-
ology in lead discovery. Functional and mech-
anistic characterization of hit compound–target
protein interactions by biophysical methods is
an active field of research. Currently there is
much hype throughout the pharmaceutical and
biotechnology industry about using artifact-free
detection technologies. It is generally assumed
that label-free detection is less influenced by mea-
suring artifacts than fluorescence detection.
Changing Paradigm in Detection
Technologies for Drug Discovery?
Until about 1990, radioactivity was the detection
technology of choice for assays and early screening
methods; today it is almost only used for scintillation
proximity assays in HTS. In the period between 1990
and 2005 HTS was strongly dominated by fluores-
cence spectroscopy almost exclusively used as detection
method. Later fluorescence imaging for high-content
screening became increasingly important. Label-free
detection really started to kick in around 2005 and
beyond with calorimetric methods, such as differential
scanning calorimetry (DSC) and isothermal titration
calorimetry (ITC); biosensor techniques on plates and
chips; and, most importantly, mass spectrometry (MS)
applied to screening, imaging, and profiling. Although
the label-free methods hold great promise for future
applications in applied science, a current comparison
of advantages and disadvantages of fluorescence detec-
tion versus label-free detection methods reveals a clear
strategy for method selection.
The major benefit of any label-free detection tech-
nique is the lack of a sensor or tracer linked to the
receptor or ligand. The usual concern is that a la-
bel might influence the molecular recognition event.
With appropriate care taken by the scientist to per-
form all necessary controls and by keeping in mind
that the fluorescently labeled ligand is nothing more
than a tool for competition screening, most fluores-
cence assay techniques will deliver reliable results. MS
detection is currently successfully applied for complex
enzymatic screens, such as lipid metabolic assays with
mass changes involved. Such assays are sometimes dif-
ficult to run with fluorescence detection. ITC, DSC,
and light-scattering methods start to substantially affect
lead optimization and triaging of hit lists. Only multi-
dimensional fluorescence fluctuation analysis at single-
molecule resolution applied in competition studies with
a labeled surrogate ligand and a series of inhibitors can
provide better results. Currently, fluorescence meth-
ods are superior to label-free detection techniques be-
cause of their high sensitivity down to single-molecule
and single-cell resolution. Together with the possibil-
ity to measure ligand binding in homogeneous mix-
tures at equilibrium, fluorescence techniques are su-
perior in delivering quantitative thermodynamic and
kinetic data (dissociation constants, rate constants of
Hintersteiner & Auer: Single-bead, Single-molecule, Single-cell Fluorescence
individual reaction steps necessary to resolve molecu-
lar mechanisms). The picosecond to microsecond time
scale of fluorescence detection corresponds to the time
scale of dynamic events in molecules. Fluorescence de-
tection techniques are therefore most suitable to resolve
conformational and dynamic information of molecules
and their complexes in solution. The current level of
resolution of label-free detection remains much lower,
at least for the time being. Successful MS detection
of noncovalent ligand–protein complexes in homo-
geneous mixtures strongly depends on the nature of
the protein, compound affinities, and buffer conditions
and has not yet been developed to a universal method
in lead discovery. Biosensor chip–based systems, on the
other hand, are heterogeneous and suffer from a lower
time resolution in kinetic experiments than transient-
state kinetics. The current trend in screening and pro-
filing groups throughout the industry is to increase the
palette of biophysical methods and to include them at
appropriate time points of the lead discovery process
into their flowcharts and decision trees.
The Playground for Fluorescence Detection
throughout the Discovery Process
The essential advantage of fluorescence detection
techniques compared with other biophysical methods
is their breadth of application throughout the individ-
ual process elements of the discovery phase. Fluores-
cence methods currently affect basic research, target
identification, assay development, HTS, hit and lead
profiling, and lead optimization. However, the differ-
ent kinds of fluorescence techniques are applied in an
unconnected way addressing specific individual project
needs. These individual fluorescence tool sets often do
not address the uncertainties in the identification of
chemical ligands by automated HTS or the difficul-
ties in mechanistic target validation on a cellular level.
For a thorough assessment of the drug target poten-
tial of a specific protein, it is necessary to modulate
its function within the cellular context in a precisely
timed manner. That is, for target validation one would
ideally like to have a small-molecular inhibitor in the
first place. Small-molecular ligands provide precious
tools to perform a thorough investigation on the ef-
fects of a chemically induced conditional knockdown
of a protein under investigation. Therefore, the need
to generate an arsenal of small-molecular modulators
of protein function has contributed to the development
of chemical biology as a field of research at the cross-
roads of chemistry and biology.7 Consequently, there
is a need for alternative in vitro ligand identification
tracks, which can be followed in parallel to the stan-
dard drug discovery process. The main purpose of such
3
an alternative ligand identification platform would be
to deliver chemical starting points for drug discovery, to
present ligands for chemical biology, and to accelerate
target validation without undergoing full-blown, costly
screening campaigns. It is this prospect of identifying
ligands to bridge the gap between basic and clinical
research that has made screening and ligand identifi-
cation an actively pursued field of research also within
the academic world.8–10
Ideally, such alternative chemical biology screen-
ing platforms should be characterized by the following
advantages:
• High resource efficiency to allow simple and cost-
effective testing of virtually any target protein.
• Fast discovery cycle times to allow rapid assess-
ment of potential candidate proteins for their
drug target potential (i.e., druggable? disease rel-
evant? toxic?).
• The method should not depend on large com-
pound archives accessible only to big pharma
companies.
• The method should be flexible with regard to the
chemical input to allow target-guided selection
of chemical space.
By precisely addressing the issues and needs dis-
cussed above, we developed an integrated bead-based
synthesis and screening process for lead identification
and for mechanistic and quantitative target validation
by chemical biophysics (FIG. 1). Our process integrates
a diverse set of specific fluorescence techniques that
provide both qualitative and quantitative information
in the different areas of lead discovery. Our key Inte-
grated Chemical Biophysics (ICB) process steps incor-
porate the following components:
• high-resolution confocal scanning and imaging,
resolving multiple fluorescence parameters, such
as intensity, polarization, and lifetime;
• confocal fluorescence fluctuation analysis
at single-molecule resolution (usually called
“single-molecule spectroscopy”) for multidimen-
sional hit confirmations with less than 50 pmol
of primary hit compounds; and
• a rich tool set of analysis methods working at the
single-molecule and single-cell level for mech-
anistic target validation, including fluorescence
lifetime imaging (FLIM), Förster resonance en-
ergy transfer (FRET), anisotropy, fluorescence
correlation spectroscopy (FCS), fluorescence
cross-correlation spectroscopy, fluorescence in-
tensity distribution analysis (FIDA), and burst in-
tegrated fluorescence lifetime (BIFL).
4 Annals of the New York Academy of Sciences

FIGURE 1. The ICB process. Today, drug discovery is pursued by following a standardized set of
stages/steps that are carried out by specialized teams. The classical HTS process usually comprises little
more than assay development, the screen itself, and a secondary hit confirmation assay. In contrast,
the ICB process, as described herein, ranges from basic research to hit/lead profiling. The integrating
elements in this process are the use of highly flexible bead-based chemistry and confocal fluorescence
(micro) spectroscopy techniques. (In color in Annals online.)

FIGURE 2. Outline and issues of a conventional OBS process. Starting from a one-bead, one-
compound combinatorial library, the resin beads are incubated with fluorescently labeled target protein.
Binding of target protein to the bead-immobilized compounds produces hit beads which must be isolated.
Hit structures are then obtained by MS analysis of the cleaved compound fractions. After decoding, the
hit structures can be resynthesized in larger quantities and subjected to solution testing. The main issues in
the presented process flow include variable bead autofluorescence, possible deviations between solution
and surface binding thermodynamics, variable compound purity, and the limited amount of compounds
derived from single hit beads. (In color in Annals online.)

Results sis, that is, on-bead screening (OBS), holds great

Advantages and Difficulties of Bead-based
Screening
The concept of screening combinatorial com-
pound libraries directly at the site of their synthe-
potential for improving HTS beyond speed opti-
mization and assay miniaturization (FIG. 2). Solid-
phase combinatorial synthesis techniques can gener-
ate thousands to even millions of compounds in only
a few reaction steps.11 The beauty of this numbers
Hintersteiner & Auer: Single-bead, Single-molecule, Single-cell Fluorescence
game is usually limited only by the inherently low
reaction scales and the need to purify the generated
substances for further solution testing. However, us-
ing the bead as both a screening and synthesis com-
partment for generating one-bead, one-compound li-
braries provides distinct advantages:
•
Screening the compounds directly on bead for
their ability to bind fluorescently labeled target
proteins overcomes the need to purify library
members before carrying out the screening assay.
•
Consequently, the overall library production
time and synthesis costs are dramatically reduced
compared with traditional library synthesis and
purification protocols.
•
Furthermore, using the bead as a screening com-
partment allows performing compound screen-
ing with only 10–90 pmol of substance (the
amount of compound present on one individual
bead). The actual assay volume is thereby more
than 1000-fold reduced (i.e., reduced essentially
to the volume of a 90-µm TentaGel bead) com-
pared with the few-microliter assay volume of a
1536-well microtiter plate used in solution HTS.
Overall, the concept of screening compounds at the
site of their synthesis, namely, on bead, would mean
shifting resources from inactive toward active library
members. Only compounds showing activity in the
initial OBS assay would be followed up. Decoding,
resynthesis, purification, and further testing of hit com-
pounds in more sophisticated assay systems would be
optimally performed only with active compounds.
Unfortunately, the OBS process as outlined above is
associated with several inherent issues: First, different
chemistries put on the beads will lead to different levels
of autofluorescence. As a result, positive hit beads can
be difficult to distinguish from autofluorescent beads.12
Second, the amount of target protein bound to individ-
ual hit beads is not only governed by the affinity of the
ligand to the target protein but also by the compound
purity on bead. Third, and most important, true HTS
of large libraries usually generates hundreds of hits
(similar to screening in solution). For a repetitive, sta-
ble, and reliable process it is impractical to isolate a sig-
nificant number of these on-bead active compounds, to
decode the structures, and to resynthesize hundreds of
compounds with the goal of determining their affinities
to the target protein in solution to distinguish between
true binders and artifacts.
Elements of a Reliable Bead-based
Screening Process
A reliable OBS process satisfying drug discovery
standards must fulfill several criteria:
5
• A screening method that allows detecting spe-
cific target binding, irrespective of the bead back-
ground fluorescence (preferably in an automated
fashion).
• A method for the reliable retrieval of hit beads in
combination with an effective structure decoding
process.
• A method to determine whether the target bind-
ing event detected on bead also results in com-
plex formation between ligand and receptor in
homogeneous solution.
• Ideally, one would further aim at demonstrat-
ing the activity of the identified ligands in a cell
by using a combination of cellular imaging, mi-
crospectroscopy, and functional assays.
Confocal Nanoscanning (CONA) and Bead
Picking
For the primary OBS process, scientific groups in the
field either use a commercially available bead sorter
(complex objects and particle sorter) in combination
with “manual eye inspection and manual bead pick-
ing” on standard fluorescence microscopes or rely com-
pletely on manual procedures. In contrast, we have de-
veloped and implemented automated CONA for OBS.
This unique high-resolution optical method allows the
reliable detection and quantification of binding events
in the outer approximately 1–2 µm of (TentaGel) beads
with superior sensitivity and effectively suppresses the
background intensity arising from the bead matrix
(FIG. 3).
Standard fluorescence imaging methods in OBS
detect the fluorescence signal of the entire bead vol-
ume. However, binding of fluorescently labeled target
proteins is restricted mainly to the surface of a bead,
although smaller proteins and peptides diffuse into the
bead matrix as a function of time and particle size.
Overall, the fluorescence intensity signal relevant for
OBS is mapped to the outer bead layers. In CONA,
a confocal laser focus scans through the beads, slightly
below the equatorial plane. The spatial fluorescence in-
tensity profile of a an approximately 2-µm-thick layer
in the image plane is collected. The confocal scanning
method works by cross sectioning. Beads with bound
target protein on the surface appear as discs with an
intensity-enhanced outer part, a so-called fluorescent
ring on the outer surface.
Measuring the fluorescence intensity with a spa-
tial resolution of approximately 5 µm provides a high
signal-to-noise ratio and allows highly sensitive and
specific detection of target–compound interactions on
bead, despite various background levels. Furthermore,
this method is inherently quantitative, and ranking
of hit beads according to the amount of protein
6 Annals of the New York Academy of Sciences

FIGURE 3. OBS by confocal microscopic nanoscanning. In CONA, a focused laser beam is used to
sequentially scan through beads in the x –y plane. The scanning is conducted slightly below the equatorial
plane of the beads (at ∼30-µm height for standard 90-µm TentaGel beads), forming a monolayer in a
microtiter plate well. This scanning process monitors an optical cross-section of the beads, and the high
spatial resolution due to the small sampling volume allows distinguishing autofluorescence of the bead
interior from target protein binding on the outer surface of the beads. Binding of fluorescently labeled
target protein to the bead surface produces a fluorescent ring in the confocal scan images. The PS02
instrument (bottom left) is based on an inverted Olympus IX70 microscope. The basic microscope unit is
drawn in black and features two detection channels, an avalanche photodetector–based signal detection
and the possibility to use several laser sources. For OBS, the instrument was adapted (Perkin Elmer, former
Evotec Technologies, Hamburg, Germany) by integrating a high-precision scanning table and a picker
stage, which consist of a robot arm and a precision syringe pump system (blue parts). A picker capillary
is mounted onto the picker arm and with this capillary hit beads are picked in a semiautomated fashion.
The picker capillary in action is depicted in the insert on the right (red frame). (In color in Annals online.)

bound per bead yields an objective hit prioritization.
The instrument series used for CONA consists of a
modified Olympus IX70 microscope (Olympus Cor-
poration Tokyo, Japan) with an integrated confocal
detection system for two detection channels. The in-
struments are also equipped with a robotics system for
picking of individual hit beads by virtue of a capillary
that is positioned at the x–y coordinates of a selected
hit bead. The hit bead is then sucked into the cap-
illary through hydraulic underpressure and deposited
into either individual vials or microtiter plates. Us-
ing avalanche photodetectors as detectors, we perform
high-resolution scans for resolving on-bead binding
mechanisms. With charge-coupled device–based de-
tection and a Nipkow spinning disk (Yokogawa Elec-
tric Corporation, Tokyo, Japan) integrated in a sec-
ond type of microscope, a throughput of approximately
6.5 h for a maximum of 180,000 compounds can be
achieved. Therefore, the full instrument capacity is ap-
proximately 360,000 compounds per day.
Confirmation of On-bead Screening
Hits in Homogeneous Solution
Although the possible background fluorescence
from the bead matrix needs to be considered a po-
tential problem in OBS, using the bead as a screening
compartment also provides some advantages over so-
lution screening. Of particular importance is the high
local compound concentration on individual beads.
Even low-affinity binders, up to millimolar dissoci-
ation constants (K d s), may result in a positive on-
bead binding event.13 The range of detectable affini-
ties can be adjusted by modulating the stringency of
the screening conditions.14 Nevertheless, it might eas-
ily happen that a target–compound interaction found
on bead by fluorescent ring formatior cannot be re-
produced with standard fluorescence detection tech-
nologies in homogeneous solution because of too low
affinity and the inability to reach the necessary con-
centration of a protein in the incubation compart-
ment. Hit compounds with affinities to a protein
Hintersteiner & Auer: Single-bead, Single-molecule, Single-cell Fluorescence
target below approximately 100 µM K d will hardly
ever be followed up in a chemistry program. For an
efficient screening process, it is therefore of crucial
importance to have a second affinity determination
method available for solution testing to eliminate un-
desirably low affinity binders or real false positives.
The importance of such secondary confirmation as-
says has recently received much attention.15 Reported
confirmation methods with resynthesized material
comprise Biacore,16,17 ITC,18 photo-crosslinking,19,20
pulldown experiments from cell lysates, and fluores-
cence anisotropy.21 Some of these methods are time-
consuming and/or need a large amount of com-
pound, such as photo-affinity labeling and ITC,
respectively. In lead discovery processes, high-
throughput methods, which allow determining K d s of
hit compounds in solution, are clearly preferable (e.g.,
fluorescence anisotropy titrations). However, modify-
ing hit compounds for use in secondary assays (e.g.,
incorporating tethers and tagging groups) and the
case-by-case choice of a detection method are not
suitable for a screening process with medium to high
throughput.
Tagged libraries represent a generic solution for
this confirmation problem and were used in chemical
genomics and proteomics systems.22–25 Besides facili-
tating biological activity testing in vitro, (fluorescently)
tagged libraries hold great potential for studying a com-
pound’s molecular mode of action in vivo.26 Because of
the inherent advantages of tagged library methods, we
have combined our OBS approach with such a com-
pound tagging method for generic solution and cellular
confirmation.
Labeling Beads Postsynthesis
and Postscreening
Two main types of detection signals are used during
the multistep standard OBS process.
The primary readout in the OBS phase is target
binding to immobilized compounds on beads. Usu-
ally, only after hit bead isolation, cleavage, MS-high-
performance liquid chromatography (HPLC) analysis,
decoding, and resynthesis, the decisive interaction be-
tween target and potential hit compound in solution
is measured with a secondary assay. The first molecu-
lar recognition on the solid surface should ideally help
isolating all beads with active compounds of any affin-
ity to the screening target and exclude most inactive
compounds faster and more effectively than any solu-
tion process. The second molecular recognition event
in solution is relevant for the definition of a hit list. A
full-blown OBS may easily result in several hundred
to thousand hit beads. The selection criterion in the
7
OBS, in our setup the target binding effect to com-
pounds on beads, can at best give a rough estimation
of how strong the interaction in solution might be.
Based on the experience gained with many screens,
the ranking of on-bead binding signals from target
fluorescence intensities can only in a limited way be
used as a definite parameter for the hit prioritization.
With the inherent variability in the amount of com-
pound on bead, chemistry-dependent presentation of
compounds to the solvent, strong electrostatic effects,
and the reverse stoichiometric compound target ratios,
there is always a certain percentage of on-bead bind-
ing events that do not translate into a target–compound
interaction in solution.
By using CONA in combination with state-of-the-
art tandem MS (MSn ) methods, the isolation and de-
coding of many hits is in principle feasible. However,
in a real screening situation, only a limited number of
structures can be resynthesized, ideally the ones with
the best odds for exhibiting solution binding. Hence,
the relevant solution data can be obtained only for
some of the initial hits.
Rationally, to overcome this central dilemma, a di-
rect link between on-bead binding and the secondary
solution binding assay needs to be established. Argue-
ably one of the most advantageous ways to achieve
this link might be to label compounds on individual
hit beads with a fluorescent dye only after the chem-
ical synthesis of the compound on bead and after
the primary screen. We therefore call this new con-
cept postsynthesis/postscreening (PS/PS) labeling. In-
troduction of the dye after synthesis and after screen-
ing is particularly advantageous because it gives a free
choice of labeling reagent, e.g., for color selection. It
therefore overcomes the chemical stability problem in-
herent to most dyes and finally allows using confocal
spectroscopy for solution confirmation.
In the PS/PS concept, each isolated hit beads de-
rived from CONA OBS is individually derivatized with
a fluorophore, using the copper-catalyzed version of
Huisgen’s 1,3-dipolar cycloaddition reaction of azides
and acetylenes, known as click reaction. Compounds
are then cleaved from the individual beads. With
the same suitable, long-wavelength fluorescent dye on
every compound, all available single-molecule spectro-
scopic methods for analyzing target–compound inter-
actions (e.g., two-dimension FIDA (2D-FIDA), FIDA,
FCS, fluorescence intensity multiple distribution anal-
ysis) can immediately be used in the secondary assay.
As a consequence, resynthesis decisions are based on
the relevant solution detection signal, and the K d be-
tween compound and target allows focusing the efforts
on the “best” hit compounds.
8 Annals of the New York Academy of Sciences
Cellular Imaging and Microspectroscopy
Using Fluorescent Ligands
Every hit compound from our bead-based ligand
discovery process comes with a known K d for tar-
get complexation and additional predesigned chemical
tagging possibilities. The next critical step is demon-
strating the biological activity of the identified ligands
in disease-relevant cell types. This goal is achieved with
a combination of functional and biophysical cellular
imaging and microspectroscopy assays. Testing the la-
beled and unlabeled hit compound for target binding
and for inhibitory effects in cells is currently used as
the final stage of the ICB screening process.
For this purpose, PS03—a high-resolution, mul-
tiparameter, multimodality microspectroscope—has
been developed. The innovative combination of (1)
3D high-resolution (confocal) multiparameter fluores-
cence imaging, (2) single-molecule fluorescence spec-
troscopy at selected points or regions of interest, and
(3) a comprehensive set of analysis methods renders
our instrument an ideal tool for detecting and quanti-
fying the activity of screening compounds in a cellular
context. The currently integrated and applied analy-
sis methods of PS03 are FLIM, polarization imaging,
FRET (donor lifetime quenching and sensitized emis-
sion), FCS and fluorescence cross-correlation spec-
troscopy, FIDA, 2D-FIDA anisotropy, and BIFL. These
analysis methods provide, for instance, information
on the binding status, chemical environment, reac-
tion rate constants, molecular distances, weight, shape,
rotation, and aggregation. The subsequently enumer-
ated application examples aim at demonstrating the
current microspectroscopy possibilities of PS03 for cell-
based assays and compound testing. For instance, PS03
was used to measure the 3D localization of transgenic
and endogenous proteins with submicron resolution;
the nuclear translocation of low abundant endoge-
nous protein in primary peripheral blood mononu-
clear cells upon T cell activation; multicolor, multi-
component imaging of fluorescent protein–conjugated
proteins and dye-labeled compounds; ligand–receptor
association kinetics; affinities and competitions; and
intracellular complex formations, as well as to unravel
molecular subpopulations in homogeneous solution.
Newly Established Ligand
Identification Process
The core ICB process consists of seven steps:
1. sample preparation and automated confocal
scanning;
2. data evaluation and bead picking;
3. PS/PS labeling of the individual hit beads and
cleavage from resin;
4. quantitative affinity determination of fluores-
cently labeled hit compounds to target proteins
by 2D-FIDA and/or FCS;
5. MS-based decoding of hit compounds, exhibit-
ing the highest affinity to target proteins in solu-
tion;
6. resynthesis of hit structures and secondary testing
of direct binding by 2D-FIDA; and
7. confirmation of ligand activity in cellular context
by using the fluorescently tagged and untagged
form of the identified ligand for fluorescence mi-
crospectroscopy and in functional assays, respec-
tively (FIG. 4).
To run such an OBS cycle, the wells of a 96-
well microtiter plate are filled with approximately
1 mg of beads, equal to roughly 2000 beads per well.
Specific treatment of the beads, including swelling
in buffer (e.g., phosphate-buffered saline, containing
small quantities of additives, such as Tween 20 or
Pluronic) and sonication (∼0.5 min) to disrupt bead
clusters, results in a monolayer of beads filling the area
of each well. The beads are then incubated with fluo-
rescently labeled target protein overnight and automat-
ically scanned well by well. Typical CONA scan times
are 6–7 h. The optical slicing through the equatorial
plane of spherical beads during the confocal scanning
process results in disc-like distributions of fluorescence
intensity. Pattern recognition software (BeadEval from
Perkin Elmer, former Evotec Technologies, Hamburg,
Germany) recognizes the bead discs and calculates two
characteristic parameters for each bead—the average
fluorescence intensity on the disc surface (the bead
intensity) and the mean intensity on the border area
of the disc (the ring intensity). Hit beads with fluo-
rescently labeled target bound to the bead exhibit an
increased fluorescence ring intensity compared with
the surface intensity. The quantitative image analysis
procedure selects positive beads on the basis of quanti-
tative criteria and determines their x–y coordinates in
the respective well from the image.
The beads with the highest ring versus surface in-
tensities are selected for picking with a semiautomatic
picking device. The picking procedure typically takes
approximately 1 min of actual picking time per bead.
The retrieved hit beads are then subjected to PS/PS la-
beling for generating fluorescently labeled compounds
and cleaved from the solid support. For PS/PS label-
ing, a large excess of an azide-modified fluorophore
is used. The labeling is carried out overnight accord-
ing to standard protocols for click chemistry, and the
beads are excessively washed before final cleavage from
resin. Twenty percent of labeled substance from one
Hintersteiner & Auer: Single-bead, Single-molecule, Single-cell Fluorescence 9

FIGURE 4. Process outline of the ICB ligand identification process. The ICB process starts by generating biased or
unbiased libraries on TentaGel beads (90 µm). The beads are then distributed into the wells of a 96-well plate and incubated
with fluorescently labeled target protein(s). Next, CONA is performed and the beads are quantitatively analyzed to detect
ring- and surface fluorescence intensities. The highest-ranked hit beads are isolated using the PS-instrument and are subjected
to PS/PS labeling for generating fluorescently labeled compounds, which are cleaved from the solid support. The material
from individual hit beads is then analyzed for target binding in solution by fluorescence fluctuation spectroscopy (direct
binding of fluorescently tagged compounds to unlabeled target proteins). Compounds active in solution are subjected to
MS methods and the hit structures are decoded. Finally, the best hit compounds are resynthesized in labeled and unlabeled
form. Currently, the final step of this process is the cellular/mechanistic characterization of PS/PS-labeled and/or unlabeled
binders and inhibitors in functional microspectroscopic assays. (In color in Annals online.)

individual 90-µm TentaGel hit bead is then taken for
analyzing target binding in solution by fluorescence
fluctuation spectroscopy titrations (direct binding of
fluorescence-tagged compounds to unlabeled target
proteins in solution). Compounds active in solution are
then subjected to MS-based decoding methods, such
as matrix-assisted laser desorption ionization MS/MS
or electrospray ionization MS/MS. Finally, the best
hit compounds are resynthesized in labeled and unla-
beled form. The final step of this process is to apply the
PS/PS-labeled and/or unlabeled compounds for cel-
lular/mechanistic characterization and in functional
assays.
Discussion
Hit confirmation in a reliable and quantitative way
is a central problem in OBS, as well as in any screen-
ing process in general. Especially all “heterogeneous”
screening methods, such as chip- or bead-based sys-
tems, need to translate their primary detection signals
into the relevant solution binding information. Start-
ing from the idea of directly linking an on-bead binding
event to its confirmation in homogeneous solution, we
developed a method for derivatizing compounds on
single beads after synthesis and screening. This trans-
formation of compounds into fluorescent ligands serves
as the integrating element for using fluorescence-based
detection technologies throughout all stages of a ligand
discovery and target validation process.
With the PS/PS labeling in place, it is possible to
obtain quantitative solution affinity data for all OBS
hits directly, without resynthesis. Moreover, the 10–
90 pmol of labeled compound from one hit bead are
theoretically sufficient for approximately 500 single-
molecule spectroscopic confirmation assays in 5-µL
volume wells. For a bead-based HTS process, this pos-
sibility to generate tens to hundreds of assay points
from each hit bead, plus an HPLC-based quality con-
trol and a MS decoding link, represents for the first
time a way to remove the central bottleneck of a tech-
nology which was always appealing for chemists and
biologists alike but never achieved a breakthrough in
10
pharmaceutical industry. The link of bead-based pri-
mary screening with quantitative solution testing opens
up new routes for designing and running a resource-
efficient bead-based ligand identification process. The
most significant features of such a new ligand identifi-
cation process include the following:
•
Multiplexing. By screening multiple targets si-
multaneously, one can generate a quantitative
interaction–specificity profile for each hit com-
pound. In contrast to hitherto existing multiplex-
ing approaches, the assignment of the respective
target to a certain hit bead can be accomplished
in the solution confirmation step by testing ev-
ery hit compound against each target in the
screen. Doing so not only increases the obtain-
able throughput but also generates additional
specificity information about the obtained hit
compounds. Such interaction–specificity profiles
are of special interest for considering biased li-
braries and/or related targets or whole target
classes.
•
Multiple confirmation measurements before
resynthesis. With the methods introduced herein,
on average 50 pmol of fluorescent ligand is ob-
tained from primary OBS. This material can be
used to carry out several analyses in parallel.
Microdialysis and size-exclusion HPLC are two
nonspectroscopic biophysical methods, which
have been miniaturized to the single bead level to
expand the available set of secondary confirma-
tion methods for compound profiling in the ICB
process. These two methods, together with the
confocal fluorescence fluctuation analyses such
as 2D-FIDA or FCS, and with HPLC quality
control, rely on fluorescence as the detection
principle. Besides of derivation for fluorescence-
based detection, compounds may just be cleaved
off the hit bead and used to measure molecular
recognition events label free on instruments like
the SRU Reader or SRU Scanner (SRU Biosys-
tems, Woburn, MA).
• Cellular confirmation. The generation of ligands
with “tunable” fluorescence properties broadens
the potential scope of the ICB process. The ulti-
mate goal is to resolve the mechanistic mode of
action for a fluorescently tagged hit compound
in a cell or in a model organism.
• Resource efficiency and speed. Compounds are
decoded only after solution confirmation and
ranking. This approach effectively reduces the
number of beads to be analyzed by HPLC-
MS or matrix-assisted laser desorption ioniza-
Annals of the New York Academy of Sciences
tion MS. Secondary functional and biophys-
ical assay systems based on fluorescence de-
tection imaging and microspectroscopy meth-
ods provide further compound characterization
and lead to an early and stepwise exclusion of
artifacts.
Overall, one of the key advantages of our ICB dis-
covery platform lies in the identification of fluores-
cently labeled ligands in a generically applicable way.
Moreover, taking a case-by-case choice any dye (or any
other tagging group) can potentially be used to label
a given library without deciding beforehand (during
the synthesis phase) on a specific tracer. Thus, the
described process can be used to set up an affinity-
based multimodal screening process. “Multimodal”
here means the simultaneous usage of one chemical li-
brary (input) for at least three different screening strate-
gies: (1) in vitro biochemical screening, especially on the
Evotec Mark-III and equivalent platforms; (2) cellular
affinity and inhibition screening; and (3) affinity-based
OBS.
Taken together, combining bead-based chemistry
with the breadth and sensitivity of fluorescence-based
detection technologies into one screening and valida-
tion technology provides a fully integrated, miniatur-
ized, and quantitative chemical biology process.
Acknowledgments
We thank Rene Amstutz, Roman Bauer, Christof
Buehler, Geraldine Garavel, Marion Kala, Thierry
Kimmerlin, Nicole Meisner, Torsten Schindler, Jan-
Marcus Seifert, Mario Schmied, and Volker Uhl for
their contribution to the ICB process development.
Conflict of Interest
The authors declare no conflicts of interest.
References
1. FRANTZ, S. 2006. Pipeline problems are increasing the urge
to merge. Nat. Rev. Drug Discov. 5: 977–979.
2. PREZIOSI, P. 2004. Opinion: science, pharmacoeconomics
and ethics in drug R&D: a sustainable future scenario?
Nat. Rev. Drug Discov. 3: 521–526.
3. ALANINE, A. et al. 2003. Lead generation—enhancing the
success of drug discovery by investing in the hit to lead
process. Comb. Chem. High Throughput Screen. 6: 51–
66.
4. RATTI, E. & D. TRIST. 2001. The continuing evolution of
the drug discovery process in the pharmaceutical industry.
Farmaco 56: 13–19.
Hintersteiner & Auer: Single-bead, Single-molecule, Single-cell Fluorescence
5. ADANG, A.E.P. & P.H.H. HERMKENS. 2001. The contribu-
tion of combinatorial chemistry to lead generation: an
interim analysis. Curr. Med. Chem. 8: 985–998.
6. FISHMAN, M.C. & J.A. PORTER. 2005. Pharmaceuticals: a
new grammar for drug discovery. Nature 437: 491–493.
7. SCHREIBER, S.L. 2005. Small molecules: the missing link in
the central dogma. Nat. Chem. Biol. 1: 64–66.
8. DUFFNER, J.L., P.A. CLEMONS, & A.N. KOEHLER. 2007.
A pipeline for ligand discovery using small-molecule mi-
croarrays. Curr. Opin. Chem. Biol. 11: 74–82.
9. STEIN, R.L. 2003. High-throughput screening in academia:
the Harvard experience. J. Biomol. Screen 8: 615–619.
10. GORDON, E.J. 2007. Small-molecule screening: it takes a
village. ACS Chem. Biol. 2: 9–16.
11. LAM, K.S., M. LEBL & V. KRCHNAK. 1997. The one-bead-
one-compound combinatorial library method. Chem.
Rev. 97: 411–448.
12. OLIVOS, H.J., K. BACHHAWAT-SIKDER & T. KODADEK. 2003.
Quantum dots as a visual aid for screening bead-bound
combinatorial libraries. Chembiochem 4: 1242–1245.
13. YING, L. et al. 2005. A topologically segregated one-bead-
one-compound combinatorial glycopeptide library for
identification of lectin ligands. J. Comb. Chem. 7: 372–
384.
14. REDDY, M.M., K. BACHHAWAT-SIKDER, & T. KODADEK.
2004. Transformation of low-affinity lead compounds
into high-affinity protein capture agents. Chem. Biol. 11:
1127–1137.
15. KODADEK, T. & K. BACHHAWAT-SIKDER. 2006. Optimized
protocols for the isolation of specific protein-binding pep-
tides or peptoids from combinatorial libraries displayed on
beads. Mol. Biosyst. 2: 25–35.
16. BEEBE, K.D. et al. 2000. Determination of the binding speci-
ficity of the SH2 domains of protein tyrosine phosphatase
SHP-1 through the screening of a combinatorial phospho-
tyrosyl peptide library. Biochemistry 39: 13251–13260.
11
17. SWEENEY, M.C. et al. 2005. Decoding protein-protein inter-
actions through combinatorial chemistry: sequence speci-
ficity of SHP-1, SHP-2, and SHIP SH2 domains. Bio-
chemistry 44: 14932–14947.
18. ALLURI, P.G. et al. 2003. Isolation of protein ligands from
large peptoid libraries. J. Am. Chem. Soc. 125: 13995–
14004.
19. LIN, H.J. & T. KODADEK. 2005. Photo-induced oxida-
tive cross-linking as a method to evaluate the specificity
of protein-ligand interactions. J. Pept. Res. 65: 221–
228.
20. AMINI, F. et al. 2003. Using oxidative crosslinking and
proximity labeling to quantitatively characterize protein-
protein and protein-Peptide complexes. Chem. Biol. 10:
1115–1127.
21. LIU, B. et al. 2005. A potent transactivation domain mimic
with activity in living cells. J. Am. Chem. Soc. 127: 8254–
8255.
22. UTTAMCHANDANI, M. et al. 2004. Microarrays of tagged
combinatorial triazine libraries in the discovery of small-
molecule ligands of human IgG. J. Comb. Chem. 6: 862–
868.
23. MITSOPOULOS, G., D.P. WALSH & Y.T. CHANG. 2004. Tagged
library approach to chemical genomics and proteomics.
Curr. Opin. Chem. Biol. 8: 26–32.
24. KHERSONSKY, S. M. et al. 2003. Facilitated forward chem-
ical genetics using a tagged triazine library and ze-
brafish embryo screening. J. Am. Chem. Soc. 125: 11804–
11805.
25. BACKES, B. J. et al. 2000. Synthesis of positional-scanning
libraries of fluorogenic peptide substrates to define the
extended substrate specificity of plasmin and thrombin.
Nat. Biotechnol. 18: 187–193.
26. ALEXANDER, M. D. et al. 2006. A central strategy for con-
verting natural products into fluorescent probes. Chem-
biochem 7: 409–416.