Femtomolar Detection of Prostate-Specific Antigen: An Immunoassay Based on Surface-Enhanced Raman Scattering and Immunogold Labels

Article abstract of DOI:10.1021/ac034356f

A novel reagent for low-level detection in immunoadsorbent assays is described. The reagent consists of gold nanoparticles modified to integrate bioselective species (e.g., antibodies) with molecular labels for the generation of intense, biolyte-selective

Full text of DOI:10.1021/ac034356f

Anal. Chem. 2003, 75, 5936-5943
Femtomolar Detection of Prostate-Specific
Antigen: An Immunoassay Based on
Surface-Enhanced Raman Scattering and
Immunogold Labels
Desiree S. Grubisha, Robert J. Lipert,* Hye-Young Park, Jeremy Driskell, and Marc D. Porter*
Microanalytical Instrumentation Center, Ames LaboratorysUSDOE, and Department of Chemistry, Iowa State University,
Ames, Iowa 50011
tering (SERS),14-16 quantum dots,17-22 microcantilevers,23,24 and
atomic force microscopy25-27 have been devised. Because of its
inherently high sensitivity, fluorescence-based detection is among
the most used readout modality.
This paper describes our latest findings from a continuing
investigation of SERS as a multiplexed, immunoassay readout
concept.16 In its traditional formats (i.e., normal and resonance),
A novel reagent for low-level detection in immunoadsor-
bent assays is described. The reagent consists of gold
nanoparticles modified to integrate bioselective species
(e.g., antibodies) with molecular labels for the generation
of intense, biolyte-selective surface-enhanced Raman scat-
tering (SERS) responses in immunoassays and other
bioanalytical applications. The reagent is constructed by
coating gold nanoparticles (3 0 nm) with a monolayer of
an intrinsically strong Raman scatterer. These monolayer-
level labels are bifunctional by design and contain disul-
fides for chemisorption to the nanoparticle surface and
succinimides for coupling to the bioselective species.
There are two important elements in this label design; it
both minimizes the separation between label and particle
surface and maximizes the number of labels on each
particle. This approach to labeling also exploits several
other advantages of SERS-based labels: narrow spectral
bandwidth, resistance to photobleaching and quenching,
and long-wavelength excitation of multiple labels with a
single excitation source. The strengths of this strategy are
demonstrated in the detection of free prostate-specific
antigen (P SA) using a sandwich assay format based on
monoclonal antibodies. Detection limits of ∼1 pg/ mL in
human serum and ∼4 pg/ mL in bovine serum albumin
have been achieved with a spectrometer readout time of
6 0 s. The extension of the method to multianalyte assays
(e.g., the simultaneous determination of the many com-
plexed forms of P SA) is discussed.
(5) Brown, C. R.; Higgins, K. W.; Frazer, K.; Schoelz, L. K.; Dyminski, J. W.;
Marinkovich, V. A.; Miller, S. P.; Burd, J. F. Clin. Chem. 1 9 8 5 , 31, 1500-
5.
(6) Hayes, F. J.; Halsall, H. B.; Heineman, W. R. Anal. Chem. 1 9 9 4 , 66, 1860-
5.
(7) Butler, J. E. J. Immunoassay 2 0 0 0 , 21, 165-209.
(8) Lyon, L. A.; Musick, M. D.; Natan, M. J. Anal. Chem. 1 9 9 8 , 70, 5177-83.
(9) Knoll, W.; Zizlsperger, M.; Liebermann, T.; Arnold, S.; Badia, A.; Liley, M.;
Piscevic, D.; Schmitt, F. J.; Spinke, J. Colloids Surf., A 2 0 0 0 , 161, 115-37.
(10) Liebermann, T.; Knoll, W.; Sluka, P.; Herrmann, R. Colloids Surf. A 2 0 0 0 ,
169, 337-50.
(11) Nelson, B. P.; Grimsrud, T. E.; Liles, M. R.; Goodman, R. M.; Corn, R. M.
Anal. Chem. 2 0 0 1 , 73, 1-7.
(12) Lee, H. J.; Goodrich, T. T.; Corn, R. M. Anal. Chem. 2 0 0 1 , 73, 5525-31.
(13) Wegner, G. J.; Lee, H. J.; Corn, R. M. Anal. Chem. 2 0 0 2 , 74, 5161-8.
(14) Rohr, T. E.; Cotton, T.; Fan, N.; Tarcha, P. J. Anal. Biochem. 1 9 8 9 , 182,
388-98.
(15) Dou, X.; Takama, T.; Yamaguchi, Y.; Yamamoto, H.; Ozaki, Y. Anal. Chem.
1 9 9 7 , 69, 1492-5.
(16) Ni, J.; Lipert, R. J.; Dawson, G. B.; Porter, M. D. Anal. Chem. 1 9 9 9 , 71,
4903-8.
(17) Chan, W. C. W.; Nie, S. Science (Washington, D. C.) 1 9 9 8 , 281, 2016-8.
(18) Mattoussi, H.; Mauro, J. M.; Goldman, E. R.; Anderson, G. P.; Sundar, V.
C.; Mikulec, F. V.; Bawendi, M. G. J. Am. Chem. Soc. 2 0 0 0 , 122, 12142-
50.
(19) Sun, B.; Xie, W.; Yi, G.; Chen, D.; Zhou, Y.; Cheng, J. J. Immunol. Methods
2 0 0 1 , 249, 85-9.
(20) Goldman, E. R.; Anderson, G. P.; Tran, P. T.; Mattoussi, H.; Charles, P. T.;
Mauro, J. M. Anal. Chem. 2 0 0 2 , 74, 841-7.
A host of different immunoassay readout techniques has been
developed in past years.1 The more established approaches include
scintillation counting,2 fluorescence,3,4 chemiluminescence,5 elec-
trochemical,6 and enzymatic methods.1,7 Recently, strategies based
on surface plasmon resonance,8-13 surface-enhanced Raman scat-
(21) Speckman, D. M.; Jennings, T. L.; LaLumondiere, S. D.; Moss, S. C. Mater.
Res. Soc. Symp. Proc. 2 0 0 2 , 676, Y3.6.1-6.
(22) Tran, P. T.; Goldman, E. R.; Anderson, G. P.; Mauro, J. M.; Mattoussi, H.
Proc. SPIE-Int. Soc. Opt. Eng. 2 0 0 2 , 4636, 23-30.
(23) Wu, G.; Datar, R. H.; Hansen, K. M.; Thundat, T.; Cote, R. J.; Majumdar, A.
Nat. Biotechnol. 2 0 0 1 , 19, 856-60.
(24) Grogan, C.; Raiteri, R.; O’Connor, G. M.; Glynn, T. J.; Cunningham, V.; Kane,
M.; Charlton, M.; Leech, D. Biosens. Bioelectron. 2 0 0 2 , 17, 201-7.
(25) Allen, S.; Chen, X.; Davies, J.; Davies, M. C.; Dawkes, A. C.; Edwards, J. C.;
Roberts, C. J.; Sefton, J.; Tendler, S. J.; Williams, P. M. Biochemistry 1 9 9 7 ,
36, 7457-63.
* To whom correspondence should be addressed. E-mail: blipert@
porter1.ameslab.gov.
(1) Immunoassay; Diamandis, E. P., Christopoulos, T. K., Eds.; Academic
Press: New York, 1996.
(2) Gutcho, S.; Mansbach, L. Clin. Chem. 1 9 7 7 , 23, 1609-14.
(3) Vuori, J.; Rasi, S.; Takala, T.; Vaananen, K. Clin. Chem. 1 9 9 1 , 37, 2087-
92.
(4) Xu, Y. Y.; Pettersson, K.; Blomberg, K.; Hemmila, I.; Mikola, H.; Lovgren,
T. Clin. Chem. 1 9 9 2 , 38, 2038-43.
(26) Jones, V. W.; Kenseth, J. R.; Porter, M. D.; Mosher, C. L.; Henderson, E.
Anal. Chem. 1 9 9 8 , 70, 1233-41.
(27) Allen, S.; Chen, X.; Davies, J.; Davies, M. C.; Dawkes, A. C.; Edwards, J. C.;
Roberts, C. J.; Tendler, S. J. B.; Williams, P. M. Appl. Phys. A: Mater. Sci.
Process. 1 9 9 8 , A66, S255-61.
5936 Analytical Chemistry, Vol. 75, No. 21, November 1, 2003
10.1021/ac034356f CCC: $25.00 © 2003 American Chemical Society
Published on Web 09/23/2003

Raman spectroscopy lacks the sensitivity required for a readout
strategy. Using silver and gold nanoparticles as enhancing
substrates, however, recent reports have shown that the sensitivity
of SERS can rival that of fluorescence.28-30 This breakthrough has
led to renewed interest in the exploration of this information-rich
Scheme 1
31-36
spectroscopy as a tool for rapid, low-level readout.16,
SERS has three more characteristics that point to its potential
as a multiplexed readout technique.16,37 First, most Raman bands
are 10-100 times narrower than those of fluorescence, reducing
the likelihood of spectral overlap when multiple labels are used.
Second, the optimum SERS excitation wavelength is dependent
on the chemical/ physical properties of the enhancing substrate
and not the photophysics of the scatterer, facilitating multilabel
readout by requiring only one excitation wavelength. Third, Raman
responses are much less susceptible to photobleaching than
fluorescence, enabling the use of extended signal averaging to
lower detection limits.
The list of attributes can be expanded upon examining
scenarios for the development of a SERS-based readout concept.
Scheme 1, which idealizes our past approach for a dual-analyte
sandwich immunoassay,16 serves as a basis for identifying more
subtle attributes. In Scheme 1, a set of immobilized antibodies
selectively captures the target antigens, which are then detected
after the directed uptake of gold nanoparticles labeled with both
tracer antibodies and intrinsically strong Raman scatters (i.e.,
Raman reporter molecules (RRMs)). Multiple analytes can there-
fore be determined by immobilizing a mixture of different capture
antibodies and using nanoparticles coated with different sets of
RRMs and tracer antibodies.
Scheme 2
The added advantages in this strategy reflect the use of gold
colloids in label development. First, gold is a strongly enhancing
surface at long-wavelength excitation. Readout is therefore less
susceptible to interference by native fluorescence. Second, each
nanoparticle is coated with a large number of RRMs (103-104).
As such, the response of an individual binding event is markedly
amplified. Third, gold surfaces are readily modified by thiols and
disulfides.38 This chemistry provides a versatile platform for label
construction.
Herein, we discuss advances in our labeling and detection
strategy that yield a significant improvement in detection capabili-
ties. Our earlier procedure16 (Scheme 2A) prepared tracer nano-
particles by the physisorption of antibodies on gold colloids that
had been previously coated with a partial monolayer of RRMs
based on aromatic thiols. However, the exchange of physisorbed
(28) Nie, S.; Emory, S. R. Science (Washington, D. C.) 1 9 9 7 , 275, 1102-6.
(29) Kneipp, K.; Wang, Y.; Kneipp, H.; Perelman, L. T.; Itzkan, I.; Dasari, R. R.;
Feld, M. S. Phys. Rev. Lett. 1 9 9 7 , 78, 1667-70.
antibodies between nanoparticles with different RRMs degraded
performance. The new procedure (Scheme 2B) addresses this
problem by producing particles coated with a thiolate-based
monolayer that has a terminal succinimide group. This group can
then react with the amines of a protein to form an amide linkage.
The revised procedure also increases RRM surface concentration
and employs monoclonal antibodies as biospecific elements,
amplifying the response and binding affinities, respectively. This
paper applies this strategy to the low-level detection of prostate-
specific antigen (PSA).
(30) Kneipp, K.; Kneipp, H.; Itzkan, I.; Dasari, R. R.; Feld, M. S. Chem. Rev. 1999 ,
99, 2957-75.
(31) Freeman, R. G.; Grabar, K. C.; Allison, K. J.; Bright, R. M.; Davis, J. A.;
Guthrie, A. P.; Hommer, M. B.; Jackson, M. A.; Smith, P. C.; et al. Science
1 9 9 5 , 267, 1629-31.
(32) Keating, C. D.; Kovaleski, K. M.; Natan, M. J. J. Phys. Chem. B 1 9 9 8 , 102,
9404-13.
(33) Graham, D.; Mallinder, B. J.; Smith, W. E. Biopolymers 2 0 0 0 , 57, 85-91.
(34) Graham, D.; Mallinder, B. J.; Smith, W. E. Angew. Chem., Int. Ed. 2 0 0 0 ,
39, 1061-3.
(35) Graham, D.; Mallinder, B. J.; Whitcombe, D.; Smith, W. E. ChemPhysChem
2 0 0 1 , 2, 746-8.
(36) Cao, Y. C.; Jin, R.; Mirkin, C. A. Science 2 0 0 2 , 297, 1536-40.
(37) Vo-Dinh, T.; Stokes, D. L.; Griffin, G. D.; Volkan, M.; Kim, U. J.; Simon, M.
I. J. Raman Spectrosc. 1 9 9 9 , 30, 785-93.
Prostate cancer is the second leading cause of cancer-related
deaths in adult males in the United States. PSA, a 33-kDa
glycoprotein, has been used as a prostate cancer marker since
(38) Shipway, A. N.; Katz, E.; Willner, I. ChemPhysChem 2 0 0 0 , 1, 18-52.
Analytical Chemistry, Vol. 75, No. 21, November 1, 2003 5937

1988.39 In blood plasma, PSA exists in both complexed and free
forms, with normal levels of total PSA between 4 and 10 ng/ mL.39
The predominant form of serum PSA is a complex with R1-
antichymotrypsin, with lesser amounts bound with R2-macroglo-
bulin and R1-antitrypsin.40,41 Assays today distinguish between the
different forms of PSA through unreacted epitopes in each
complex. The distinction between complexed and free PSA is
clinically relevant because the occurrence probability of cancer
increases as the percentage of free PSA decreases.42,43 Moreover,
the reappearance of PSA after radical prostatectomy may signal
cancer reemergence.44,45 Early detection, however, is difficult
because the low levels of PSA present in early stages of recurrence
are below the detection capabilities of most assays.46 There is also
evidence that an ultrasensitve PSA assay may be of value in
detecting breast cancer in females.47 Hence, there is a clear need
for a simple, rapid, sensitive PSA detection method.48 We show
in this paper that SERS-based readout methods can detect PSA
in human serum at very low levels (∼1 pg/ mL). We also briefly
discuss the potential application of this method to multiplexed
assays.
Scheme 3
EXPERIMENTAL SECTION
Reagents. Suspensions of unconjugated colloidal gold (32.2
( 4.4-nm diameter, 2 × 1011 particles/ mL) were purchased from
Ted Pella, Inc. The matched pair of monoclonal antibodies utilized
for the sandwich assay was obtained from Research Diagnostics.
The pair consisted of mouse anti-human free PSA clone PSA-F65,
which was used as the capture antibody after immobilization on
gold-coated glass chips, and mouse anti-human PSA clone PSA-
66, which was employed as the tracer antibody after conjugation
to the gold particles (see below).
Serum PSA (10-30% free PSA) was purchased from Bios
Pacific, and buffer packs and ImmunoPure normal human serum
were acquired from Pierce Biotechnology. N-hydroxysuccinimide
(NHS), 1,3-dicyclohexylcarbodiimide (DCCD), Tween 80, 5,5′-
dithiobis(2-nitrobenzoic acid) (DNBA), and bovine serum albumin
(BSA) were obtained from Aldrich. Unless otherwise specified,
all reagents were used as received or were reconstituted according
to standard methodologies. The preparation of dithiobis(succin-
imide undecanoate) (DSU) followed a modification to recent
literature procedures.49,50
Synthesis of 5,5 ′-Dithiobis(succinimidyl-2 -nitrobenzoate)
(DSNB). To 50 mL of dry tetrahydrofuran were added 0.50 g of
DNBA (1.3 mmol), 0.52 g of DCCD (2.5 mmol), and 0.29 g of
NHS (2.5 mmol) in a 100-mL round-bottom flask equipped with a
drying tube. The mixture was magnetically stirred at 25 °C for 12
h, filtered, and then rotoevaporated to remove solvent. The crude
product was recrystallized from acetone/ hexane, yielding a yellow
3
powder: 1H NMR (CDCl3) δ ) 8.13 (d, 2H, JH,H ) 8 Hz, C6H4),
3
(39) Polascik, T. J.; Oesterling, J. E.; Partin, A. W. J. Urol. 1 9 9 9 , 162, 293-306.
(40) Stenman, U. H.; Leinonen, J.; Alfthan, H.; Rannikko, S.; Tuhkanen, K.;
Alfthan, O. Cancer Res. 1 9 9 1 , 51, 222-6.
7.85 (d, 2H, JH,H ) 8 Hz, C6H4), 7.97 (s, 2H, C6H4), 2.91 (s, 8H,
CH2).
P reparation of Raman Reporter-Labeled Immunogold
Colloids. To build the new labels, we started with various
derivatives of dithiobis(benzoic acid), which could easily be
converted to the corresponding succinimide ester with NHS. Of
those tested, DSNB is a particularly attractive example because
of the strong scattering cross section of its symmetric NO2 stretch.
As such, treatment of colloidal gold with this derivative (Scheme
3) yields a coating of the thiolate of DSNB, which can couple to
the primary amines of a tracer antibody by formation of an amide
linkage.49 We add that this design strategy minimizes the distance
between the gold surface and label scattering center. This
minimization is particularly significant because, according to a
simplified electromagnetic model,51 enhancement varies inversely
(41) Lilja, H.; Christensson, A.; Dahlen, U.; Matikainen, M. T.; Nilsson, O.;
Pettersson, K.; Lovgren, T. Clin. Chem. 1 9 9 1 , 37, 1618-25.
(42) Luderer, A. A.; Chen, Y. T.; Soriano, T. F.; Kramp, W. J.; Carlson, G.; Cuny,
C.; Sharp, T.; Smith, W.; Petteway, J.; Brawer, M. K. Urology 1 9 9 5 , 46,
187-94.
(43) Catalona, W. J.; Smith, D. S.; Wolfert, R. L.; Wang, T. J.; Rittenhouse, H. G.;
Ratliff, T. L.; Nadler, R. B. JAMA, J. Am. Med. Assoc. 1 9 9 5 , 274, 1214-20.
(44) Ellis, W. J.; Vessella, R. L.; Noteboom, J. L.; Lange, P. H.; Wolfert, R. L.;
Rittenhouse, H. G. Urology 1 9 9 7 , 50, 573-9.
(45) Yu, H.; Diamandis, E. P.; Wong, P. Y.; Nam, R.; Trachtenberg, J. J. Urol.
1 9 9 7 , 157, 913-8.
(46) Ward, A. M.; Catto, J. W. F.; Hamdy, F. C. Ann. Clin. Biochem. 2 0 0 1 , 38,
633-51.
(47) Melegos, D. N.; Diamandis, E. P. Clin. Biochem. 1 9 9 6 , 29, 193-200.
(48) Black, M. H.; Grass, C. L.; Leinonen, J.; Stenman, U. H.; Diamandis, E. P.
Clin. Chem. 1 9 9 9 , 45, 347-54.
(49) Wagner, P.; Hegner, M.; Kernen, P.; Zaugg, F.; Semenza, G. Biophys. J.
1 9 9 6 , 70, 2052-66.
(50) Niemz, A.; Jeoung, E.; Boal, A. K.; Deans, R.; Rotello, V. M. Langmuir 2 0 0 0 ,
16, 1460-2.
(51) Kneipp, K.; Kneipp, H.; Itzkan, I.; Dasari, R. R.; Feld, M. S. J. Phys.: Condens.
Matter 2 0 0 2 , 14, R597-624.
5938 Analytical Chemistry, Vol. 75, No. 21, November 1, 2003

with the 12th power of the separation distance between the
scatterer and the metal particle center.
The particle workup consisted of two steps. In step 1, 100 µL
of a 2.5 mM DSNB solution in acetonitrile was added to 1 mL of
the unconjugated colloidal gold suspension and the mixture
reacted for 3-5 h. The reporter-labeled colloids were then
separated from solution by centrifugation at 10000g for 7 min. The
clear supernatant was discarded, and the loose red sediment was
resuspended in 1 mL of borate buffer (2 mM, pH 9).
In step 2, mouse anti-PSA was coupled to the gold particles
via the succinimidyl terminus of the DSNB-derived coating. As
such, 35 µg of detection antibody (7 µL of 5 mg/ mL PSA-66
solution) was added to the 1-mL suspension of the reporter-labeled
colloid. The mixture was then incubated at room temperature for
1 h. After centrifugation at 10000g for 7 min and removal of the
supernatant, the red sediment was resuspended in 1 mL of 2 mM
Tris buffer (Tris-HCl (pH 7.6), 1% BSA). We note that the use of
BSA, Tween 80, or both in all of the preparative steps and in the
assay protocol is part of a general procedure designed to minimize
complications from nonspecific adsorption.
Figure 1. Experimental setup for measuring PSA levels in human
serum using surface-enhanced Raman spectroscopy.
P reparation of Capture Antibody Substrates. Glass slides
were cleaned in an ultrasonic bath under dilute surfactant solution
(Micro, Cole-Parmer), deionized water, and methanol, each for
30 min. The slides were then loaded into an Edwards 306A metal
evaporator and coated with 15 nm of chromium and 300 nm of
gold at 0.2 nm/ s at pressures less than 5 × 10-6 Torr. Next, the
gold substrates were removed from the evaporator and exposed
for ∼30 s to an octadecanethiol (ODT)-soaked poly(dimethylsil-
oxane) stamp, which had a 5-mm-diameter hole cut in its center.
This step “inks” the outer portion of the gold substrate with a
monolayer of ODT.52 After inking, the substrates were rinsed with
ethanol, dried under a stream of nitrogen, and immersed in a 1
mM ethanolic solution of DSU for 6-12 h. Upon removal from
solution, the substrates were rinsed again with ethanol and dried
under a stream of nitrogen. The result is a 5-mm-diameter domain
of the succinimide ester-terminated monolayer on each substrate,
surrounded by a hydrophobic ODT coating. We note that the ODT
coating serves as a hydrophobic barrier that localizes aqueous
protein solutions when pipetted onto the area of the substrate
defined by the DSU-derived monolayer.
Anti-free PSA antibodies (PSA-65) were immobilized by pipet-
ting 40 µL of the protein solution (100 µg/ mL in 0.05 M borate
buffer (pH 9) and 1% Tween 80) onto the localized domain of the
DSU-modified monolayer. The reaction was allowed to progress
overnight at room temperature. After rinsing three times with
buffer 1 (0.01 M borate buffer (pH 9), 30 mM NaCl, 0.5% Tween
80), 40 µL of blocking buffer (5% BSA in 0.05 M borate buffer
(pH 9)) was pipetted onto the surface and incubated for 1 h. The
substrates were then rinsed three times with buffer 1.
onto a capture antibody-coated substrate and allowed to react for
3 h at room temperature. After rinsing three times with buffer 2
(10 mM PBS buffer (pH 7.5), 0.5% Tween 80, 0.02% NaN3), the
sample was exposed to 40 µL of the immunogold detection reagent
for 6 h. All samples were then rinsed three times with buffer 2
and once with deionized water and dried under a stream of
nitrogen before SERS characterization. We found that Tween
concentrations of 0.1-0.5% in the rinse buffer were generally
effective in minimizing nonspecifically bound protein, while
maintaining the hydrophobic integrity of the ODT domain.
Instrumentation. (i) SERS Measurements. Figure 1 shows
the spectroscopic setup. A fiber-optic-based Raman system, Nano-
Raman I, from NanoRaman Instruments was used for all Raman
data generation. The system consists of three major subas-
semblies: laser light source, spectrograph, and fiber-optic probe.
The light source is a 30-mW, 632.8-nm HeNe laser, while the
spectrograph consists of an f/ 2.0 Czerny-Turner imaging spec-
trometer (6-8-cm-1 resolution, no moving parts) and a thermo-
electrically cooled (0 °C) Kodak 0401E CCD. The fiber-optic probe
(1.75 × 2.5 × 6 in) utilizes band-pass and long-pass filters for laser
light (OD 6) and fiber background (OD 4) rejection. The probe
objective provides a numerical aperture of 0.65 while maintaining
a relatively long working distance of 3 mm. The laser spot size
on the sample surface is ∼22 µm in diameter. A Windows-based
Visual Basic program controls the system. All spectra were
collected with a 60-s integration time. The positions of the Raman
bands were determined by comparisons to the known positions
of bands for solid naphthalene.53
Immunoassay P rotocol. Free PSA dose-response curves
were constructed using matrixes consisting of normal human
serum, 10 mM phosphate-buffered saline (PBS, KH2PO4/ K2HPO4
(pH 7.5), 150 mM NaCl, 0.1% BSA, 0.5% Tween 80, 0.02% NaN3),
and a 1:1 mixture of human serum and PBS, following the typical
procedure for a sandwich-type assay. For each matrix, 40-µL
aliquots of PSA solutions of various concentrations were pipetted
(ii) Infrared Spectroscopy. Infrared reflection spectra were
acquired with a Nicolet 850 FT-IR spectrometer, purged with liquid
N2 boil-off, and equipped with a liquid N2-cooled HgCdTe detector.
Spectra were obtained using p-polarized light incident at 80° with
respect to the surface normal. The spectra were recorded as -log-
(R/ R0), where R is the sample reflectance and R0 is the reflectance
of an octadecanethiolate-d37 monolayer-coated Au reference. The
(52) Chen, C. S.; Mrksich, M.; Huang, S.; Whitesides, G. M.; Ingber, D. E.
(53) McCreery, R. L. http:/ / chemistry.ohio-state.edu/ ˜ rmccreer/ freqcorr/ im-
ages/ nap.html.
Biotechnol. Prog. 1 9 9 8 , 14, 356-63.
Analytical Chemistry, Vol. 75, No. 21, November 1, 2003 5939

(amide I), and 1540 cm-1 (amide II), have appeared. The new
bands reflect the presence of amides inherent in the native
antibody as well as those formed by the reaction of the succin-
imidyl groups of DSU with amines in the protein. Coupled with
earlier reports,26,49,55 which, in part, studied the hydrolysis rate of
the succinimidyl terminal group of DSU-derived monolayer under
similar conditions,26 the differences in Figure 2 support the
covalent attachment of anti-free PSA to the underlying substrate.
The XPS characterizations are in agreement with the IRS
findings. The results for both modified surfaces are given in Table
1. As expected, survey spectra for the two different coatings
showed only the presence of carbon, oxygen, nitrogen, and sulfur.
For the DSU-based coating, the C(1s) region was composed of a
lower energy band (284.4 eV), attributed to the alkyl chain
structure of the coating, and a higher energy band (289.0 eV),
assigned to the different types of carbonyl carbon.56 Two bands
were also observed in the O(1s) and S(2p) regions. In the O(1s)
region, the band at 534.9 eV is ascribed to the oxygen of the ester
linkage and that at 532.5 eV is assigned to the remaining carbonyl
oxygens.56,57 In the S(2p) region, the positions of the features in
the doublet that arises from spin-orbit coupling (2p3/ 2, 161.8 eV;
2p1/ 2, 162.9 eV) are consistent with the presence of gold-bound
thiolates formed by the adsorption of thiols and disulfides on
gold.58,59 In contrast, there was only one band observed in the
N(1s) region. The location of this band (401.9 eV) agrees with
the presence of an electron-withdrawing group attached to
nitrogen, as is the case for the succinimidyl nitrogen.56
After treatment with anti-free PSA, all XPS features undergo a
general broadening, which limited the ability to carry out an in-
depth compositional analysis. There were, however, two readily
identifiable changes that support the coupling of anti-free PSA to
the DSU-derived coating. First, the N(1s) band shifts from 401.9
to 400.5 eV. This shift is ascribed to the loss of the succinimidyl
end group and the appearance of the numerous nitrogen func-
tionalities in the immobilized protein.60 Second, the intensity of
the S(2p) couplet is greatly diminished. The decrease in intensity
reflects the attenuation of photoelectrons by the immobilized
protein.
Figure 2. Infrared reflection spectra of a DSU-derived monolayer
on gold before (spectrum A) and after (spectrum B) exposure to the
anti-free PSA capture antibody.
spectra are an average of 512 sample and reference scans, taken
at 4-cm-1 resolution with Happ-Genzel apodization.
(iii) X-ray P hotoelectron Spectroscopy. X-ray photoelectron
spectra (XPS) were acquired at room temperature with a Physical
Electronics Industries 5500 multitechnique surface analysis sys-
tem. This system is equipped with a hemispherical analyzer, a
toroidal monochromator, a multichannel detector at 45°, and
monochromatic Al KR excitation radiation (1486.6 eV, 250 W). A
pass energy of 29.35 eV was used, giving a half-width of the
Au(4f7/ 2) peak of ∼0.8 eV.
RESULTS AND DISCUSSION
Chip Characterization. The capture antibody substrate con-
sisted of anti-free PSA bound to a gold-coated glass chip via the
DSU-derived coupling agent. DSU chemisorbs to gold through
cleavage of the sulfur-sulfur bond, and the formation of the
resulting gold-bound thiolate and its subsequent coupling to anti-
free PSA can be readily confirmed by infrared reflection spec-
troscopy (IRS) and XPS. The IRS results are presented in Figure
2. The three bands around 1800 cm-1 in the spectrum of the layer
formed from DSU (Figure 2A) are assigned to the carbonyl
stretches of the ester (1816 cm-1) and of the succinimidyl end
group (1787 (in-phase) and 1750 cm-1 (out-of-phase)).54 The
presence of these bands, along with the succinimidyl bands at
1219 and 1078 cm-1 and the methylene stretches between 3000
and 2800 cm-1, verifies the formation of the DSU-based coating.
IRS was also used to confirm the covalent binding of anti-free
PSA to the terminal group of the gold-bound coupling layer
(Figure 2B). Since the acyl carbon of the succinimidyl ester group
is strongly susceptible to nucleophilic attack, reaction with the
sterically accessible amines in the protein should immobilize anti-
free PSA via amide linkages. As evident in Figure 2B, treatment
of the DSU-modified substrate with anti-free PSA causes a marked
decrease in the magnitude of the bands for the succinimidyl group
(e.g., 1750, 1219, and 1078 cm-1). Moreover, three readily
identifiable bands, which are located at 3304 (N-H stretch), 1654
Detection Reagent Characterization. The new RRMs, which
were prepared by reacting DNBA with NHS to form the bis-
(succinimide ester), yield a coating on gold that can act as a
coupling agent in the same manner as the DSU-based monolayer.
Figure 3 shows the IRS spectra for a monolayer of DSNB
spontaneously adsorbed on gold-coated glass before and after
exposure to anti-PSA. The as-formed layer has carbonyl stretches
at 1812, 1789, and 1748 cm-1 and strong symmetric and asym-
metric nitro stretches at 1343 and 1533 cm-1, respectively. As with
the DSU-based monolayer, the spectrum for the DSNB-derived
monolayer undergoes a similar set of changes following exposure
to anti-PSA. Three new bands appear (3292 (N-H stretch), 1665
(55) Duhachek, S. D.; Kenseth, J. R.; Casale, G. P.; Small, G. J.; Porter, M. D.;
Jankowiak, R. Anal. Chem. 2 0 0 0 , 72, 3709-16.
(56) Delamarche, E.; Sundarababu, G.; Biebuyck, H.; Michel, B.; Gerber, C.;
Sigrist, H.; Wolf, H.; Ringsdorf, H.; Xanthopoulos, N.; Mathieu, H. J.
Langmuir 1 9 9 6 , 12, 1997-2006.
(57) Zaugg, F. G.; Spencer, N. D.; Wagner, P.; Kernen, P.; Vinckier, A.; Groscurth,
P.; Semenza, G. J. Mater. Sci.: Mater. Med. 1 9 9 9 , 10, 255-63.
(58) Bain, C. D.; Biebuyck, H. A.; Whitesides, G. M. Langmuir 1 9 8 9 , 5, 723-7.
(59) Biebuyck, H. A.; Whitesides, G. M. Langmuir 1 9 9 3 , 9, 1766-70.
(60) Jiang, L.; Glidle, A.; Griffith, A.; McNeil, C. J.; Cooper, J. M. Bioelectrochem.
Bioenerg. 1 9 9 7 , 42, 15-23.
(54) Frey, B. L.; Corn, R. M. Anal. Chem. 1 9 9 6 , 68, 3187-93.
5940 Analytical Chemistry, Vol. 75, No. 21, November 1, 2003

Table 1. Binding Energies (eV) and Compositional Assignments for XPS Spectra of DSU and DSNB Monolayers on
Gold before and after Antibody Derivitization
core level
DSU/ Au
anti-free PSA/ DSU/ Au
DSNB/ Au
anti-PSA/ DSNB/ Au
Au(4f7/ 2
)
83.9
161.8
162.9
83.9
161.8
162.9
83.9
162.2
83.9
162.2
nd
S(2p3/ 2
S(2p1/ 2
C(1s)
)
)
nda
289.0
288.5
288.4
288.1
C(1s)
N(1s)
O(1s)
284.4
401.9
532.5, 534.9
284.9, 286.4 (sh)
400.5
284.6
405.6, 401.2
532.2
284.8, 285.9 (sh)
405.2, 400.0
532.2
532.1
a
nd, not detected.
Figure 4. Raman spectra of the reporter compound: (A) Spectrum
of DSNB powder; (B) SERS spectrum of gold nanoparticles following
reaction with DSNB.
Figure 3. Infrared reflection spectra of a DSNB-derived monolayer
on gold before (spectrum A) and after (spectrum B) exposure to the
anti-PSA tracer antibody.
band at 851 cm-1 to the nitro scissoring vibration. The band at
1566 cm-1 is assigned to an aromatic ring mode (8a),61 and the
large band at 1079 cm-1 is probably a succinimidyl N-C-O
stretch overlapping with aromatic ring modes.
(amide I), and 1530 cm-1 (amide II)), and the three carbonyl
stretches exhibit a notable decrease. The detection of residual
succinimide groups is expected because the presence of an
immobilized antibody will sterically hinder protein binding to
neighboring succinimide moieties. Thus, BSA was used in the
immunogold reagent preparation procedure to react with any
exposed residual succinimide groups in order to preclude their
participation in nonspecific binding.
The XPS characterizations (Table 1) also strongly mimic those
for the DSU-based layer. In this case, however, there are two N(1s)
bands for each sample. For the as-formed layer, bands at 401.2
and 405.5 eV are indicative of the succinimidyl nitrogen and nitro
nitrogen on the aromatic ring, respectively. After exposure to anti-
PSA, the band at 401.2 eV disappears and one at 400.0 eV appears.
This change again parallels that for the DSU-derived coating.
SERS of Reporter-Labeled Immunogold Reagent. Raman
spectra for the RRM DSNB are shown in Figure 4 before and
after coupling to the gold nanoparticles. The nanoparticle sample
was prepared by drop casting a small amount of the labeled colloid
solution onto a gold-coated glass slide and evaporating the water-
based solvent. The powder spectrum (Figure 4A) is dominated
by the symmetric nitro stretch at 1342 cm-1, and we attribute the
Many of the bands in the powder spectrum are present in the
spectrum after DSNB is chemisorbed onto the gold nanoparticles
(Figure 4B), though some have undergone a small change in
position. For example, the symmetric nitro stretch has shifted from
1342 to 1338 cm-1, and the 8a mode has moved from 1566 to 1558
cm-1. These shifts are indicative of interactions between neighbor-
ing adsorbates and between the adsorbates and the gold surface.
Additionally, since there was no detectable Raman signal for the
monolayer formed from DSNB adsorbed to a smooth gold surface,
the spectrum in Figure 4B illustrates that the immobilization of
DSNB on the gold nanoparticles results in a significant level of
enhancement, the magnitude of which will be examined in detail
in future studies. Together, these results confirm that the particles
have been effectively modified with the DSNB-based RRMs.
SERS Immunoassay Detection of Free P SA. The results
of our SERS-based determinations for free PSA in normal human
serum are shown in Figure 5. Test solutions were made by serial
dilution in human serum of a 1 mg/ mL PSA standard to cover
the range from 1 µg/ mL (30 nM) to 1 pg/ mL (30 fM). The spectra
(61) Varsanyi, G. Assignments for Vibrational Spectra of Seven Hundred Benzene
Derivatives; John Wiley & Sons: New York, 1974.
Analytical Chemistry, Vol. 75, No. 21, November 1, 2003 5941

localized concentrations of binding sites and, therefore, higher
particle concentrations. Images of the surface using atomic force
microscopy (AFM), however, argue that particle densities are
reasonably homogeneous over areas irradiated by the laser source.
On the other hand, AFM imaging revealed a small number of
particle aggregates that could account for the hot spots. A third
possibility arises from the existence of “hot particles”. Recent
studies have shown that enhancement factors are strongly
dependent on particle size, shape, and excitation wavelength and
that a small fraction of particles exhibit markedly larger enhance-
ments.28,29 We are currently designing experiments to better
understand the origin of the hot spots and potentially utilize them
in future assays.
The dose-response curve shows that our detection platform
can determine free PSA at very low concentrations in human
serum. The detection limit is ∼1 pg/ mL.62 These results compare
favorably with commercial assays based on radiometric, chemi-
luminescent, and ELISA methods, which have detection limits
ranging from 3 to 1000 pg/ mL free PSA.46 Additional studies were
performed in which PSA was added to 10 mM PBS that contained
0.1% BSA and 0.5% Tween 80. These studies yielded a detection
limit of 4 pg/ mL free PSA, based on a concentration that produces
a signal three times the standard deviation of the background.
Similar results were obtained in an analyte matrix of a 1:1 mixture
of PBS/ serum. Thus, the assay appears applicable to a range of
sample matrixes.
The ability to detect exceedingly small amounts of analyte
using our monoclonal-based assay format is underscored by a
rough estimate of the number of molecular recognition events
responsible for the response at the limit of detection. At a detection
limit of 4 pg/ mL, a 40-µL solution of free PSA contains 160 fg
(∼3 × 106 molecules) of free PSA. If we assume that (1) the
capture surface exhaustively binds all of the proteins, (2) the
captured antigens are uniformly distributed across the 5-mm-
diameter surface of the capture substrate, and (3) the binding
stoichiometry between the nanoparticles and captured antigen is
1:1, then there are only ∼60 PSA molecules in the 22-µm-diameter
area irradiated by the laser source. This analysis shows that the
combination of surface enhancement with respect to the close
proximity of the scattering site to the particle surface, the
amplification due to the large number of RRMs coating each
particle (preliminary estimates indicate that there are ∼103 RRMs
tethered to each nanoparticle), and the binding affinity of mono-
clonal antibodies leads to an extremely low level of detection.63
Based on the estimate of the number of recognition events
detected in our PSA assay, projections can be made which strongly
Figure 5. Demonstration of a SERS-based free PSA immunoassay.
(A) SERS spectra, offset for clarity, acquired at various PSA
concentrations. (B) Dose-response curve for free PSA in human
serum. The dose-response curve was constructed by calculating the
average reading of the response for 6-8 different locations on the
surface of each sample, which typically varied by 10% (see text for
further details).
in Figure 5A were obtained using 60-s integrations after comple-
tion of the immunoassay protocol outlined above. As is evident,
the features diagnostic of the DSNB-labeled nanoparticles exhibit
a strong increase as the PSA level increases. These changes span
more than 6 orders of magnitude, thus encompassing concentra-
tion levels critical to prostate cancer diagnosis.39
The lower limit of detection is defined by the nonspecific
adsorption of the labeled nanoparticles, as demonstrated by the
signal observed for the blank serum sample. Blanks prepared
without BSA and Tween 80 as additives yielded signals that were
several times larger than those obtained with the use of additives.
In contrast, the packing constraints imposed by the labeled particle
size should control the upper limit of the dynamic range. Though
only examined in a preliminary manner, tests place the upper limit
at ∼10 µg/ mL.
(62) Recall that the capture antibody was anti-free PSA while the antigen was
serum PSA containing only 10-30% free PSA. The projected detection limit
estimates are, however, based on the total PSA concentration of the antigen
standard and therefore the detection limits for free PSA should be somewhat
lower.
(63) Our past work16 reported a limit of detection of ∼30 ng/ mL for an assay
using polyclonal antibodies. The ∼30000× improvement described herein
reflects a combination of several factors. These factors include a doubling
in the label surface concentration on each particle, a larger Raman scattering
cross section for the new label (∼5×), and the use of monoclonal vs
polyclonal antibodies (∼1000× higher affinity). We have also employed a
different Raman spectrometer. This instrument uses a slightly shorter
excitation wavelength for greater enhancement and has an improved
collection efficiency (>4× improvement in signal-to-noise ratio). While only
rough estimates, it is the combination of these improvements that lead to
the markedly lower limit of detection in our PSA assay.
A more detailed treatment of our findings is presented by the
dose-response curve in Figure 5B. This curve was constructed
by plotting the scattering intensity of the symmetric nitro stretch
(1338 cm-1, full width at half-maximum of 22 cm-1). Each data
point represents the average of six to eight readings across the
sample surface. Variations in signal strength across the surface
of each chip were typically ∼10%. However, signal strengths up
to twice as large as those represented in the plot were observed
∼20% of the time. The dose-response curve was constructed by
omitting the data for these “hot spots”. These hot spots could
possibly reflect the presence of domains where there are higher
5942 Analytical Chemistry, Vol. 75, No. 21, November 1, 2003

argue that the technique can be extended to single-molecule
detection. There are two clear avenues to reach such a level. The
first avenue uses labels that undergo both resonance and surface
enhancement. With resonance enhancement, intensities can be
2-6 orders of magnitude greater than those based on normal
Raman scattering.64 The second avenue takes advantage of recent
reports that have shown that the surface enhancement for slightly
larger gold particles (e.g., 60 nm for our excitation wavelength)
is greater than that for 30-nm particles.65 Taken together, the
ability to detect the binding of a single antigen appears to be well
within reach and should be of immense value in the ultra-low-
level detection of a wide range of biomarkers used in early disease
diagnosis and other assay applications. Low-level detection be-
comes even more important as the degree of multiplexing
increases, e.g., in instances where screening for multiple analytes
at a single address is of interest.
is potentially capable of encompassing a wide range of applications,
especially in view of the opportunities to multiplex through the
judicious design of more labeled nanoparticles. As such, multiple
analytes could be concurrently identified by the position of a
characteristic feature of the Raman label and then quantified by
its intensity. Assays could therefore be developed for the high-
sensitivity, simultaneous screening of a battery of cancer markers
using a single serum sample, saving time, reducing assay costs,
and potentially leading to earlier diagnosis. Experiments are
underway to test this concept by developing an assay for the
concurrent determination of free and total PSA, a ratio shown to
improve the reliability of prostate cancer diagnosis with respect
to those that rely on interpretations from the levels for one
marker.42,43
ACKNOWLEDGMENT
This work was supported by Concurrent Analytical, Inc.,
through a grant from the DARPA CEROS program and by the
Microanalytical Instrumentation Center of Iowa State University.
The Ames Laboratory is operated for the U.S. Department of
Energy by Iowa State University under contract W-7405-eng-82.
CONCLUSIONS
This report has shown that an important biomarker for early
cancer diagnosis can be detected in serum samples at very low
concentrations by a SERS-based readout method. This strategy
(64) Morris, M. D.; Wallan, D. J. Anal. Chem. 1 9 7 9 , 51, 182A-3A, 185A6A,
188A, 190A, 192A.
(65) Krug, J. T., II.; Wang, G. D.; Emory, S. R.; Nie, S. J. Am. Chem. Soc. 1 9 9 9 ,
121, 9208-9214.
Received for review April 7, 2003. Accepted August 25,
2003.
AC034356F
Analytical Chemistry, Vol. 75, No. 21, November 1, 2003 5943