Labeling and purification of cellulose-binding proteins for high resolution fluorescence applications

Article abstract of DOI:10.1021/ac901183b

The study of enzymatic reactions through fluorescence spectroscopy requires the use of bright, functional fluorescent molecules. In the case of proteins, labeling with fluorescent dyes has been carried out through covalent reactions with specific amino ac

Full text of DOI:10.1021/ac901183b

Anal. Chem. 2009, 81, 7981–7987
Labeling and Purification of Cellulose-Binding
Proteins for High Resolution Fluorescence
Applications
Jose M. Moran-Mirabal,*,† Stephane C. Corgie,†,§ Jacob C. Bolewski,†,§ Hanna M. Smith,†,§
Benjamin R. Cipriany,‡,§ Harold G. Craighead,‡,§ and Larry P. Walker*,†
Department of Biological and Environmental Engineering and School of Applied and Engineering Physics, Cornell
University, Ithaca, New York 14850
The study of enzymatic reactions through fluorescence
spectroscopy requires the use of bright, functional fluo-
rescent molecules. In the case of proteins, labeling with
fluorescent dyes has been carried out through covalent
reactions with specific amino acids. However, these
reactions are probabilistic and can yield mixtures of
unlabeled and labeled enzymes with catalytic activities that
can be modified by the addition of fluorophores. To have
meaningful interpretations of results from the study of
labeled enzymes, it is then necessary to reduce the
variability in physical, chemical, and biological charac-
teristics of the labeled products. In this paper, a solid
phase labeling protocol is described as an advantageous
alternative to free solution labeling of cellulose-binding
proteins and is applied to tag cellulases with three
different fluorophores. The products from the labeling
reactions were purified to remove the unreacted dye and
separate labeled and unlabeled enzymes. Characterization
of the catalytic and spectroscopic properties of the isolated
labeled species confirmed that highly homogeneous popu-
lations of labeled cellulases can be achieved. The protocol
for the separation of labeled products is applicable to any
mixture of labeled proteins, making this an attractive
methodology for the production of labeled proteins suit-
able for single molecule fluorescence spectroscopy.
changes,5,6 and catalytic activity.5,7 The application of SMFS to
enzyme studies requires tagging the proteins with bright, fluo-
rescent moieties, while retaining wild-type activity. Organic dyes
are small molecules that can confer fluorescence capabilities with
minimal effect on enzymatic activity.8,9 Additionally, because the
size of organic dyes lies in the 500-1500 Da range, they offer the
advantage of minimally increasing the enzyme’s mass even at high
degrees of labeling (DoL). Limitations of performing measure-
ments on molecules labeled with a single fluorophore are intensity
fluctuations and bleaching,10,11 which calls for enzymes labeled
with multiple fluorophores without loss of catalytic activity.
Labeling with organic dyes yields mixtures of labeled enzymes
based on the quantity and accessibility of reactive sites.12 This
polydispersity of labeled molecules, along with unlabeled enzymes,
can introduce significant variability, both in spectral and functional
properties. Thus, to carry out SMFS and obtain results that
accurately describe biological activities, it is important to purify
the labeled mixtures by separating the different populations of
labeled enzymes and characterize those populations to ensure
wild-type catalytic activity. Few studies have fully characterized
labeling through a variety of labeling molar ratios (LMRs),8
concentrations,13 and labeling methods,14-16 explored labeled
product distribution, or evaluated the SM performance of labeled
enzymes12,16-18 and their functionality.8,19 This manuscript pre-
sents methods to label Thermobifida fusca cellulases Cel5A, Cel6B,
and Cel9A (as models of cellulose-binding proteins), with three
(5) Zhuang, X.; Bartley, L. E.; Babcock, H. P.; Russell, R.; Ha, T.; Herschlag,
Single molecule fluorescence spectroscopy (SMFS) techniques
provide measurements of molecular events and reveal behaviors
otherwise obscured by ensemble averages. SMFS has been used
to study molecular associations,1,2 displacement,3,4 conformational
D.; Chu, S. Science 2000, 288, 2048–2051
(6) Schuler, B.; Lipman, E. A.; Eaton, W. A. Nature 2002, 419, 743–747
(7) Lu, H. P.; Xun, L.; Xie, X. S. Science 1998, 282, 1877–1882
(8) Voloshina, N. P.; Haugland, R. P.; Bishop, J.; Bhalgat, M.; Millard, P.; Mao,
F.; Leung, W. Y.; Haugland, R. P. Mol. Biol. Cell 1997, 8, 2017–2017
(9) Kapanidis, A. N.; Weiss, S. J. Chem. Phys. 2002, 117, 10953–10964
(10) Rasnik, I.; McKinney, S. A.; Ha, T. Nat. Methods 2006, 3, 891–893
(11) Vogelsang, J.; Kasper, R.; Steinhauer, C.; Person, B.; Heilemann, M.; Sauer,
M.; Tinnefeld, P. Angew. Chem., Int. Ed. 2008, 47, 5465–5469
(12) Craig, D. B.; Dovichi, N. J. Anal. Chem. 1998, 70, 2493–2494
(13) Huang, S. J.; Wang, H. Y.; Carroll, C. A.; Hayes, S. J.; Weintraub, S. T.;
Serwer, P. Electrophoresis 2004, 25, 779–784
(14) Pinto, D. M.; Arriaga, E. A.; Sia, S.; Li, Z.; Dovichi, N. J. Electrophoresis
1995, 16, 534–540
(15) Kantchev, E. A. B.; Chang, C. C.; Cheng, S. F.; Roche, A. C.; Chang, D. K.
Org. Biomol. Chem. 2008, 6, 1377–1385
(16) Dismer, F.; Hubbuch, J. J. Chromatogr., A 2007, 1149, 312–320
(17) Grunwaldt, G.; Haebel, S.; Spitz, C.; Steup, M.; Menzel, R. J. Photochem.
Photobiol., B 2002, 67, 177–186
(18) Teske, C. A.; Simon, R.; Niebisch, A.; Hubbuch, J. Biotechnol. Bioeng. 2007,
98, 193–200
.
.
.
.
.
.
* To whom correspondence should be addressed. E-mail: jmm248@cornell.edu
(J.M.M.-M.); lpw1@cornell.edu (L.P.W.).
.
† Department of Biological and Environmental Engineering.
‡ School of Applied and Engineering Physics.
.
§ E-mail: scc37@cornell.edu (S.C.C.); jcb348@cornell.edu (J.C.B.); hms42@
cornell.edu (H.M.S.); brc34@cornell.edu (B.R.C.); hgc1@cornell.edu (H.G.C.).
(1) Ha, T.; Enderle, T.; Ogletree, D. F.; Chemla, D. S.; Selvin, P. R.; Weiss, S.
.
.
Proc. Natl. Acad. Sci. U.S.A. 1996, 93, 6264–6268
(2) Schwille, P.; Meyer Almes, F. J.; Rigler, R. Biophys. J. 1997, 72, 1878–
1886
(3) Noji, H.; Yasuda, R.; Yoshida, M.; Kinosita, K. Nature 1997, 386, 299–
302
(4) Funatsu, T.; Harada, Y.; Tokunaga, M.; Saito, K.; Yanagida, T. Nature 1995,
374, 555–559
.
.
.
.
.
.
.
.
10.1021/ac901183b CCC: $40.75  2009 American Chemical Society
Published on Web 09/03/2009
Analytical Chemistry, Vol. 81, No. 19, October 1, 2009 7981

different dyes, separate the labeled products, and test their
catalytic and optical properties.
Table 1. Fluorophore and Enzyme Propertiesa
λEx
MW
charge
lysine residues
Cellulases are enzymes that catalyze the hydrolysis of cellulose
by cleaving glucosidic bonds within cellulose chains (endocellu-
lases) or from chains’ ends (exocellulases) producing soluble
oligosaccharides.20 They are classified as either random or
processive depending on whether they catalyze single or multiple
bond cleavages in succession. Cellulases are key in the production
of soluble sugars from lignocellulosic material that can be used
for biofuel and bioproducts production and have become the focus
of many fluorescence studies that explore their catalytic activity,
processivity, and interaction with cellulosic materials.21-27 Yet,
these studies have lacked the resolution to elucidate processivity
and catalysis during cellulose depolymerization at the nanoscale,
which could be provided by SMFS.
The goals of this paper were (1) to develop and optimize a
solid phase (SP) labeling method, where cellulases are im-
mobilized on cellulose, as an advantageous alternative to free
solution (FS) labeling, where the proteins freely diffuse in an
aqueous medium; (2) to separate the labeled mixtures obtained
into their components; (3) to characterize the activities and optical
properties of the separate labeled products.
molecule (nm) CF
ε
(Da) (at pH 7.0) (accessible/total)
AF350
AF488
AF647
Cel5A
Cel6B
Cel9A
346 0.19 19000
495 0.11 71000
410
643
0
-1
650 0.03 239000 ∼1300
-3
97100 46300
-16
-37
-52
3/10
4/11
10/20
115150 59600
210670 90400
a Data for fluorophores from Alexa Fluor Amine-Reactive Probes
Catalog, Molecular Probes. The number of accessible lysine residues
was calculated as a minimum of 30% surface exposure.
based on binding capacity (Figure S-1 in the Supporting Informa-
tion). SP labeling was done in Spin-X columns (0.45 µm nylon
membrane, Costar, Lowell, MA). For the optimization of SP
labeling, two protocols were tested: SPA was carried out with
constant agitation under the same volume conditions as FS,
while SPB used reduced volume and no agitation (details for
the two protocols are available in the Supporting Information).
The binding matrix was loaded onto the columns, and excess
buffer was removed by centrifugation. Then, enzymes were added
in 300 µL of labeling buffer and allowed to bind for 1 h at 4 °C
with end-over-end agitation, after which the buffer was removed
by centrifugation. The fluorophore was added, and the volume
was completed to the specified volume by addition of labeling
buffer and 1 M sodium bicarbonate. After labeling, unreacted dye
was removed by six consecutive centrifugations (5000g, 2 min),
followed by additions of 500 µL of labeling buffer. Between washes,
samples were agitated using a magnetic stir bar (500 rpm, 5 min),
after which the sidewalls of the tube were scrubbed using an
external magnet. The labeled enzymes were recovered by three
elutions with chilled ethylene glycol (EG). Two washes were done
with 400 µL of EG, incubation in ice for 10 min, and centrifugation
(6000g, 5 min). The third was done with 200 µL of EG and
centrifugation (10 000g, 10 min). The flow-throughs were im-
mediately diluted 3× in labeling buffer and kept in ice.
EXPERIMENTAL SECTION
Cellulase Production and Labeling. T. fusca cellulases
Cel5A, Cel6B, and Cel9A were expressed and purified by size
exclusion as previously described.21,28 Cellulases were labeled with
amine-reactive AF350, AF488, or AF647 (Invitrogen, Carlsbad, CA)
in FS or on SP. In both strategies, 3000 pmols of enzyme were
labeled at 10 µM concentration with LMRs of 10:1, 30:1, or 100:1.
All reactions were conducted in triplicate at 4 °C, for 24 h, in the
dark, and were stopped by unreacted dye removal.
Free Solution Labeling. Enzymes and dye were mixed in a low
adhesion tube; the volume was completed to 270 µL with 35 mM
boric acid, 50 mM NaCl buffer pH 8.3 (labeling buffer), and 30
µL of 1 M sodium bicarbonate was added. Removal of unreacted
dye from labeling mixtures was performed as previously described
through native polyacrylamide gel electrophoresis (n-PAGE), and
protein was recovered via electroelution.26
Concentration and Characterization. Removal of EG, buffer
exchange with 20 mM MES pH 6.0 (MES buffer), and concentra-
tion of labeled products was done in Vivaspin4 columns (10 kDa
MWCO, Sartorius, Bohemia, NY). Protein concentrations (Cp) and
D/P were calculated by absorbance according to
Solid Phase Labeling. An 8:1 (w/w) mixture of CF11-BMCC
resuspended at 32 mg/mL was prepared as the SP binding matrix
(CF11 cellulose powder from Whatman, Piscataway, NJ, and
bacterial microcrystalline cellulose (BMCC) from Monsanto
Cellulon, San Diego, CA). The ratio of the components was chosen
[A280 - (AEX × CF)]
CP )
(1)
(2)
εP
(19) Berlier, J. E.; Rothe, A.; Buller, G.; Bradford, J.; Gray, D. R.; Filanoski, B. J.;
Telford, W. G.; Yue, S.; Liu, J. X.; Cheung, C. Y.; Chang, W.; Hirsch, J. D.;
Beechem, J. M.; Haugland, R. P.; Haugland, R. P. J. Histochem. Cytochem.
AEX
εF × CP
2003, 51, 1699–1712
(20) Wilson, D. B. Ann. N. Y. Acad. Sci. 2008, 1125, 289–297, DOI: 10.1196/
.
D/P )
annals.1419.026.
(21) Jeoh, T.; Wilson, D. B.; Walker, L. P. Biotechnol. Prog. 2002, 18, 760–769
(22) Hong, J.; Ye, X. H.; Zhang, Y. H. P. Langmuir 2007, 23, 12535–12540
(23) Pinto, R.; Amaral, A. L.; Carvalho, J.; Ferreira, E. C.; Mota, M.; Gama, M.
Biotechnol. Prog. 2007, 23, 1492–1497
(24) Pinto, R.; Carvalho, J.; Mota, M.; Gama, M. Cellulose 2006, 13, 557–569
(25) Jervis, E. J.; Haynes, C. A.; Kilburn, D. G. J. Biol. Chem. 1997, 272, 24016–
24023
(26) Moran-Mirabal, J. N.; Santhanam, N.; Corgie, S. C.; Craighead, H. G.;
Walker, L. P. Biotechnol. Bioeng. 2008, 101, 1129–1141
(27) Santhanam, N.; Walker, L. P. Biol. Eng. 2009, 1, 5–23
(28) Jung, E. D.; Lao, G. F.; Irwin, D.; Barr, B. K.; Benjamin, A.; Wilson, D. B.
Appl. Environ. Microbiol. 1993, 59, 3032–3043
.
where A280 is the absorbance at 280 nm, AEx is the absorbance
at the fluorophore’s excitation wavelength, CF is the fluoro-
phore correction factor, and ε is the extinction coefficient (Table
.
.
.
1
). The term D/P was used to denote the presence of labeled
and unlabeled proteins in the labeled mixture and differentiate it
.
from the DoL for pure labeled species.18
.
Separation of Labeled Mixtures. Separation of labeled
mixtures was done by fast protein liquid chromatography (FPLC)
¨
at 21 °C with MES buffer as the mobile phase, using an AKTA
.
.
7982 Analytical Chemistry, Vol. 81, No. 19, October 1, 2009

Explorer 10S FPLC system and a Resource Q column (1 mL, GE
Healthcare, Piscataway, NJ). The column was equilibrated to the
initial NaCl concentration (80 mM NaCl for Cel5A, 250 mM for
Cel9A, and 250 mM for Cel6B), and 1 mL of labeling mixture
was injected. Proteins were eluted by applying a 250 mM linear
salt gradient (3 mL/10 mM NaCl) at a 1.5 mL/min flow rate.
Absorbance was recorded at 280 nm for proteins and at 346, 495,
or 650 nm for samples labeled with AF350, AF488, and AF647,
respectively. Eluted proteins were collected in 1 mL fractions and
partitioned according to the peaks observed in the chromato-
grams. Fractions were consolidated and concentrated, and buffer
was exchanged with 50 mM sodium acetate pH 5.5 using Vivaspin
4 columns (Sartorius, Bohemia, NY). Cp and DoL were measured
for the separate labeled products.
Hydrolysis Assay. The hydrolytic activities of labeled cellu-
lases were quantified on BMCC by measuring the total oligosac-
charides produced and comparing them to those of unlabeled
enzymes. Hydrolysis reactions were carried out in 600 µL of
sodium acetate buffer with 1.5 mg of BMCC and 5 pmol of
enzyme. Hydrolysis reactions were performed in triplicate at 50
°C for 24 h under constant agitation. Filtration through the Spin-X
columns (5000g, 5 min) stopped the reaction by removing the
cellulose substrate. The flow-through was incubated for 1 h on
ice in a column with fresh BMCC to capture remaining enzymes
and then recovered by centrifugation (5000g, 5 min). Oligosac-
charides were measured by refractive index using a Shimadzu
HPLC (Shimadzu, Columbia, MD). Hydrolysis samples (injection
volume: 50 µL) were eluted at 0.6 mL/min through an HPX-87P
Bio-Rad Aminex column (BioRad, Hercules, CA), heated to 84
°C, using 18.2 MΩ water as the mobile phase. The chromatograms
for the products were baseline corrected, and the peaks were fitted
to Gaussian functions using Origin8Pro (OriginLab, Northhamp-
ton, MA). Standards containing cellotetraose, cellotriose, cello-
biose, and glucose were used to calculate the oligosaccharide
concentrations from peak areas. The total oligosaccharides
produced were converted into either cellobiose (Cel5A and Cel6B)
or cellotetraose equivalents (Cel9A), and hydrolytic activities were
expressed as nmol of cellobiose or cellotetraose equivalents/nmol
enzyme/h.
Characterization of Photon Output and Imaging. SM
fluorescence from AF647 labeled Cel9A cellulases was recorded
using an inverted microscope (IX-71 Olympus, Center Valley, PA)
equipped for laser-induced fluorescence. A 647 nm laser beam
(ArKr, Melles-Griot, Carlsbad, CA) incident on a dual-band
dichroic mirror (488/647, Chroma Technology, Rockingham, VT)
was focused on a microfluidic channel through a 60×/1.2 NA
water-immersion objective (UPlanSAPO, Olympus, Center Valley,
PA). Fluorescence was filtered through an emission filter (680/
40M, Chroma Technology, Rockingham, VT) and collected using
a 100 µm diameter core multimode optical fiber (OZ Optics,
Ottawa, ON, Canada). Photons emitted by single AF647-labeled
cellulases were detected by avalanche photodiodes (APDs, Perkin-
Elmer, Waltham, MA). A high-speed correlator (correlator.com)
was used to record the APD photon count. Microfluidic channels
were fabricated as previously described29,30 to constrict the
excitation volume to submicrometer dimensions and achieve SM
confinement at 1nM concentration.
The microfluidic channels were initially filled with running
buffer (4× PBS containing 0.5% Triton X-100 v/v, both from Sigma
Aldrich, St. Louis, MO). Electrokinetic drive was established by
applying 50 V across the fluidic channel. The combined autofluo-
rescence of the channel and buffer was determined by illumination
with 500 µW laser power. The channels were then cleared and
filled with 200 pM labeled enzymes in running buffer, and
fluorescence from the SMs was recorded. Fluorescence photon
counting/burst histogram analysis was performed using a custom
Matlab routine.30 The threshold for a SM event was established
at 5σ above background. The photon output was collected for a
total of 10 000 molecules. A histogram of SM data was plotted
using 25 photon bin, and the resulting distributions were fitted
to Gaussian functions to determine average photon output. Images
of Cel9A cellulases (100 fM) bound to fluorescently labeled
cellulose were acquired using a FV1000 confocal microscope and
a 100×/1.45 NA oil immersion objective (Olympus, Center Vally,
PA).
Statistical Analysis. All statistical analyses were performed
with raw data using Origin8Pro (OriginLab, Northampton, MA).
The comparison between D/P ratios for all labeling treatments
and cellulases was performed via a two-way ANOVA and Bonfer-
roni significance test using confidence intervals (CI) set to 99%.
D/P ratios for each cellulase were compared by pairing the data
grouped by LMR with either the fluorophore or labeling method
(while maintaining the other factor constant). Statistical analysis
of catalytic activity was performed via a two-tail Student t test for
the mean of unpaired samples. The activity of each assayed labeled
fraction was directly compared with the unlabeled enzyme and
tested at 95 and 99% CI.
RESULTS AND DISCUSSION
Cellulase Labeling. Removal of unreacted dye after labeling
is important to stop the labeling reaction and avoid excess dye
that contributes to fluorescence background in quantitative ap-
plications. The two labeling protocols applied different strategies
for unbound fluorophore removal: in FS, n-PAGE followed by
electroelution of the labeling mixtures was used to effectively
separate the free dye from the proteins (Figure S-2 in the
Supporting Information); in SP, free dye was eluted through
centrifugation and washing (Figure S-3 in the Supporting Informa-
tion). Both strategies successfully removed excess dye from
labeled samples.
The common measure of labeling efficiency is the ratio of dye
to protein molecules present in the sample (referred to as dye to
protein ratio (D/P) for labeled mixtures or DoL for labeled
products after FPLC separation). Preliminary experiments labeling
Cel9A with AF647 showed that, for FS, the D/P varied linearly
with LMRs between 10 and 200 (Figure S-4 in the Supporting
Information). Thus, only LMRs of 10, 30, and 100 were compared
in the labeling experiments. Figure 1 shows that, as the LMR
increases, the D/P for all enzyme-fluorophore-labeling method
combinations significantly increases (CI > 99%) due to a greater
reaction rate between fluorophores and reactive sites. Additional
(29) Stavis, S. M.; Corgie, S. C.; Cipriany, B. R.; Craighead, H. G.; Walker, L. P.
(30) Stavis, S. M.; Edel, J. B.; Samiee, K. T.; Craighead, H. G. Lab Chip 2005,
Biomicrofluidics 2007, 1, 034105
.
5, 337–343.
Analytical Chemistry, Vol. 81, No. 19, October 1, 2009 7983

Figure 1. Comparison of labeling efficiency for the different fluorophore-enzyme combinations measured through D/P. SPA and FS labeling
protocols were compared for labeling with three fluorophores (AF350, AF488, and AF647) at three LMRs (10:1, 30:1, 100:1).
labeling experiments comparing SPA and SPB protocols showed
that D/P was also improved if the reaction volume was reduced,
increasing the fluorophore concentration (Figure S-5 in the
Supporting Information). Thus, SPB enabled labeling reactions
with minimal volume and was used to increase the D/P without
increasing the fluorophore loading.
ing find was that labeled mixtures with D/Ps above two still
contained unlabeled proteins, meaning that characterization of
mixtures based solely on D/P does not guarantee that all
molecules possess at least one label. The separated fractions were
partitioned according to the peaks observed in the chromatograms
and characterized for their DoL.
Under identical labeling conditions, each fluorophore had
different labeling efficiency (Figure 1), with AF350 exhibiting the
highest and AF488 exhibiting the lowest efficiency. This was
attributed to the structural characteristics of each fluorophore.
Because AF350 is the smallest fluorophore, it has greater acces-
sibility to lysine residues and yields higher D/P. Charge and
hydrophobicity can also affect the interactions with proteins and
alter the labeling efficiency. It was also observed that, for any given
fluorophore, the D/P for each cellulase is different (Figure 1) and
correlates with the number of accessible lysine residues (Table
1). Thus, both the amino acid sequence and protein conformation
are as important as the size, charge, and hydrophobicity of the
fluorophore and can affect the labeling efficiency.
Comparison of the labeling methods in FS and SP showed that
their D/P was significantly different for all fluorophore-enzyme
combinations (Figure 1, CI > 99), except Cel5A-AF488, Cel6B-
AF350, and Cel9A-AF488, where no difference was observed.
Overall, labeling of any enzyme with AF647 was more efficient in
FS, labeling of Cel6B with AF488 was more efficient in FS, and
labeling of Cel5A and Cel9A with AF350 was more efficient with
SP. Other fluorophore-enzyme combination labeling efficiencies
were similar for FS or SP. The general trend was that, under
identical conditions, SP yielded lower D/Ps than FS. Cellulose
binding changes the accessibility to lysine residues either via
conformational changes or by steric hindrance. The fact that
AF647 consistently labeled more efficiently with FS suggests that,
as the enzymes bind to cellulose, access to lysine residues
becomes restricted, preventing labeling by large fluorophores.
Separation of Labeled Mixtures. FPLC chromatograms
(Figure 2 and Figures S-6 and S-7 in the Supporting Information)
show the separation of mixtures obtained from labeling cellulases
in FS or SP with AF647, AF350, and AF488. The first eluted peak
represents the unlabeled protein, which only shows absorbance
at 280 nm. All other peaks showing absorbance at both 280 nm
and the fluorophore absorbance wavelength represent labeled
products. Consistent with an increase in D/P, as LMR increased,
the amount of unlabeled protein decreased and in some cases
disappeared, while the total labeled products increased. A surpris-
The separation between unlabeled and labeled enzymes
increased with increasing fluorophore charge (Figure 2, Figures
S-6 and S-7 in the Supporting Information, and Table 1). This was
also true for the separation of the labeled products, where more
negatively charged fluorophores had labeled peaks that were
better resolved than those from less charged fluorophores. Peaks
that were well resolved represent pure labeled species, as their
measured DoL was always close to an integer. On the other hand,
convoluted peaks can represent a mixture of two or more labeled
species. Products with the same DoL, but different elution times,
can also represent different combinations of labeled sites. Overall,
molecules with higher DoL eluted later than those with lower DoL,
and protein distribution shifted from lower to higher DoL products
as LMR increased (Figure 2 and Figures S-6 and S-7 in the
Supporting Information). The results suggest that anionic ex-
change FPLC can separate enzymes labeled with few fluorophores
better than those that contain multiple tags. Yet, products with
high DoL represent combinations of few labeled products and are
still more homogeneous than labeled mixtures.
The chromatograms obtained from enzymes labeled in FS
versus SP showed significantly different distributions. Although
some of the labeled peaks coincide for both labeling methods,
the profiles are different in number of peaks, their size, and
distribution (Figure 2). Because, under identical conditions, the
D/P was higher for FS than for SP, differences in the amplitude
of the labeled species were expected between methods. However,
differences in the number and separation between species were
also observed. Chromatograms for SP usually had fewer and better
resolved peaks than those for FS. Thus, the comparison between
the separations obtained from mixtures labeled with FS and SP
showed that cellulase binding to cellulose alters the labeling by
modifying the accessibility of some primary amines.
Catalytic Activity of Labeled Cellulases. Fluorescence label-
ing of enzymes can affect their binding and catalytic activity
through steric hindrance and accessibility to functional sites,
changes in hydrophobicity and charge, and even a disruption of
the overall 3D structure of the enzyme. The effect of the labeling
method and DoL on oligosaccharide production for cellulases
7984 Analytical Chemistry, Vol. 81, No. 19, October 1, 2009

Figure 2. FPLC chromatograms of the separation of samples labeled with AF647 either by FS or by SPA. Chromatograms have been normalized
to the total area under the A280 curve. Unlabeled protein elutes earlier than labeled products. Identified peaks representing populations of labeled
products were consolidated and characterized for their DoL. Labels in red represent products where not enough enzyme was recovered to
measure DoL.
Cel5A, Cel6B, and Cel9A was evaluated through a hydrolysis
assay. Results show that Cel9A produced cellotetraose, cellotriose,
cellobiose, and glucose, Cel5A produced cellobiose and cellotriose,
and Cel6B produced almost exclusively cellobiose (Figure 3a).
When binding to the cellulose strand, Cel6B and Cel9A depoly-
merize cellulose by processively cleaving multiple oligosaccharide
units (of cellobiose and cellotetraose, respectively) before detach-
ing from the strand.31-33 On the other hand, Cel5A cleaves a
single cellobiose unit every time it binds to the cellulose strand.31
Thus, Cel5A and Cel6B matched the expected oligosaccharide
production, while Cel9A showed products other than cellotetraose,
suggesting the enzymatic conversion of cellotetraose in solution.
The results from cellulose depolymerization are presented in
Figure 3b-d. Cel9A yielded the most labeled products, and only
a few were assayed to cover the spectrum of DoLs obtained. On
the other hand, all the AF647-labeled Cel5A and Cel6B products
were assayed. None of the labeled products recovered for Cel5A
presented significant changes in activity (Figure 3b, CI 95%),
indicating that adding a single AF647 fluorophore to Cel5A does
not alter its catalytic activity. Labeled Cel6B (Figure 3c) showed
differences only for products labeled in FS (DoLs of 1.00, 1.90,
and 1.98), where activity was increased by the addition of one
fluorophore but decreased to wild-type levels upon further labeling.
The increase in activity for AF647-labeled Cel6B was an unex-
pected result that cannot be fully explained with current data.
Cel9A showed significant decrease in activity for products labeled
in FS with DoLs of 4.09 and 4.50 and in SP with a DoL of 6.21
(Figure 3d). The decrease in activity observed for labeled enzymes
from FS with DoL > 4 and from SP for DoL > 6 supports the
hypothesis that SP labeling protects the enzymatic activity.
Single Molecule Characterization of Photon Output. The
effect of DoL on photon output was evaluated for AF647-labeled
Cel9A. A sample plot of binned data for the photon output
measured from molecules with DoL 2.03 is presented in Figure
4a. Figure 4b presents the binned data for all the labeled products
tested. In addition to labeled products obtained after FPLC
separation, a labeled mixture was run as a comparison (dashed/
gray line). Products with DoL of 1 labeled with FS or SP have
overlapping distributions, showing that their emission properties
are similar. As DoL is increased (Figure 4b), the distributions
become wider as a result of the combinations of excitation states
for the fluorophores incorporated in the molecule. While mol-
ecules that have a single fluorophore can only be in either a dark
or a bright state, molecules with higher DoL can have all possible
combinations of dark and bright states for each fluorophore
attached. The trend for the photon output (Figure 4b) was that,
as DoL increased (up to a DoL of 5), the distribution shifted to
the right, i.e., more photons were emitted per molecule. When
the average photon output is plotted against DoL (Figure 4c), it
is observed that there is a monotonic increase as a function of
DoL, until the DoL surpasses a value of 4. Molecules with DoL >
4 do not show a linear increase in the average photon emission.
The extreme case is the DoL of 6.21, where a significant reduction
in photon output is observed. In theory, each added fluorophore
should increase the photon output by an equal amount. In practice,
the addition of multiple fluorophores can increase the photon
output only if they are incorporated at distances greater than their
Fo¨rster radius. Thus, the decrease in the photon output for heavily
labeled molecules can be attributed to quenching. These results
(31) Barr, B. K.; Hsieh, Y. L.; Ganem, B.; Wilson, D. B. Biochemistry 1996, 35,
586–592
(32) Irwin, D.; Shin, D. H.; Zhang, S.; Barr, B. K.; Sakon, J.; Karplus, P. A.; Wilson,
D. B. J. Bacteriol. 1998, 180, 1709–1714
(33) Sakon, J.; Irwin, D.; Wilson, D. B.; Karplus, P. A. Nat. Struct. Biol. 1997,
4, 810–818
.
.
.
Analytical Chemistry, Vol. 81, No. 19, October 1, 2009 7985

Figure 3. Evaluation of catalytic activities of labeled cellulases. (a) HPLC chromatogram with sample profile of oligosaccharides produced.
(b)-(d) Comparison of catalytic activities of labeled and native Cel5A, Cel6B, and Cel9A. Statistical significance (n g 3) is marked as * for 95%
and ** for 99%.
Figure 4. Photon output for single AF647 labeled Cel9A cellulases. (a) Raw distribution of binned data for photon output collected from ∼10 000
enzymes with DoL 2.03. Gaussian fit presented in red. (b) Distribution of photon output for data collected from labeled products assayed. (c)
Average photon output plotted vs DoL. (d) Confocal images of AF647 labeled Cel9A cellulases (red) bound to fluorescently labeled cellulose
(green). Cellulases with a DoL of 1 and 4 were incubated at 100 fM and imaged under identical conditions.
7986 Analytical Chemistry, Vol. 81, No. 19, October 1, 2009

show that overlabeling of Cel9A with AF647 does not necessarily
translate into brighter molecules and that a DoL of 4 is optimal.
Figure 4d exemplifies the advantage of the use of highly labeled
molecules in SM imaging. The comparison shows Cel9A mol-
ecules with a DoL of 1 and 4 bound to fluorescently labeled
cellulose. The higher average photon output for a DoL of 4 allows
greater signal-to-noise ratio, which makes SM localization more
accurate.
possible to eliminate the presence of any unlabeled enzymes and
separate labeled species with well-defined DoLs. The separation
of labeled species was done based on the charge carried by the
fluorophore, which allows this separation method to be applied
to separate any mixture of labeled proteins.
The purification of more homogeneous labeled products
allowed the thorough characterization of functional properties of
AF647-labeled cellulases. Measurement of the catalytic activity of
labeled cellulases on BMCC showed that labeling with few
fluorophores did not change their catalytic activity, but overla-
beling could reduce it. This could be due to labeling of lysine
residues that participate in binding or catalysis or to an overall
disruption of the enzyme structure. The threshold for overlabeling
was lower for FS than for SP, which shows that functional residues
are protected by cellulose binding. A special case was Cel6B,
where activity increased by 30% for a DoL of 1 when labeled in
FS. The root of the increase cannot be traced with current data
but could be related to enhancement of hydrophobic interactions
with cellulose. Finally, the characterization of the emission
characteristics of labeled cellulases by SM experiments showed
that overlabeling does not always result in an increased photon
output. The labeled cellulases obtained from the labeling and
separation methods can be used in the future for SMFS studies
with confidence that their performance mirrors that of native
enzymes.
CONCLUSIONS
In this paper, methods have been described to label cellulose-
binding proteins and separate the resulting mixtures into labeled
products with well-defined optical and functional characteristics.
A solid phase labeling method for cellulose-binding proteins was
developed and optimized as an alternative to free solution labeling.
Although the FS labeling protocol can be applied to any protein,
the SP method presents several advantages: (1) protection of
binding and catalytic residues during labeling, (2) reduced volume
for labeling reactions, (3) ease of removal of unreacted dye
through centrifugation and washes, and (4) high recovery of
labeled proteins. Provided a suitable immobilization matrix, these
advantages can be extended to other proteins. Comparison of the
labeling methods, solely through D/P, indicated that an increase
in LMR yields higher D/Ps; fluorophore size, charge, and
hydrophobicity affect the labeling efficiency, with AF350 being
the dye that labels cellulases most efficiently; and the labeling
efficiency for cellulases depends on the number of surface-
accessible lysines. Significant differences were observed between
the labeling efficiency of FS and SP, with some dye-cellulase
combinations exhibiting higher labeling efficiency with FS and
others with SP.
Separation of the labeled mixtures through the FPLC method
developed showed significant amounts of unlabeled enzymes
present, along with a multiplicity of labeled products. Thus,
characterization solely based on the D/P is not an accurate
depiction of the labeled mixtures. Use of labeled mixtures in SM
experiments could translate into the observation of an average
behavior of the different labeled products. Therefore, separation
of labeled mixtures into its components is necessary for the
isolation of more homogeneous labeled molecules, with known
functional characteristics. A comparison of the separated mixtures
from FS against those from SP showed differences in the
distribution of labeled products. This is a reflection of the changes
in lysine accessibility induced by cellulase binding onto the
cellulosic matrix. With the FPLC protocol developed, it was
ACKNOWLEDGMENT
We thank Professor David Wilson and Diana Irwin for assis-
tance with enzyme production and purification, Ted Thannhauser
and Simon Hucko for assistance with FPLC protocols, Ed Evans and
Dustin McCall for technical assistance in labeling reactions, and
Deborah Sills and Sarah Munro for assistance with HPLC protocols.
This research made use of equipment and facilities from the
Biofuels Research Laboratory at Cornell University, a facility
funded by the NYS Empire Development Corporation. This
research was funded by DOE Contract GO18084 and by the New
York State Foundation for Science, Technology and Innovation.
SUPPORTING INFORMATION AVAILABLE
Seven additional figures available as Supporting Information.
This material is available free of charge via the Internet at
http://pubs.acs.org.
Received for review May 31, 2009. Accepted August 19,
2009.
AC901183B
Analytical Chemistry, Vol. 81, No. 19, October 1, 2009 7987