Reverse-Phase versus Sandwich Antibody Microarray, Technical

Jul 3, 2007 - Michael Seidel , Reinhard Niessner. Analytical and ... Ulrike Korf , Christian Löbke , Florian Haller , Holger Sültmann , Annemarie Po...
2 downloads 0 Views 731KB Size
Download PDF
Anal. Chem. 2007, 79, 5817-5825

Reverse-Phase versus Sandwich Antibody Microarray, Technical Comparison from a Clinical Perspective K. Ja 1 rås,*,† A. Ressine,‡ E. Nilsson,§ J. Malm,† G. Marko-Varga,| H. Lilja,†,⊥ and T. Laurell‡

Department of Laboratory Medicine, Division of Clinical Chemistry, Lund University, Malmo¨ University Hospital, Malmo¨, Sweden, Department of Electrical Measurement, Lund University, Lund, Sweden, Department of Laboratory Medicine, Division of Tumour Biology, Lund University, Malmo¨ University Hospital, Malmo¨, Sweden, Department of Analytical Chemistry, Lund University and AstraZeneca R&D Lund, Lund, Sweden, and Departments of Clinical Laboratories, Surgery (Urology), and Medicine (GU-Oncology), Memorial Sloan-Kettering Cancer Center, New York, New York 10021

Protein microarrays are powerful tools to quantify and characterize proteins in multiplex assays. They have great potential within clinical diagnostics and prognostics, as they minimize consumption of both analyte and biological sample. Assays that do not require labeling of the biological specimen, henceforth called label-free, are vital for ease of clinical sample processing. Here, we evaluate two label-free techniques, reverse-phase and sandwich antibody assays, using microarrays on high-performance porous silicon surfaces and fluorescence detection. In view of increasing interest in reverse microarrays, this paper focuses on analytical sensitivity of the reverse assays compared to the more complex but highly sensitive sandwich assay. Sensitivity, linear range, and reproducibility of the two assays were compared using prostatespecific antigen (PSA) in buffer. The sandwich assay displayed 5 orders of magnitude lower detection limit (0.7 ng/mL) compared to the reverse assay (70 µg/mL). PSA at 50 nM (1.5 µg/mL) in cell lysates was detected by the sandwich assay but not by the reverse assay, demonstrating again a far lower detection limit for sandwich microarrays. In independent assay runs of PSA spiked in female serum, the sandwich assay had good linearity (R2 > 0.99) and reproducibility (coefficient of variation e15%), and the detection limit could be improved to 0.14 ng/mL. Without further signal amplification, the sandwich assay would be our choice for PSA analysis of clinical samples using a microarray technology platform. DNA microarrays have contributed tremendously to new biological information by allowing multiplex comparisons of gene * To whom correspondence should be addressed. E-mail: [email protected]. Fax: +46 46 222 45 27. † Department of Laboratory Medicine, Division of Clinical Chemistry, Lund University. ‡ Department of Electrical Measurement, Lund University. § Department of Laboratory Medicine, Division of Tumour Biology, Lund University. | Department of Analytical Chemistry, Lund University and AstraZeneca R&D Lund. ⊥ Memorial Sloan-Kettering Cancer Center. 10.1021/ac0709955 CCC: $37.00 Published on Web 07/03/2007

© 2007 American Chemical Society

expression. However, as known from gene expression studies, mRNA levels and protein levels do not necessarily correlate.1-3 In addition, protein functionality is affected by posttranslational modifications and protein interactions. Analogous to the DNA microarrays, protein microarrays now form the basis of sensitive and multiplex miniaturized assays and constitute a powerful tool for quantification and characterization of proteins.4-7 Protein microarrays have considerable potential within clinical diagnostics and prognostics, because the consumption of biological sample can be reduced. The impact of protein microarrays was already foreseen more than 15 years ago by Ekins.8,9 Only recently, however, have protein microarrays made significant inroads into proteomics research, diagnostics, and drug discovery. Three common ways of detecting protein biomarkers are direct labeling of the sample analyte, sandwich antibody arrays, and reverse-phase antibody arrays. Direct labeling with two different fluorophores is a useful way of analyzing up- and downregulation of proteins in two different samples.10-15 Biotinylation of diluted (1) Gygi, S. P.; Rochon, Y.; Franza, B. R.; Aebersold, R. Mol. Cell. Biol. 1999, 19, 1720-1730. (2) Linck, B.; Boknik, P.; Eschenhagen, T.; Muller, F. U.; Neumann, J.; Nose, M.; Jones, L. R.; Schmitz, W.; Scholz, H. Cardiovasc. Res. 1996, 31, 625632. (3) Anderson, N. L.; Anderson, N. G. Electrophoresis 1998, 19, 1853-1861. (4) Joos, T. O.; Stoll, D.; Templin, M. F. Curr. Opin. Chem. Biol. 2002, 6, 7680. (5) Zhu, H.; Bilgin, M.; Bangham, R.; Hall, D.; Casamayor, A.; Bertone, P.; Lan, N.; Jansen, R.; Bidlingmaier, S.; Houfek, T.; Mitchell, T.; Miller, P.; Dean, R. A.; Gerstein, M.; Snyder, M. Science 2001, 293, 2101-2105. (6) Templin, M. F.; Stoll, D.; Schrenk, M.; Traub, P. C.; Vohringer, C. F.; Joos, T. O. Trends Biotechnol. 2002, 20, 160-166. (7) MacBeath, G.; Schreiber, S. L. Science 2000, 289, 1760-1763. (8) Ekins, R. P. J. Pharm. Biomed. Anal. 1989, 7, 155-168. (9) Ekins, R.; Chu, F.; Biggart, E. Ann. Biol. Clin. 1990, 48, 655-666. (10) Miller, J. C.; Zhou, H. P.; Kwekel, J.; Cavallo, R.; Burke, J.; Butler, E. B.; Teh, B. S.; Haab, B. B. Proteomics 2003, 3, 56-63. (11) Zhou, H. P.; Bouwman, K.; Schotanus, M.; Verweij, C.; Marrero, J. A.; Dillon, D.; Costa, J.; Lizardi, P.; Haab, B. B. Genome Biol. 2004, 5. (12) Knezevic, V.; Leethanakul, C.; Bichsel, V. E.; Worth, J. M.; Prabhu, V. V.; Gutkind, J. S.; Liotta, L. A.; Munson, P. J.; Petricoin, E. F.; Krizman, D. B. Proteomics 2001, 1, 1271-1278. (13) Tannapfel, A.; Anhalt, K.; Hausermann, P.; Sommerer, F.; Benicke, M.; Uhlmann, D.; Witzigmann, H.; Hauss, J.; Wittekind, C. J. Pathol. 2003, 201, 238-249. (14) Hudelist, G.; Pacher-Zavisin, M.; Singer, C. F.; Holper, T.; Kubista, E.; Schreiber, M.; Manavi, M.; Bilban, M.; Czerwenka, K. Breast Cancer Res. Treat. 2004, 86, 281-291.

Analytical Chemistry, Vol. 79, No. 15, August 1, 2007 5817

Figure 1. Schematic illustration of the reverse-phase and sandwich antibody microarray protocols for detection of PSA. The reverse assay (A) entails dispensing of the biological specimen (1), subsequent washing followed by incubation with FITC-labeled 2E9 monoclonal anti-PSA antibody (2), and a final washing step before readout with a fluorescence microscope (3). In the sandwich assay (B), the H117 monoclonal anti-PSA antibody is dispensed (1), followed by washing. The chip is subsequently incubated with the biological sample (2), washed, and finally incubated with the same FITC-labeled antibody, 2E9 (3), followed by microscope readout (4).

serum samples17,18 and Cy dye labeling of fractionated cell culture supernatant19 have successfully been reported. However, bulk sample labeling is a challenging and time-consuming task, and in clinical diagnostics, a quick high-throughput assay is preferable. Assays that do not involve labeling of the biological specimen we refer to as label-free. Reverse-phase antibody arrays (shown schematically in Figure 1A) constitute a very convenient and label-free way of analyzing biological samples by simply arraying the specimen and detecting the biomarkers with an antibody, either by fluorescent labeling of the antibody or by catalyzed signal amplification and colorimetric readout.20-24 In addition, the technique has the potential for detecting more than one biomarker in the same sample, simply by using multiple monospecific antibodies labeled with different fluorophores.25 Sandwich immunoassays (Figure 1B) are another label-free approach, which offer the advantage of very high sensitivity. Sandwich assays are widely used for diagnostics, frequently in (15) Lin, Y.; Huang, R. C.; Cao, X.; Wang, S. M.; Shi, Q.; Huang, R. P. Clin. Chem. Lab. Med. 2003, 41, 139-145. (16) Sreekumar, A.; Nyati, M. K.; Varambally, S.; Barrette, T. R.; Ghosh, D.; Lawrence, T. S.; Chinnaiyan, A. M. Cancer Res. 2001, 61, 7585-7593. (17) Wingren, C., I. J.; Dexlin, L.; Szul, D.; Borrebaeck, C. Unpublished work, Department of Immunotechnology, Lund Institute of Technology, 2006. (18) Ingvarsson, J., L. A.; Sjo¨holm, A.; Truedsson, L.; Jansson, B.; Borrebaeck, C.; Wingren, C. Unpublished work, Department of Immunotechnology, Lund Institute of Technology, 2006. (19) Ingvarsson, J.; Lindstedt, M.; Borrebaeck, C. A. K.; Wingren, C. J. Proteome Res. 2006, 5, 170-176. (20) Espina, V.; Woodhouse, E. C.; Wulfkuhle, J.; Asmussen, H. D.; Petricoin, E. F.; Liotta, L. A. J. Immunol. Methods 2004, 290, 121-133. (21) Nishizuka, S.; Charboneau, L.; Young, L.; Major, S.; Reinhold, W. C.; Waltham, M.; Kouros-Mehr, H.; Bussey, K. J.; Lee, J. K.; Espina, V.; Munson, P. J.; Petricoin, E.; Liotta, L. A.; Weinstein, J. N. Proc. Natl. Acad. Sci. U.S.A. 2003, 100, 14229-14234. (22) Wulfkuhle, J. D.; Aquino, J. A.; Calvert, V. S.; Fishman, D. A.; Coukos, G.; Liotta, L. A.; Petricoin, E. F. Proteomics 2003, 3, 2085-2090. (23) Grubb, R. L.; Calvert, V. S.; Wulkuhle, J. D.; Paweletz, C. P.; Linehan, W. M.; Phillips, J. L.; Chuaqui, R.; Valasco, A.; Gillespie, J.; Emmert-Buck, M.; Liotta, L. A.; Petricoin, E. F. Proteomics 2003, 3, 2142-2146. (24) Paweletz, C. P.; Charboneau, L.; Bichsel, V. E.; Simone, N. L.; Chen, T.; Gillespie, J. W.; Emmert-Buck, M. R.; Roth, M. J.; Petricoin, E. F.; Liotta, L. A. Oncogene 2001, 20, 1981-1989. (25) Haab, B. B. Mol. Cell. Proteomics 2005, 4, 377-383.

5818

Analytical Chemistry, Vol. 79, No. 15, August 1, 2007

96-well formats. The translation of these assays into a miniaturized format is an attractive approach to minimize the consumption of sample and analyte. The most difficult step is obtaining the matched sandwich antibodies. Many studies have been performed on sandwich microarrays.16,26-33 MacBeath and Schweitzer et al.31,34 have reported that up to 40-50 different antibody pairs can be used on the same microarray, without cross-reactions. In the development of protocols for protein microarrays, much focus has been put on the capture molecules. As a complement to this, our group has previously emphasized the development of new high-performance surfaces for protein arraying. The substrates used in this work are our 3D porous silicon surfaces, developed in-house by electrochemical porosification of silicon wafers. These have proven to be highly compatible with protein chip technology, and they outperform most commercial surfaces with respect to spot quality,36 spot density (noncontact printing mode),36 linear range,37 and sensitivity.37 Interestingly, these surfaces are also compatible with matrix-assisted laser desorption/ ionization mass spectrometry, enabling a dual-readout of detection, i.e., fluorescence and mass spectrometry.38 (26) Gembitsky, D. S.; Lawlor, K.; Jacovina, A.; Yaneva, M.; Tempst, P. Mol. Cell. Proteomics 2004, 3, 1102-1118. (27) Saviranta, P.; Okon, R.; Brinker, A.; Warashina, M.; Eppinger, J.; Geierstanger, B. H. Clin. Chem. 2004, 50, 1907-1920. (28) Nielsen, U. B.; Geierstanger, B. H. J. Immunol. Methods 2004, 290, 107120. (29) Nielsen, U. B.; Cardone, M. H.; Sinskey, A. J.; MacBeath, G.; Sorger, P. K. Proc. Natl. Acad. Sci. U.S.A. 2003, 100, 9330-9335. (30) Wang, C. C.; Huang, R. P.; Sommer, M.; Lisoukov, H.; Huang, R. C.; Lin, Y.; Miller, T.; Burke, J. J. Proteome Res. 2002, 1, 337-343. (31) Schweitzer, B.; Roberts, S.; Grimwade, B.; Shao, W. P.; Wang, M. J.; Fu, Q.; Shu, Q. P.; Laroche, I.; Zhou, Z. M.; Tchernev, V. T.; Christiansen, J.; Velleca, M.; Kingsmore, S. F. Nat. Biotechnol. 2002, 20, 359-365. (32) Tam, S. W.; Wiese, R.; Lee, S.; Gilmore, J.; Kumble, K. D. J. Immunol. Methods 2002, 261, 157-165. (33) Huang, J. X.; Mehrens, D.; Wiese, R.; Lee, S.; Tam, S. W.; Daniel, S.; Gilmore, J.; Shi, M.; Lashkari, D. Clin. Chem. 2001, 47, 1912-1916. (34) MacBeath, G. Nat. Genet. 2002, 32, 526-532. (35) Schweitzer, B.; Kingsmore, S. F. Curr. Opin. Biotechnol. 2002, 13, 14-19. (36) Ressine, A.; Ekstrom, S.; Marko-Varga, G.; Laurell, T. Anal. Chem. 2003, 75, 6968-6974. (37) Steinhauer, C.; Ressine, A.; Marko-Varga, G.; Laurell, T.; Borrebaeck, C. A. K.; Wingren, C. Anal. Biochem. 2005, 341, 204-213.

The target gene product that we focused on in the current paper is widely recognized in prostate cancer diagnostics. Prostate cancer is the most common cancer in males in Western Europe and the United States and also a major cause of death. Kallikreinrelated peptidase 3 (KLK3) is more commonly known as prostatespecific antigen (PSA) and is widely used as a serum biomarker to detect prostate cancer and to assess response to treatment and recurring cancer in patients with established cancer.39 All cancers considered, PSA is the most valuable of all tumor markers. PSA is an androgen-regulated serine protease that is mainly expressed in the prostate epithelium. It is released into seminal plasma upon ejaculation with a concentration of ∼1 g/l. Its concentration in serum is ∼106 times lower. PSA exists in both zymogen (i.e., with a propeptide) and active (i.e., without propeptide) forms in semen.40 Active PSA is an N-glycosylated single-chain protein of 237 amino acids41,42 with a mass of ∼28 kDa.43 In vitro, proPSA can be converted to active PSA by human kallikrein 2 (hK2), trypsin, or prostase.44-47 These findings might suggest a connection between the proteins in vivo as well. The enzyme activity of PSA, mainly directed to semenogelin I and II and fibronectin,48-51 induces liquefaction of semen and subsequent release of motile spermatozoa. PSA is primarily inactivated by SERPINA5 (also called protein C inhibitor, PCI or PAI-3) in seminal plasma and by SERPINA3 (R1-antichymotrypsin or ACT) or R2-macroglobulin in blood.52 Earlier reports from our group demonstrated proof of principle for PSA detection by antibody microarrays using our highperformance porous silicon chips. However, that assay required direct labeling of the antigen, i.e., FITC labeling of PSA. In this work, we have compared two label-free approaches, reverse-phase and sandwich antibody microarrays, for the detection of PSA. Analytic limit of detection (LOD or analytical sensitivity), dynamic range, and spot reproducibility were analyzed on the porous chips. EXPERIMENTAL SECTION Proteins and Reagents. PSA in seminal fluid and recombinant proPSA, produced in insect cells as previously described,46 were purified with immunoaffinity chromatography using four mono(38) Finnskog, D.; Ressine, A.; Laurell, T.; Marko-Varga, G. J. Proteome Res. 2004, 3, 988-994. (39) Steuber, T.; Helo, P.; Lilja, H. World J. Urol. 2007, 25, 111-119. (40) Becker, C.; Noldus, J.; Diamandis, E.; Lilja, H. Crit. Rev. Clin. Lab. Sci. 2001, 38, 357-399. (41) Lundwall, A.; Lilja, H. FEBS Lett. 1987, 214, 317-322. (42) Schaller, J.; Akiyama, K.; Tsuda, R.; Hara, M.; Marti, T.; Rickli, E. E. Eur. J. Biochem. 1987, 170, 111-120. (43) Belanger, A.; Vanhalbeek, H.; Graves, H. C. B.; Grandbois, K.; Stamey, T. A.; Huang, L. H.; Poppe, I.; Labrie, F. Prostate 1995, 27, 187-197. (44) Takayama, T. K.; McMullen, B. A.; Nelson, P. S.; Matsumura, M.; Fujikawa, K. Biochemistry 2001, 40, 15341-15348. (45) Takayama, T. K.; Fujikawa, K.; Davie, E. W. J. Biol. Chem. 1997, 272, 21582-21588. (46) Lovgren, J.; Rajakoski, K.; Karp, M.; Lundwall, A.; Lilja, H. Biochem. Biophys. Res. Commun. 1997, 238, 549-555. (47) Kumar, A.; Mikolajczyk, S. D.; Goel, A. S.; Millar, L. S.; Saedi, M. S. Cancer Res. 1997, 57, 3111-3114. (48) Lilja, H.; Oldbring, J.; Rannevik, G.; Laurell, C. B. J. Clin. Invest. 1987, 80, 281-285. (49) Lilja, H.; Abrahamsson, P. A.; Lundwall, A. J. Biol. Chem. 1989, 264, 18941900. (50) Lilja, H.; Lundwall, A. Proc. Natl. Acad. Sci. U.S.A. 1992, 89, 4559-4563. (51) McGee, R. S.; Herr, J. C. Biol. Reprod. 1988, 39, 499-510. (52) Christensson, A.; Laurell, C. B.; Lilja, H. Eur. J. Biochem. 1990, 194, 755763.

clonal anti-PSA antibodies, 2E9, 5A10, 2C1, and 2H11, coupled to Affigel 10 (Bio-Rad, Hercules, CA). The eluted protein solution was further purified by gel filtration (Sephacryl S-200 HR, Pharmacia Biotech, Uppsala, Sweden), and size and purity were confirmed by SDS/PAGE and Western blot. Monoclonal antibodies 2E9 and H117 were produced and characterized as described.53,54 2E9 was labeled with fluorescein isothiocyanate (FITC) isomer I-celite (Sigma St. Louis, MO) and separated on a PD10 column (Amersham, Uppsala, Sweden). Protein Extraction from LNCaP Cells. LNCaP cells (an androgen receptor-positive human prostate cancer line) were purchased from American Type Culture Collection (Manassas, VA) and cultured in 5% CO2 at 37 °C. The 5 × 105 cells were seeded in 100-mm culture dishes and grown in phenol red-free RPMI medium (GIBCO, Paisley, UK) supplemented with 10% fetal bovine serum, 10 mM Hepes, 1 mM sodium pyruvate, 4.5 g/L glucose, and 2 mM L-glutamine. After 48 h, cells were rinsed once in phosphate-buffered saline (PBS), and medium was changed to DMEM (GIBCO) supplemented with 4% dextran-charcoalstripped fetal bovine serum (GIBCO) and synthetic androgen R1881 (0.1 or 1 nM) (Sigma). Cells were grown for an additional 48 h; thereafter, medium was collected for analysis, and cells were rinsed once in ice-cold PBS and lysed for 5 min in ice-cold lysis buffer (150 mM NaCl, 50 mM Tris-HCl pH 8, 1% NP-40, 0.1% SDS, and 1 µg/mL Trasolyl). Cells were detached from the plate with a cell scraper, disrupted by passing through a pipet, and centrifuged at 10000g for 5 min at 4 °C. The supernatant (cell lysate) was transferred to a new microcentrifuge tube and stored at -20 °C. Human Female Serum Spiked with PSA. Human female blood was prepared using BD Vacutainers (Bedfordshire, UK), aliquoted, and stored at -80 °C. The serum was spiked with purified proPSA before analysis. Porous Silicon Fabrication. The 3D porous silicon surfaces were obtained by electrochemical dissolution of silicon wafers. The liquid cell for electrochemical silicon etching and the fabrication setup are described in more detail elsewhere.55 We used a two-compartment electrochemical cell with sapphire glass (Melles Griot BV) on one side to allow for illumination during anodization. The cell was designed for 3-in. wafers and to give the wafer contact with the electrolyte on both sides. P-Type wafers with crystal orientation 100 were purchased from Topsil Semiconductor Materials A/S (Frederikssund, Denmark; resistivity 10-15 Ω cm) and from Addison Engineering Inc. (San Jose, CA; resistivity 6-8 Ω cm). Wafers were etched at constant current 2 mA/cm2. The back sides were illuminated throughout the anodization period using a 100-W halogen lamp (Osram, Germany) at a distance of 10 cm from the transparent window on the back of the electrochemical cell. The electrolyte was a 1:10 mixture by volume of 40% hydrogen fluoride and 99.8% dimethyl formamide. Reverse-Phase Immunoassay. The 100-pL droplets of LNCaP lysate or purified PSA in 10 mM PBS were spotted onto (53) Lilja, H.; Christensson, A.; Dahlen, U.; Matikainen, M. T.; Nilsson, O.; Pettersson, K.; Lovgren, T. Clin. Chem. 1991, 37, 1618-1625. (54) Petterson, K.; Vehnianen, M.; Viloma, S.; Makinen, M. L.; Manstala, P.; Lovgren, T. Clin. Chem. 1995, 41, S81-S81. (55) Drott, J.; Lindstrom, K.; Rosengren, L.; Laurell, T. J. Micromech. Microeng. 1997, 7, 14-23.

Analytical Chemistry, Vol. 79, No. 15, August 1, 2007

5819

Figure 2. (A) Scanning electron microscope images of the porous silicon surfaces: left) view from top, right) cross-section picture. The vertical microcylindrical walls are perforated by a nanoporous network, not visible in the image. (B) Spots and spot intensity profile from one of the sandwich antibody arrays. (C) Spots and spot intensity profile from one of the sandwich antibody arrays subjected to nonoptimal dipping, showing smearing effects.

the porous silicon chips by a piezoelectric microdispenser developed in-house. The proteins were allowed to bind the surface by physical adsorption. The microdispenser, earlier described in detail,56-58 was used to create the array with a spot-to-spot distance of 100 or 150 µm. To remove loosely bound material, the chips were washed in PBS-Tween (0.05% Tween 20 in 10 mM PBS): once in 10 mL and twice in 5 mL. The chips were blocked for 30 min in 25 µL of 5% nonfat dry milk in PBS-Tween to prevent unspecific binding. The chips were then washed 3 times in 5 mL of PBS-Tween and incubated with 25 µL of FITC-labeled 2E9 monoclonal mouse anti-PSA antibody for 3 h. The chips were subsequently washed 3 times in 5 mL of PBS-Tween, quickly dipped in distilled water, and dried by high-pressure air to remove liquid. Fluorescence detection was performed with a microscope BX51WI (Olympus) with a confocal addition, an oil immersion 20× objective, an ion laser IMA101010BOS (Melles Griot Laser Group) with excitation wavelength of 488 nm, and a Fluoview scanner unit (Fluoview, Olympus). Fluoview 300 software was used for image analysis. Sandwich Antibody Array. The 100-pL droplets of monoclonal mouse capture antibody H117 were spotted onto the silicon chips at a spot-to-spot distance of 100 or 150 µm. The proteins (56) Laurell, T.; Wallman, L.; Nilsson, J. J. Micromech. Microeng. 1999, 9, 369376. (57) Miliotis, T.; Kjellstro ¨m, S.; Nilsson, J.; Laurell, T. J. Mass Spectrom. 2000, 35, 369-377. (58) O ¨ nnerfjord, P.; Nilsson, J.; Wallman, L.; Laurell, T. Anal. Chem. 1998, 70, 4755-4760.

5820

Analytical Chemistry, Vol. 79, No. 15, August 1, 2007

were allowed to bind the surface by physical adsorption. To remove loosely bound material, the chips were washed in PBSTween: once in 10 mL and twice in 5 mL. The chips were blocked for 30 min in 25 µL of 5% nonfat dry milk in PBS-Tween. The chips were then washed 3 times in 5 mL of PBS-Tween and incubated with 25 µL of sample for 1 h. After additional washing steps, 25 µL of FITC-labeled 2E9 monoclonal mouse anti-PSAantibody was added, followed by 1-h incubation. The chips were washed 3 times in 5 mL of PBS-Tween, quickly dipped in distilled water, and dried by high-pressure air. Fluorescence detection was performed as described above. Mean Spot Intensity and Limit of Detection. The mean intensity of each spot was quantified using the Fluoview 300 and the circle method. Local background was measured in the same way and subtracted from the mean intensity of each spot, generating the mean spot intensities presented in the graphs and tables. The LOD was defined as the lowest detectable PSA concentration corresponding to a mean spot intensity at least two standard deviations above the mean intensity of the background, calculated from 16 background locations. DELFIA. As a reference assay, Prostatus PSA Free/Total DELFIA (Perkin-Elmer, Turku, Finland) was used to determine the total PSA concentration. This immunoassay is commercially available and based on the original report by Mitrunen et al.59 The assay is a solid-phase, time-resolved immunofluorometric method based on a sandwich technique, which makes use of three murine monoclonal antibodies. One antibody (H117) recognizing both free PSA and PSA bound to ACT is used to capture PSA. Two differently labeled detection antibodies are used, one europium-labeled antibody (5A10) binding only free PSA and one samarium-labeled antibody (H50) binding both free PSA and PSA-ACT complex with identical affinity. Time-resolved recording of the fluorescence intensity at the wavelengths characteristic for samarium and europium, respectively, has previously been demonstrated to be highly proportional and specific to the concentrations of free and total PSA in the sample.59 RESULTS AND DISCUSSION General Remarks. Sensitivity in a noncompetitive immunoassay, as described by Ekins, is maximized when the capture molecule concentration approaches infinity.60 In the reverse antibody array, the capture molecule is the biomarker to be measured, which in our model system is the PSA. The biomarker, attached to the surface by physical adsorption, is then detected by fluorescently labeled antibodies. In the sandwich antibody microarray, the capture molecule is an antibody adsorbed onto the surface at high concentrations. Once the sample is applied, the biomarker is concentrated onto the spots by the antibodies. The captured biomarker is subsequently detected with a fluorescently labeled antibody. Because the concentration of capture molecules is much higher in the sandwich assay, theory predicts that this assay will be more sensitive (i.e., have lower limit of detection) than the reverse-phase antibody array, at least in the absence of secondary signal amplification. However, since the reverse-phase immunoassays are gaining increased attention, we (59) Mitrunen, K.; Pettersson, K.; Piironen, T.; Bjork, T.; Lilja, H.; Lovgren, T. Clin. Chem. 1995, 41, 1115-1120. (60) Ekins, R. P. Clin. Chem. 1998, 44, 2015-2030.

Figure 3. (A) Titration of PSA in 10 mM PBS in the reverse-phase assay. Standard deviations were calculated from the 16 spots in the inset images. (B) Titration of PSA in 10 mM PBS in the sandwich antibody assay. Note that this figure covers PSA concentrations 1000-fold lower than those in (A). Standard deviations were calculated from the 16 spots in the inset images. The inset graph shows data for the lowest three concentrations plotted on a linear scale.

chose to compare the two label-free methods. To make the comparison as relevant as possible, we used identical reagents and highly similar assay protocols. Arraying and Washing of the Capture Molecule. Antibody or PSA solution was microarrayed onto 3D porous silicon (PS) surfaces by piezoelectric microdispensing (Figure 2A). Before the first blocking step, the arrayed chips were washed in PBS-Tween, to remove loosely bound proteins. The method of dipping the chips into the wash solution was, however, found to be of critical importance. Different dipping procedures were tested by dispensing FITC-labeled anti-rabbit IgG onto the PS chips and analyzing them before and after the different dipping procedures (data not shown). Sliding the side of the chip into the solution with a turning motion gave the best results. If the chip was simply submerged with no turning motion into the buffer solution, smearing effects became very evident (Figure 2B,C). Correct wetting of the chip surface is important, as this influences spot homogeneity and quality of readout and automated spot detection. Poor spot

qualities are frequently seen in the literature. To improve quantitative properties of antibody and protein microarrays, reproducible and homogeneous spot shapes and intensity profiles are needed. PSA Titration in Sandwich and Reverse Antibody Microarrays. To compare the detection sensitivity, dynamic range, and reproducibility of the two setups, titration curves of PSA in 10 mM PBS were generated (Figure 3A and B). In the reverse antibody microarray, this was achieved by dispensing the different PSA solutions, from low to high concentration, and cleaning the dispenser carefully in between. In the sandwich antibody microarray procedure, the same capture molecule concentration was dispensed onto all the chips, followed by incubation in PSA at various concentrations in 10 mM PBS. Limit of Detection. The titrations indicated that the limit of detection was 0.7 ng/mL for the sandwich assay and 70 µg/mL for the reverse assay, a difference of 5 orders of magnitude. This difference can be explained by the enrichment of PSA onto the capture molecule spots in the sandwich assay. This difference can Analytical Chemistry, Vol. 79, No. 15, August 1, 2007

5821

Figure 4. Microscope images, of 70 and 7 µg/mL PSA in PBS analyzed by reverse antibody microarrays on BSA-coated and non-BSA-coated porous silicon surfaces. If denaturation of the PSA was pronounced, the albumin would prevent PSA from denaturing and improve the signal intensity. Note the better detection of the noncoated chips, indicating no prominent PSA denaturation.

also be explained from principles in the following theoretical calculations. The law of mass action can describe binding events, in our case, between antigen (An) and antibody (Ab). A binding event is here referred to as [An‚Ab], where this denotes the concentration of antigen-antibody complex. The fractional occupancy (F) is the fraction of all antibodies that are bound to antigens. The total antibody concentration, [Ab]tot, could be written as the sum of [An‚Ab] and [Ab], where [Ab] denotes the concentration of unoccupied antibodies.

fractional occupancy )

[An‚Ab] [Ab‚An] ) [Ab]tot [Ab‚An] + [Ab]

A capture antibody, when exposed to an antigen solution, will bind the antigen to a fractional extent dependent on the affinity constant and the antigen concentration in solution. In all immunoassays, the signal intensity reflects the fractional occupancy of the capture molecules, which in turn is a measurement of the analyte concentration in solution. In this work, the proteins arrayed onto the surface are referred to as capture molecules, and proteins in solution as analytes. Assume a fractional occupancy, F, a capture molecule concentration, Ccapture-molecule (for reverse assay, PSA concentration ) 1 ng/mL or 3.3 × 10-11 mol/L and, for the sandwich assay, antibody (H117) concentration ) 3.1 × 10-6 mol/ L), and a volume deposited per spot, Vspot )100 pL. The PSA concentration (1 ng/mL) was chosen just under the frequently used diagnostic cut point of 4 ng/mL. The H117 concentration, 3.1 × 10-6 mol/L, was the one used in the experiments. The following estimations demonstrate a theoretical sensitivity difference between reverse and sandwich microarray assays. Reverse antibody microarray:

amount analyte/spot ) ccapturemolecule(PSA) × Vspot × F ) 3.3 × 10-11mol/L × 100 × 10-12L/spot × F ) 3.3 F × 10-21 mol/spot Sandwich antibody microarray:

amount analyte/spot ) ccapturemolecule(H117) × Vspot × F ) 3.1 × 10-6 mol/L × 100 × 10-12 L/spot × F ) 3.1 F × 10-16 mol/spot 5822

Analytical Chemistry, Vol. 79, No. 15, August 1, 2007

For the same PSA concentration and fractional occupancy, the sandwich assay has the potential of binding 5 orders of magnitude more analyte as compared to the reverse assay. However, the fractional occupancy of the capture molecules is probably higher in the reverse assay due to a higher concentration of analyte than in the sandwich assay. (In this model, the analyte is referred to as the FITC-labeled detection antibody in the reverse assay and PSA in the sandwich assay.) Nevertheless, to achieve maximum sensitivity, as many of the analyte molecules as possible should be captured from the analyte solution, and the sandwich assay fulfills this criterion to a greater extent. Another factor in the sensitivity of immunoassays is rebinding. The probability of an analyte molecule rebinding during a washing step increases with higher capture molecule concentration. This consideration also supports our finding greater sensitivity of the sandwich antibody array. As a control, female serum without PSA was used as sample in the sandwich assay. No signal was detectable (data not shown). To investigate whether the antigen deposited on the surface was affected by denaturation upon adsorption, we also analyzed biotinylated PSA in both reverse and sandwich mode in experiments where fluorescently labeled streptavidin was used for detection. Since biotin attached to PSA is not subject to denaturation, this experimental design could be used to illustrate the difference in LOD between reverse and sandwich mode, precluding any dependency of PSA denaturation at the surface. In this experiment, a murine monoclonal capture antibody (2E9) with slightly lower binding affinity was used in the sandwich microarray as compared to the H117 antibody. Three to four orders of magnitude difference in LOD was observed between the sandwich and the reverse antibody microarrays (data not shown), again illustrating the inherently lower LOD of the sandwich assay. To further examine whether denaturation importantly influences the LOD of our reverse antibody microarrays, porous silicon surfaces were coated by bovine serum albumin (BSA). The chips were dipped into a BSA solution (10 mg/mL in 10 mM PBS) and allowed to dry over night. The reverse assay was performed on the chips at two different PSA concentrations, 7 and 70 µg/mL. A PSA concentration of 7 µg/mL was earlier found to be a concentration below the LOD of reverse antibody microarrays on noncoated chips. We also studied a 10 times higher concentration,

70 µg/mL, i.e., the LOD established for the reverse assay. Noncoated chips were used as controls. Duplicate samples were performed for each concentration and chip type. Figure 4 shows that BSA coating does not improve the signal intensity. Only noncoated chips (70 µg/mL) gave detectable spots, indicating no pronounced PSA denaturation. In addition, a titration series of human serum albumin (0.015, 0.15, and 1.5 mg/mL) with a constant PSA concentration, 0.18 mg/mL, was analyzed with reverse antibody microarrays. If the PSA would denature upon adsorption on the surface, this could possibly be suppressed by the albumin. However, the result indicated the opposite, the higher the albumin concentration, the lower the mean spot intensity (Figure 5). Dynamic Range. In the reverse mode, the dynamic range was found to be at least 2 orders of magnitude ( 10-g1000 µg/mL). The upper limit was not investigated further, as 1000 µg/mL is far above any level of clinical interest. The sandwich antibody microarray exhibited a dynamic range from 0.7 to 7000 ng/mL, corresponding to 4 orders of magnitude. The upper limiting factor in the sandwich mode was due to saturation of the optical detector, since the same detection conditions were used throughout the titration. The PSA concentrations of clinical interest range up to a couple of thousands of nanograms per milliliter. Therefore, the sandwich assay, but not the reverse assay, had a dynamic range appropriate for this clinical application. Spot Reproducibility. In both reverse and sandwich assays, spot reproducibility was measured from 16 spots for each of 4 different concentrations of PSA; the spots are shown in the insets in Figure 3A and B. The two highest concentrations of the sandwich assay were not included since they both reached the saturation level. Table 1 shows the coefficients of variation (CV) and the mean spot intensities. The spot reproducibilities within the chips were similar for both assays. We also note that the spot reproducibilities in the two assays (CVs ∼10-20%) are in agreement with earlier published results.21,24 PSA Measurement in Cell Lysates and Cultivation Medium. To further evaluate the two microarray assays, LNCaP cell lysates and cell cultivation medium were analyzed as examples of samples with highly abundant PSA. The cell lysate was chosen as a model for tissue homogenate. To induce PSA expression, the LNCaP cells were cultivated with 0.1 and 1 nM testosterone analogue R1881. Cell lysate from both cultures and cell medium from the 0.1 nM R1881 culture were subjected to microarray analysis. In the reverse-phase antibody array, concentrated cell lysates and cultivation medium were arrayed by the piezoelectric microdispenser without any difficulties. However, the signals were below detection limit (data not shown). A comparative DELFIA analysis revealed that the highest PSA concentration was 1396 ng/mL (Table 2), which is below the detection limit of the reverse-phase assay. For the corresponding sandwich assay, the samples were diluted 70-fold (in 10 mM PBS) for analysis. High-intensity spots were generated from all samples (Table 2), again confirming the inherently higher sensitivity of the sandwich assay. The PSA concentrations from the sandwich assay results were determined using the titration in Figure 3B as standard curve. We note that a more appropriate standard curve could be constructed from PSA-

Figure 5. Spot intensity profiles of a titration series of human serum albumin (0.015, 0.15, and 1.5 mg/mL) with a constant PSA concentration, 0.18 mg/mL, analyzed by reverse antibody microarrays. The higher the albumin content the weaker the signal, again indicating no pronounced PSA denaturation.

spiked cell lysate and medium from a PSA-free system. Nevertheless, for the cell lysates, the PSA concentrations from the sandwich microarray corresponded well to the DELFIA results. The cultivation medium, however, gave a low PSA concentration in the microarray compared to the one measured in the DELFIA. This result might be explained by signal depression by components of the medium. Aspects of Reverse and Sandwich Microarray Assays. Key requirements for clinical diagnostics are assay simplicity and sufficient sensitivity. The large difference in detection limit Analytical Chemistry, Vol. 79, No. 15, August 1, 2007

5823

Figure 6. (A) PSA-spiked human female serum analyzed with sandwich antibody microarray in two independent experiments (dashed and solid lines). PSA concentrations were obtained by DELFIA assays. An LOD of 0.14 ng/mL was recorded. (B) Linearity of sandwich assays of PSA-spiked female serum in the PSA range of 1-21 ng/mL. DELFIA was used as the reference method, and least-squares analysis for linear regression. Table 1. Mean Spot Intensities and Coefficients of Variation of the Reverse and Sandwich Antibody Microarrays. concentration reverse (µg/mL) 70 315 700 1200 sandwich (ng/mL) 0.7 7 70 700

mean spot intensity

CV, % (n ) 16)

74 179 265 300

15 19 13 21

67 137 915 2826

15 13 20 14

between the reverse and sandwich antibody microarrays therefore strongly favors a sandwich assay format. Based on this, we focused on the sandwich format for further development of label-free microarray methods for PSA detection. It should be noted that the sensitivity in reverse assays can be improved by other detection methods, i.e., planar wave guide readout61 or amplified assays, which require more elaborate protocols or advanced detection principles.21,23,24 Earlier publications describe reverse arrays that use catalyzed signal amplification and colorimetric readout, yielding an LOD of ∼1 ng/mL PSA.21 Nonamplified planar waveguide detection has yielded detection limits in the range of 2.5 ng/mL in reverse mode (personal communication Zeptosens) and low picogram per milliliter in sandwich mode.61 Linearity and Reproducibility of the Sandwich Assay. The final goal of the assay development is measuring PSA levels in patient serum, in which PSA is at low abundance. Female serum was spiked with PSA, and a titration series of PSA-spiked serum, covering the dynamic range of the sandwich assay, was analyzed on two different occasions (Figure 6A). The PSA titration levels were measured by DELFIA. The sandwich assay’s limit of detection in serum was 0.14 ng/mL, covering the frequently used diagnostic cut point of 4 ng/mL. In comparison, according to Figure 3B, the obtained LOD was 0.7 ng/mL for the sandwich assay in buffer. The difference in LOD was attributed to two facts. (61) Pawlak, M.; Schick, E.; Bopp, M. A.; Schneider, M. J.; Oroszlan, P.; Ehrat, M. Proteomics 2002, 2, 383-393.

5824

Analytical Chemistry, Vol. 79, No. 15, August 1, 2007

First, spiking of female serum with catalytic PSA has previously been shown to result in a formation of a stable PSA-linked-to-R2macroglobulin complex; i.e., all PSA is thus not immunoaccessible.52,53 A 30-40% loss of immunoccessible PSA has been estimated in earlier experiments52,53 where purified PSA was added to serum. An additional plausible artifact causing a difference in LOD between buffer and serum analysis is the difficulty in obtaining reproducible dilution series spanning over several orders of magnitude. Second, we have defined our LOD as the lowest detected PSA level, at least two standard deviations above the mean intensity of the background (see Experimental Section); i.e., the two calibration series did not exactly follow the same dilution profile and the experiment performed in Figure 3B had its lowest detectable level slightly higher than the corresponding lowest measurement performed in Figure 6A. With an LOD of 0.14 ng/mL, the sandwich microarray assay has the capacity to discriminate men with early signs of either malignant or benign prostate disease. The LOD of the PSA sandwich assay is below the LOD earlier reported for the directly labeled antibody microarray.62 The graphs of the two different titration experiments (Figure 6A) demonstrate a high reproducibility between the two independent titrations. As a control, female serum was analyzed in the same manner as the rest of the chips. No signals were detectable (data not shown), indicating little unspecific binding. The coefficients of variation for the three lowest PSA concentrations, measured on the 16 spots shown in Figure 6A, are given in Table 3. The linearity of the sandwich assay was further addressed within the working range of 1-21 ng/mL PSA in female serum. The DELFIA immunoassay was again used as a reference method. The linear regression of the mean spot intensity versus PSA concentration is shown in Figure 6B. The coefficient of determination, R2, was above 0.99, indicating a good linearity within the working range. CONCLUSIONS AND FUTURE RESEARCH This work presents our development of an antibody microarray for PSA, based on our high surface area porous silicon substrate. We note that microarray scanners and readout software are available in many laboratories, enabling immediate implementation (62) Ressine, A.; Finnskog. D.; Malm, J.; Becker, C.; Lilja, H.; Marko-Varga, G.; Laurell, T. Nanobiotechnology 2005, 1, 93-104.

Table 2. Sandwich Antibody Microarray Results of LNCaP Cell Lysates and Cultivation Mediuma R1881 added during cultivation (nM)

PSA concn DELFIA (ng/mL)

PSA concn microarray (ng/mL)

microarray CV (%), n)9

LNCaP cell lysate

0.1

1396

1330

12.2

LNCaP cell lysate

1

1272

1246

7.7

LNCaP cultivation medium

0.1

572

322

6.5

sample

array image

a The microarray results were calculated from the mean intensities of the nine spots in the corresponding array image, and translated into PSA concentration using the standard curve in Figure 3B. The spot CVs are calculated from the corresponding array images.

Table 3. Mean Spot Intensities and CVs of PSA-Spiked Female Serum, Assayed in Two Independent Experiments. concentration, ng/mL sandwich assay, experiment 1 0.14 1.27 11.5 sandwich assay, experiment 2 a 1.06 9.68

mean spot intensity

CV, % (n ) 16)

53 172 828

14 14 11

59 197 903

15 11 7

a The DELFIA, used as reference assay, showed a remarkably low PSA level (0.01 ng/mL) for this sample, and the DELFIA assay could not be repeated due to insufficient sample. The corresponding mean spot intensity was plotted in Figure 6A at 0.14 ng/mL, the same concentration as determined by the DELFIA assay in experiment 1.

of protein microarray diagnostics. The assay format was designed to fill the clinical need for label-free analysis. The microarray approach was chosen to enable multiplexing of antibodies, allowing for simultaneous measurement of several biomarkers (63) Hans Lilja, D. U.; Thomas Bjo¨rk, Charlotte Becker, Angel M. Serio, JanA° ke Nilsson, Per-Anders Abrahamsson, A. J. V.; Go¨ran Berglund J. Clin. Oncol. 2007, 25. (64) Steuber, T.; Vickers, A. J.; Haese, A.; Becker, C.; Pettersson, K.; Chun, F. K. H.; Kattan, M. W.; Eastham, J. A.; Scardino, P. T.; Huland, H.; Lilja, H. Int. J. Cancer 2006, 118, 1234-1240.

related to prostate cancer. Several biomarkers in the kallikrein family have proven to be strongly associated with malignant prostate disease63,64 and are thus prime candidates for the multiplex approach. However, detection of some of these markers requires an improvement in LOD of 1-2 orders of magnitude. Current work within our group focuses on signal enhancement in the sandwich assay using new fluorescent strategies for labeling the detection antibody. ACKNOWLEDGMENT The authors thank Ingrid Wigheden, Gun-Britt Eriksson, Kerstin Håkansson, and Birgitta Frohm for providing critical reagents (antibodies and PSA) and expert technical assistance. Also thanks to Dr. Janet Novak for help with preparation of the manuscript. This study was supported by grants from the European Union sixth Framework contract LSHC-CT-2004-503011 (P-Mark), the Foundation for Strategic Research, Swedish Research Council, SWEGENE, the National Cancer Institute Contract P50-CA92629-SPORE Pilot Project 7, the Swedish Cancer Society project 3555, Wallenberg Foundation, Crafoord Foundation, Carl Trygger Foundation, Fundacion Federico, Royal Physiographical Society in Lund, and VR/SSF/Vinnova - Biomedical engineering for better health. Received for May 23, 2007.

review

May

16,

2007.

Accepted

AC0709955

Analytical Chemistry, Vol. 79, No. 15, August 1, 2007

5825