Anal. Chem. 2005, 77, 3222-3230
Cross-Correlation of Optical Microcavity Biosensor Response with Immobilized Enzyme Activity. Insights into Biosensor Sensitivity Lisa A. DeLouise,* Peng Meng Kou, and Benjamin L. Miller
Department of Dermatology and the Center for Future Health, University of Rochester Medical Center, Rochester, New York 14642
Porous silicon multilayer structures have remarkable optical and morphological properties that can be exploited for biosensing. In particular, a high internal surface area (>100 m2/cm3) and a linear response profile to changes in the dielectric environment enable fabrication of sensitive devices and a straightforward quantitation of the optical response. These essential operating characteristics are illustrated for p+ mesoporous silicon (pore diameter 15-20 nm) optical microcavities. A series of devices were prepared to permit the immobilization of glutathione-Stransferase (∼50 kDa) within the porous matrix. Enzyme activity was exploited as an indirect means to quantitate the amount of protein immobilized. Activity was positively correlated with the optical sensor response. However, at high enzyme load the activity becomes nonlinear while the microcavity response remains linear. These data were used to determine the transduction limit (minimum amount of protein required to transduce an optical response), which is reported as areal mass sensitivity ranging between 50 and 250 pg/mm2. This value is considered in context with the dynamic range of the bulk sensitivity, defined as the magnitude of the wavelength shift per refractive index unit, which was measured as a function of microcavity design parameters. This work has uncovered key parameters that can be tuned to improve the detection limit of this sensor modality. Because of the ever increasing number of emerging new biosensor technologies, defining sensor detection limits has become an ambiguous topic and a need exists to standardize measurements and sensitivity units. For chip-based devices, it seems appropriate to report sensitivity in terms of the minimum number of grams of bound target per surface area. The topic of biosensor sensitivity is the subject of great ambiguity, and as new technologies emerge, the lack of consistency in reporting detection limits makes comparative evaluation difficult. Analytical sensitivity or limits detection is frequently described in terms of molarity,1,2 grams/area,3,4 cfu5 or cells/ * To whom correspondence should be addressed. E-mail: Lisa_DeLouise@ urmc.rochester.edu. (1) Ruckstuhl, T.; Rankl, M.; Seeger, S. Biosens. Bioelectron. 2003, 18 (9), 1193-1199.
3222 Analytical Chemistry, Vol. 77, No. 10, May 15, 2005
milliliter,6 grams/milliliter,7 or molecular weight.8 Often the values reported are of a theoretical nature or based on experimental results obtained in studies of model systems,9,10 which do not effectively translate to practical situations where the complex milieu of a biological sample can interfere with the ability to measure equilibrium target/receptor binding. These latter effects are frequently handled in traditional bioassays through careful sample preparation and the operating protocol combining lengthy blocking and washing steps to minimize the occurrence of false positives from nonspecific binding.11 To enhance target sensitivity, signal amplification strategies have been employed including enzymatic reporter constructs and PCR for nucleic acid-based assays, the latter of which is reported to be capable of detecting as few as 100 copies of the target oligonucleotide per milliliter.12 Such strategies are opposed, however, to the challenge of developing rapid label-free technologies that could be used in point-of-care settings, a central focus of our research program. Novel biosensor techniques based on molecular beacons,13 photoluminescence,14-16 planar17,18 and nanoparticle19-21 surface (2) Plowman, T. E.; Reichert, W. M.; Peters, C. R.; Wang, H. K.; Christensen, D. A.; Herron, J. N. Biosens. Bioelectron. 1996, 11 (1-2,) 149-160. (3) Shumaker-Parry, J. S.; Campbell, C. T., Anal. Chem. 2004, 76 (4), 907917. (4) Vollmer, F.; Braun, D.; Libchaber, A.; Khoshsima, M.; Teraoka, I.; Arnold, S. Appl. Phys. Lett. 2002, 80 (21), 4057-4059. (5) Mathew F. P.; Alocilja, E. C. Biosens. Bioelectron. In press. (6) Chen, P.; Mao, F. C. J. Food: Drug Anal. 2004, 12 (2), 133-139. (7) Nath, N.; Chilkoti, A. Anal. Chem. 2004, 76 (18), 5370-5378. (8) Karlsson, R.; Stahlberg, R. Anal. Biochem. 1995, 28 (2), 274-280. (9) Thrush, E.; Levi, O.; Wang, K.; Harris, J. S.; Smith, S. J. IEEE-EMBS Special Topic Conference on Microtechnologies in Medicine & Biology 2002, (May), 374-379. (10) Chen, S. J.; Chien, F. C.; Lin, G. Y.; Lee, K. C. Opt Lett. 2004, 29 (12), 1390-1392. (11) Antibodies in Cell Biology; Asai, D. J., Ed.; Methods in Cell Biology; Academic Press: New York, 1993; Vol. 37. (12) Ivnitski, D.; O’Neil, D. J.; Gattuso, A.; Schlicht, R.; Calidonna, M.; Fisher, R. Biotechniques 2003, 35 (4), 862-869. (13) Tan, W.; Wang, K.; Drake T. J. Curr. Opin. Chem. Biol. 2004, 8 (5), 547553. (14) Wang, D.; Gong, X.; Heeger, P. S.; Rininsland, F.; Bazan, G. C.; Heeger, A. J. Proc. Natl. Acad. Sci. U.S.A. 2002, 99 (1), 49-53. (15) Chan, S.; Horner, S. R.; Fauchet, P. M.; Miller, B. L. J. Am. Chem. Soc. 2001, 123 (47), 11797-11798. (16) Chan, S.; Fauchet, P. M.; Li, Y.; Rothberg, L. J.; Miller, B. L. Phys. Status Solidi A 2000, 182, 541-546. (17) Jung, L. S.; Campbell, C. T.; Chinowsky, T. M.; Mar, M.; Yee, S. S. Langmuir 1998, 14, 5636-5648. (18) Wegner, G. J.; Lee, H. J.; Marriott, G.; Corn, R. M. Anal. Chem. 2003, 75, 4740-4746. 10.1021/ac048144+ CCC: $30.25
© 2005 American Chemical Society Published on Web 04/09/2005
plasmon resonance (SPR), and other evanescent field techniques22-24 hold promise for label-free analysis because the detection signal originates directly from target binding and labeling the target biomolecule is not required. The binding event causes a change in optical property, such as the extinction coefficient or refractive index, which can be measured without the need for amplification or secondary reporter constructs. The sensitivity of optical transducers to a refractive index change can be exceedingly high (10-4-10-6).17 However, how this translates to the number of target molecules that must bind within the acceptance geometry of the detector is not always clear, but progress is being made.17,25 Beyond the intrinsic sensitivity of the signal transduction mechanism, certain device architectural and operating parameters can impact the detection limit. Such parameters include the number of receptor probes immobilized per transducer surface area, whether measurements are made “real time” in a flow cell versus postincubation with a static solution, and whether the transducer element must be dried before making the measurement. In all cases, sample volume, target concentration in the sample solution, and exposure time will set an upper limit on the maximum number of target molecules that may bind the receptor. When a device is exposed to a static amount of sample solution, an equilibrium on/off rate may be established, but when the transducer is rinsed and dried for optical measurement, some fraction of bound target may be lost depending upon target/ receptor binding affinity. In consideration of these issues and given the lack of standardized protocols for assessing sensor sensitivity, it is no wonder that comparing reported detection limits for emerging technologies is difficult. For the particular case of evaluating optical sensors, a premium must also be placed on comparative control studies to validate the specificity of the optical response. In this paper, we are concerned with investigating sensitivity and the transduction limit of a p+ porous silicon (PSi) optical microcavity (MC) sensor produced by anodic electrochemical dissolution in an HF-containing electrolyte.26 The physical characteristics of the PSi MC and signal transduction mechanism that produces optical shifts in both the photoluminescence and white light reflection detection modes have been widely documented.15,16,26 Briefly, the multilayer microcavity is a 1-D photonic band gap structure that is sensitive to changes in porosity resulting from target binding to a receptor immobilized within the porous matrix. Effective medium theory27 can be used to relate porosity to index of refraction (η) when the pore morphology is in the mesoporous regime (pore diameter 2-50 nm).28,29 Simulations can be done to (19) Haes, A. J.; Van Duyne, R. P. Proc. SPIE 2003, 5221, 47-58. (20) Nath, N.; Chilkoti, A. Anal. Chem. 2004, 76, 5370-5378. (21) Zayats, M.; Baron, R.; Popov, I.; Willner, I. Nano Lett. 2005, 5 (1), 21-25. (22) Wolfbeis, O. S. Anal. Chem. 2002, 74, 2663-2678. (23) Kasili, P. M.; Song, J. M.; Vo-Dinh, T. J. Am. Chem. Soc. 2004, 126 (9), 2799-2806. (24) Vollmer, F.; Arnold, S.; Braun, D.; Teraoka, I.; Libchaber, A. Biophys. J. 2003, 85, 1974-1979. (25) Haes, A. J.; Van Duyne, R. P. J. Am. Chem. Soc 2002, 124, 10596-10604. (26) Vinegoni, C.; Cazzanelli, M.,; Pavesi, L. In Silicon based Materials; Vol. 2. Properties and Devices; Nalwa, H. S., Ed.; Academic Press: New York, 2001; pp 124-188. (27) Bruggeman, D. A. G. Ann. Phys. Paris 1935, 24, 636. (28) DeLouise, L. A.; Miller, B. L. Mater. Res. Soc. Symp. Proc. 2004 782, A5.3.1A5.3.7. (29) DeLouise, L. A.; Miller, B. L. Proc. SPIE 2004, 5357, 111-125.
estimate the magnitude of the wavelength shift as a function of pore filling, and as will be discussed in more detail later, a key advantage of the microcavity technique is that a linear dependence on pore filling is found. Moreover, because the microcavity is a resonant device, reflected light interacts with bound target multiple times increasing the sensitivity compared to single-pass planar optical techniques. Proof of principle of the PSi microcavity biosensor has been experimentally demonstrated,16 as have a variety of other porous silicon sensor modalities (including rugate filters,30-32 Bragg filters,33 single-layer Fabry-Perot films,34,35 and protein microarrays36 for detecting antigen-antibody, DNA, and protein binding. Porous silicon has also been demonstrated as an effective support for immobilized enzyme catalysis.37,38 These studies highlight many of the advantages of porous silicon, which include ease of fabrication with the possibility for integration with wafer level IC processing, and an extremely large internal surface area (on the order of several hundred m2/cm3), which depending upon depth can easily exceed 1000 times that of a planar surface of equal diameter. This enables immobilization of large amounts of receptor relative to monolayer techniques, and recently we showed that protein immobilization capacity scales with device thickness.39 This property can in principle be leveraged to enhance device sensitivity many orders of magnitude over a planar device of comparable transducer diameter as will be discussed in more detail below. However, for the microcavity to function effectively, the porous morphology must enable facile diffusion of bioreagents throughout the multilayer matrix. While pore morphology (porosity and pore diameter) is tunable over a wide range40,41 by varying etch parameters (current density, etchant formulation, wafer doping level and type), it currently remains a significant challenge to optimize pore size distribution and interface roughness to achieve facile diffusion while maintaining high-quality optical device characteristics.28 Moreover, a quantitative understanding of the dependence of microcavity optical response and detection sensitivity on morphology, multilayer device architecture, and optical resonance frequency is lacking. In this study, we take steps toward gaining insight into these important attributes. Specifically, mesoporous p+ microcavity devices were fabricated and the morphology modified28,29 to allow the quantitative immobilization of the enzyme glutathione-S(30) Kaminska, K.; Brown, T.; Beydaghyan, G.; Robbie, K. Appl. Opt. 2003, 42 (20), 4212-4219. (31) Schmedake, T. A.; Cunin, F.; Link, J. R.; Sailor, M. J. Adv. Mater. 2002, 14 (18), 1270-1272. (32) Link, J. R.; Sailor, M. J. Proc. Natl. Acad. Sci. U.S.A. 2003, 100 (19), 1060710610. (33) Andreson, M. A.; Tinsley-Bown, A.; Allcock, P.; Perkins, E. A.; Snow, P.; Hollings, M.; G. Smith, R.; Reeves, C.; Squirrell, D. J.; Nicklin, S.; Cox, T. I. Phys. Status Solidi A 2003, 197, 528-533. (34) Tinsley-Bown, A. M.; Canham, L. T.; Hollings, M.; Anderson, M. H.; Reeves, C. L.; Cox, T. I.; Nicklin, S.; Squirrell, D. J.; Perkins, E.; Hutchinson, A.; Sailor, M. J.; Wun, A. Phys. Status Solidi A 2000, 182, 547-553. (35) Dancil, K.-P. S.; Greiner, D. P.; Sailor, M. J. J. Am. Chem. Soc. 1999, 121, 7925-7930. (36) Ressine, A.; Ekstro ¨m, S.; Marko-Varga, G.; Laurell, T. Anal. Chem. 2003, 75 (24), 6968-6974. (37) Drott, J.; Lindstro ¨m, K.; Rosengren, L.; Laurell, T. J. Micromech. Microeng. 1997, 7, 14-23. (38) Letant, S. E.; Hart, B. R.; Kane, S. R.; Hadi, M. Z.; Shields, S. J.; Reynolds, J. G. Adv. Mater. 2004, 16, 689. (39) DeLouise, L. A.; Miller, B. L. Anal. Chem. 2004, 76, 6015-6920. (40) Zhang, X. G. J. Electrochem. Soc. 2004, 151, c69-c80. (41) Lehmann, V.; Stengl, R.; Luigart, A. Mater. Sci.: Eng. 2000, B69, 11-22.
Analytical Chemistry, Vol. 77, No. 10, May 15, 2005
3223
Figure 1. SEM cross section of a typical λ/2 microcavity comprised of 10 periods of alternating low (20 mA/cm2) and high (70 mA/cm2) porosity in each mirror stack.
transferase (GST), a medium-sized (50 kDa) protein dimer. A series of devices were prepared in which the amount of immobilized enzyme was varied. A unique approach, exploiting an immobilized enzyme assay recently described,42 was used to quantify the amount of functional protein, which we crosscorrelated to the optical sensor response. A positive correlation was found; however, at high enzyme load, activity becomes nonlinear while the microcavity optical response remains linear. The linearity of the microcavity response is advantageous in that enables exploitation of label-free signal amplification strategies based on multiple sequential binding as described by Sailor and co-workers.35 It also simplifies quantitation of the optical response relative to competitive techniques based on evanescent waves where the signal response decays exponentially from the surface causing sensitivity to depend on the surface-analyate distance.17 To leverage these advantages, it is essential to fundamentally understand how the microcavity response depends on design parameters such as device architecture and morphology (porosity and pore diameter). To gain insight into this problem, we have investigated the sensitivity of the microcavity response to changes in bulk refractive index unit (RIU) as a function of microcavity design parameters. EXPERIMENTAL SECTION Mesoporous microcavity fabrication, the GST immobilization process, and the enzyme assay have all been described in detail elsewhere.16,28,29,39,42 Here we briefly review the process used to fabricate nearly equivalent microcavity devices in which varying amounts of GST enzyme was immobilized. Microcavity Fabrication. High optical quality26,29 microcavity devices were fabricated by anodic electrochemical etching of a p+ silicon 〈100〉 wafer (0.01 Ω‚cm) using an electrolyte containing 70% ethanol and 30% hydrofluoric acid (48% in water). A porous silicon microcavity is composed of two multilayer λ/4 dielectric mirror stacks (alternating high- and low-porosity layers) spaced by an active layer (cavity). The cross section of a typical porous silicon microcavity is pictured in Figure 1. Here the optical (42) DeLouise, L. A.; Miller, B. L. Anal. Chem. 2005, 77, 1950-1956.
3224 Analytical Chemistry, Vol. 77, No. 10, May 15, 2005
thickness (ηd) of the active layer is twice the high-porosity layer and is hence referred to as a λ/2 microcavity. The λ/2 microcavities produced for the GST experiments were fabricated using LabView to control a constant current power supply that alternated between a low (35 mA/cm2) and high (70 mA/cm2) current density for a fixed time generating low (70-75%) and high (8085%) porosity layers before KOH treatment. Each mirror was composed of 14 periods of low- and high-porosity bilayers yielding a total device thickness of ∼7 µm. The λ/2 microcavities fabricated for the refractive index sensitivity experiments were produced by alternating between a low (20 mA/cm2) and high (70 mA/cm2) current density generating low (60-65%) and high (80-85%) porosity layers before KOH treatment. The number of low- and high-porosity periods and the total device thickness were varied in the sensitivity experiment. Our etch process is able to achieve a high degree of sample-to-sample reproducibility as evidenced by the standard error ((4 nm) in the average value of the FabryPerot resonance wavelength of microcavities produced in sequential anodization runs.29 Following etching, each the microcavities for the GST studies was immersed in 10 mL of a 1.5 mM KOH solution containing ethanol for 5 min to remove the high surface area nanostructures that fill the pore channels and increase pore diameter.29 This treatment eliminates the characteristic photoluminescence property of mesoporous silicon; however, examination of the sensors’ optical response in reflection provides a means to quantify the effect of KOH on the microcavity.29 After KOH exposure, samples were rinsed in ethanol and then water and dried under a stream of N2 gas. Prior to enzyme immobilization, the devices were thermally oxidized using a three-zone Lindberg tube furnace at 900 °C. Oxidation is required to create hydrophilic pore channels and to improve the biocompatibility and stability of the porous silicon.43 Soak test studies (not shown) indicate that oxidation plus subsequent surface functionalization (described below) stabilize the microcavity optical response, eliminating concerns for baseline drift due to corrosion in high-salt-containing solutions.34 Depending upon the initial layer porosities, the quality of the microcavity may degrade substantially following oxidation,28,29 particularly for KOHtreated samples. Degradation is characterized by substantial increase in the full width half-maximum of the resonance peak, which can make identifying the exact position of the of the FabryPerot peak wavelength difficult. In this case, red shift measurements are made at several percent refection values across a specified wavelength range. Based on SEM images analyzed using NIH ImageJ (see Supporting Information), the fabrication of layers with 35 and 70 mA/cm2 yielded an average pore size of ∼13 and 20 nm, respectively. Pore diameters range, however, in the lowporosity layer between 5 and 28 nm and in the high-porosity layer between 6 and 51 min before KOH treatment. KOH treatment widens the pore diameters such that biomolecular pore infiltration is primarily limited by diffusion through the low-porosity layer, which exhibits an average pore diameter ranging between 15 and 50 nm. To approximate the internal surface area, we assume a pore diameter of 20 nm and an interpore spacing of ∼30 nm, which in previous studies provided good comparison with experimental results.39 Using a simple geometric mode with pore channels (43) Canham, L. T.; Stewart, M. P.; Buriak, J. M.; Reeves, C. L.; Anderson, M.; Squire, E. K.; Snow, P. A. Phys. Status Solidi A 2000, 182, 521-525.
Figure 2. Time course plots of the optical response of non-KOH-treated microcavity following exposure to (a) 2% aminosilane solution and (b) 2.5% glutaradelhyde solution.
spaced on a hexagonal lattice, we estimate an internal surface on the order of ∼1000 cm2 with a surface-to-volume ratio of ∼107 m2/cm3. Microcavity Optical Characterization. Optical characteristics were checked before and after all microcavity fabrication and surface derivatization steps by white light reflection using an Ocean Optics HR2000 system. The reflection probe is a bidirectional fiber bundle in which white light is illuminated down a central fiber and reflected light is collected in six surrounding fibers. The fiber diameter is 400 µm with a numerical aperture of 0.22. The white light source illuminates a sample spot size of ∼2mm diameter. Hence, the optical response signal is derived from only a small volume of the 1.3 cm diameter × 7 µm thick microcavity chip. For purpose of discussion, we assume this volume to be ∼2.2 × 10 -5 cm3. Enzyme Immobilization. Each microcavity device was equivalently derivatized for GST enzyme immobilization using standard aminosilane and glutaraldehyde coupling chemistry described in detail elsewhere.44 The optical response of the sensor was monitored after each chemical derivatization step. First, 40 µL of an aqueous 2% aminosilane solution in 50% methanol was applied to the chip for 15-20 min. After rinsing with methanol and then water, and drying under a stream of N2, the chips were heated at 100 °C for 15-20 min prior to recording the optical shift. Time course studies of the optical shift due to silane derivatization were conducted on non-KOH-treated microcavity devices (Figure 2). A time-independent red shift of ∼15 nm is observed, suggesting that the silane layer forms quickly and that pore filling from extensive silane polymerization within the porous matrix is not a concern under the conditions utilized. Utilizing a similar procedure, eight microcavities prepared in separate anodization runs (diced from the same wafer) were silanized following KOH treatment. Compared to the non-KOH-treated microcavity a much smaller 4.6 ( 0.4 nm silane red shift is observed. We attribute this to a decrease in surface area resulting from the removal of highly branched nanostructures by KOH.29 We also note that a portion of the aminosilane red shift (2-3 nm) is likely due to rehydration of the oxide. We observe that following hightemperature oxidation the microcavity is extremely sensitive to humidity. Exposing a freshly oxidized chip to water or an aqueous methanol solution will cause a 2-5-nm red shift depending upon surface area. After the chip becomes rehydrated, the optical response remains insensitive to ambient levels of humidity. (44) Hermanson, G. Bioconjugate Techniques; Academic Press: New York, 1996.
Following silane treatment, 50 µL of a 2.5% glutaraldehyde solution in 20 mM phosphate buffer containing 1 mM EDTA (PBE pH 8.6) was applied to the chip for a fixed time. After rinsing with water and drying under a stream of N2, the optical response was recorded. Time course studies of the optical shift due to glutaraldehyde attachment were conducted on non-KOH-treated microcavity devices (Figure 2). In contrast to the silane results, reaction of glutaraldehyde with the amino surface is time dependent. Rapid uptake occurs in the first 20 min, followed by a much slower increase in deposition. To prevent excessive pore filling with polymerized glutaraldehyde, we limited the reaction time to 30 min. For the non-KOH-treated microcavity, this yields a ∼18-nm shift from 2.5% glutaraldehyde and a cumulative shift of ∼33 nm from the oxide. Utilizing a similar procedure, the eight microcavities prepared in separate anodization runs and equivalently silanized were exposed to 2.5% glutaraldehyde for 30 min. Compared to the non-KOH-treated microcavity, a much smaller 6.3 ( 0.5 nm glutaraldehyde red shift is observed yielding a cumulative shift from the oxide of 10-11 nm (silane plus glutaraldehyde). Consistent with the aminosilane results, the magnitude of the glutaraldehyde shift in KOH-treated chips is ∼3 times lower than that observed in non-KOH-treated chips. We also note that pretreatment of the chips with aminosilane is essential for glutaraldehyde attachment, as one would expect based on reactivity. No red shift results following exposure to glutaraldehyde unless the porous silicon is pretreated with aminosilane. An essential requirement for being able to successfully conduct a quantitative correlation of optical microcavity response with immobilized enzyme activity lies in our ability to immobilize a known quantity of enzyme, the activity of which must be stable over the duration of the experiment. The immobilization capacity of a KOH-treated microcavity chip can be approximated, assuming a surface area of ∼1000 cm2 as predicted by a geometrical model.39 Crystallographic data on GST indicate a globular structure with a maximum diameter of ∼10 nm.45 Hence, the planar area occupied by an immobilized GST molecule is ∼7.85 × 10-13 cm2. To approximate the GST immobilization capacity we assume that GST can bind the internal surface area of porous silicon in a monolayer fashion with zero space between dimers. While it is unlikely GST could pack this closely, the analysis predicts an immobilization capacity of ∼2.1 nmol/chip (1.3-cm diameter), and despite the uncertainty in our estimate of internal surface area and GST (45) Feil, S. C.; Wilce, M. C. J.; Rossjohn, J.; Allocati, N.; Aceto, A.; Di Ilio, C.; Parker, M. W. Acta Crystallogr. 1996, D52, 189-191.
Analytical Chemistry, Vol. 77, No. 10, May 15, 2005
3225
diameter, experimental data have shown good correlation with the geometrical model.39 A covalent attachment strategy was chosen to ensure the stability of the immobilized enzyme. Covalent immobilization of GST on a glutaraldehyde-treated surface occurs via imine formation by reaction with surface lysine side chains, the amino terminus of the protein, or both. Microcavity chips were exposed to 50 µL of a GST solution ranging in concentration from 0.05 to 2.0 mg/mL. After an immobilization time of 1.5 h, the residual supernatant was tested for enzyme activity using a solution activity assay discussed below. The lack of detectable enzyme activity in the supernatant was interpreted as indicating the enzyme was immobilized in the porous matrix. Wash steps were conducted to eliminate any enzyme weakly bound to the porous matrix. Immobilized GST microcavity chips were stored in buffer between activity measurements to prevent the enzymes from denaturing. Immobilized Enzyme Activity Assay. Glutathione-S-transferases are dimeric (25 kDa/monomer) cystolic proteins that catalyze the nucleophilic attack by the thiol group (SH) of glutathione (GSH) on the electrophilic center of lipophilic substrates.46 Numerous species-independent GST gene classes are known, all of which exhibit high enzyme activity toward the substrate 1-chloro-2,4-dinitrobenzene (CDNB).47 This activity serves as the basis for assessing immobilized GST activity. Formation of the conjugation product can be monitored spectrophotometrically at 340 nm. Details of the immobilized enzyme assay have been described elsewhere.42 Briefly, 50 µL of substrate stock solution containing 3.3 mM GSH and 3.3 mM CDNB was applied to the microcavity chip. The conjugation reaction was allowed to proceed for 1 min after which 40 µL of solution was recovered from the chip, diluted with 560 µL of PBE buffer (100 mM potassium phosphate monobasic buffer with 1.0 mM EDTA at pH 6.5), and the absorbance recorded. Residual substrate stock solution was washed from the microcavity chip with PBE buffer, and the device was allowed to soak in buffer for g10 min before repeating the measurement. Activity measurements were recorded for five independent reactions using fresh substrate stock solution, and the average absorbance value was reported. Materials. All reagents were purchased from Sigma and used without further purification. Solutions containing GST (G6511), CDNB (C6396), and reduced GSH (G6529) were made fresh each day prior to use. Aqueous solutions were mixed using PBE pH 6.5, except for the glutaraldehyde solution, which was at pH 8.6. A stock solution of 200 mM CDNB was prepared in ethanol. RESULTS Factors Affecting Porous Silicon Microcavity Sensitivity. As mentioned earlier, simulations employing effective medium theory indicate that the microcavity optical response is a linear function of pore filling. As an example, simulation indicates that the resonance peak of a λ/2 microcavity, constructed with 10 periods per mirror of high porosity (80%) and low porosity (60%) and tuned to operate in the visible at 685 nm shifts linearly to longer wavelengths by ∼129 nm when the pores are filled 100% with a fluid of refractive index increasing from η ) 1.00 to 1.33. This yields a theoretical wavelength shift sensitivity of ∼385 nm/ RIU. It is of interest to note that when the microcavity is simulated to operate in the IR, a higher nm/RIU sensitivity is achieved. For example, a microcavity of identical porosity tuned at 1500 nm will 3226 Analytical Chemistry, Vol. 77, No. 10, May 15, 2005
Figure 3. Experimentally determined optical red shift of a mesoporous silicon microcavity to bulk refractive index change. The RIU sensitivity determined from the slope is ∼360 nm/RIU for the microcavity tuned in the visible spectrum with resonance peak at 685 nm with air in the pores. Refactive indices for solvents are from an on-line source (http://chemweb.unp.ac.za/chemistry/Physical_Deat/ Solvent_properties.htm).
undergo a red shift of ∼300 nm with the pores filled 100% with water yielding a sensitivity value of ∼900 nm/RIU. Experimentally, the bulk RIU sensitivity of the optical response of a λ/2 microcavity designed to operate in the visible spectrum at 685 nm was measured by filling the pores 100% with solvents of different index of refraction. This device consisted of 10 periods of a low (60-65%) and high (80-85%) porosity layer in each mirror. The results (Figure 3) show a linear response with a sensitivity value of ∼360 nm/RIU, consistent with simulations. Additional studies were done to investigate the dependence of bulk RIU sensitivity on device design parameters including the number of layers per mirror, active layer thickness, resonance frequency, and cavity quality defined by the Q factor (Q ) λ/∆λ). The results, summarized in Table 1, suggest that the RIU sensitivity is independent of Q, the thickness of the device, and active layer thickness. The initial resonant frequency appears to be the only parameter on which the bulk RIU sensitivity is strongly dependent on. Due to limitations in our spectrophotometer, we cannot measure far out in the IR; however, given the available data the dependence appears quadratic as illustrated in Figure 4. This feature of the microcavity response can be exploited to design higher sensitivity (nm/RIU) devices that operate in the IR. Moreover, despite the independence of sensitivity on Q, a higher quality device with sharper resonance features could improve detection capability by simply improving the reliability in resolving small (subnanometer) resonance shifts. This assumes of course a high spectral resolution capability in the detector as well as sufficiently small acceptance geometry to minimize signal averaging over optical inhomogeneities within the sample analysis volume. Recent studies have documented the importance of sample analysis volume on Q.48 Clearly, more work is needed to completely understand the interplay of device architectural and instrumental factors that (46) Habig, W. H.; Pabst M. J.; Jakoby, W. B. J. Biol Chem. 1974, 249 (22), 7130-7139. (47) Ortiz-Salmero´n, E.; Yassin, Z.; Clemente-Jimenez, M. J.; Javier, F.; Las HerasVazquez, L.; Rodriguez-Vico, F.; Baro´n, C.; Garcı´a-Fuentes, L. Eur, J. Biochem. 2001, 268, 4307-4314.
Table 1. Experimental Optical RIU Sensitivity as a Function of Microcavity Parameters no. of bilayer periods/mirrora approximate layer thickness (nm):b low-porosity layer high-porosity layer active layer
5
10
15
5
10
7
10
86 130 260
86 130 260
86 130 260
98 150 300
98 150 300
98 150 2340c
106 163 326
approximate total device thickness (µm)
2.2
4.3
6.5
2.5
5.0
5.8
5.4
resonance wavelength (nm) fwhm (nm) approximate Q factor
629.9 22.7 28
581.4 4.5 130
632.1 3.2 200
718.0 22.0 32
699.8 4.7 150
691.6 4.7 146
742.6 4.9 152
approximate sensitivity (nm/RIU)
330
330
330
380
370
360
425
a Porosities of the mirror layers were 60-65% and 80-85%. b Thicknesses values are approximated from predetermined etch rate values. c The porosity of the thick active layer was 73%.
Table 2. One-Minute Conjugation Reaction Results Contrasting GST Activity in Residual Supernatant Relative to a Solution Control Using 3.3 mM GSH and 3.3 mM CDNB, 75 µL Total Volume
Figure 4. Experimentally determined RIU sensitivity as a function of the resonance wavelength illustrating an apparent quadratic dependence. More sensitive devices, defined by a larger wavelength shift per refraction index unit change, can be realized by fabricating devices tuned to operate deeper in the IR. Dotted line illustrates the quadratic fit.
impact microcavity detection sensitivity. Porosity and surface area are particularly significant because preliminary results indicate that using KOH to modify pore morphology decreases the available surface area, which consequently will lower amount of receptor that can be bound. The ability to immobilize a high receptor concentration is central to enhancing detection sensitivity, particularly for weak Kd targets. Based on equilibrium arguments (as illustrated in the example provided in the Supporting Information), the concentration of target in solution that is required to generate a threshold detection signal can lowered by several orders of magnitude by raising the receptor concentration. This is particularly important for detecting low binding affinity targets (millimolar Kd) where the off-rate is higher and it is more difficult to maintain a bound target concentration. Herein lies a potential advantage of porous silicon over a planar silicon device of equal diameter. Assuming an equivalent sample analysis area, the internal surface area of porous silicon on which receptors can be immobilized in nearly 3 orders of magnitude larger for a device ∼7 µm thick.39 Because the microcavity signal is generated from refractive index changes that occur within the volume of the device, the microcavity can potentially be much more sensitive (48) Ghulinyan, M.; Oton, C. J.; Bonetti, G.; Gaburro, Z.; Pavesi, L. J. Appl. Phys. 2003, 93 (12), 9724-9729.
soln control GST concn (nmol)
enzyme reaction concn (µg/mL)
absorbance from solution control
absorbance from residual chip supernatant
0.025 0.05 0.25 0.50 1.00
16.7 33.3 166.7 333.3 666.7
0.036 ( 0.007 0.084 ( 0.006 0.310 ( 0.017 0.416 ( 0.023 0.565 ( 0.037
-0.008 -0.005 -0.004 0.001 0.096
than a planar device. Of course, other factors including the steric demands of probe and target pore infiltration will impact this comparison. Determination of GST Microcavity Detection Limit. Previously, we demonstrated the ability to modify the microstructure of p+ mesoporous silicon using KOH to enable pore infiltration of GST (50 kDa).28,29 The immobilization capacity was quantitatively observed to scale with device thickness.39 Here we are interested in determining microcavity detection limits for GST or the minimum amount of protein that must bind the transducer to produce a detectable signal. We prepared five nearly equivalent microcavity chips, 7 µm thick with an immobilization capacity of ∼2.1 nmol. The chips were derivatized with approximately 0.05, 0.1, 0.5, 1, and 2 nmol of GST by applying 50 µL of 0.05, 0.1, 0.5, 1.0, and 2.0 mg/mL GST solutions (∼1 to ∼40 µM). After 1.5 h, 25 µL of the residual supernatant was recovered from the chip and analyzed for enzyme activity. If the amount of GST present in 25 µL of supernatant is negligible (all bound to the silicon), then the conjugation product yield (proportional to absorbance at 340 nm) produced in the 1-min assay will be negligible. In contrast, if no GST binds to the chip, then the maximum amount of enzyme contained in the 25 µL of residual aliquot would be half of the initial amount applied. The activity results in the residual supernatant relative to a control are summarized in Table 2. The data indicate that all of the GST are mobilized with the possible exception of the chip exposed to 2 nmol of GST. Hence, we conclude that active enzyme was indeed present in the supernatant for which 2 nmol of GST was applied. This means that the immobilization capacity was slightly less than expected based on the geometric model. It is difficult to estimate from data in Table 2 what fraction of the 2 nmol was absorbed, since at high Analytical Chemistry, Vol. 77, No. 10, May 15, 2005
3227
Figure 5. Conjugation product yield (proportional to absorbance at 340 nm) obtained from immobilized enzyme chips relative to a solution control. Samples are designated by the maximum moles of GST applied to the microcavity. Standard deviation errors for solution control and immobilized enzyme are 7 and 9%, respectively.
GST reaction concentrations (>200 µg/mL), the onset of nonlinearity in the product yield time course plot due to substrate limitation occurs. However, based on the absorbance value produced from the 0.05-nmol control, we estimate that ∼1.9 nmol (95%) was immobilized. To verify the immobilized GST was functional, an activity assay was performed over the entire microcavity chip. A comparison of the immobilized enzyme activity relative to a solution control (Figure 5) shows that the activities track each other; however, when the immobilized load exceeds >0.5 nmol/chip (GST reaction concentration 333 µg/mL), the absorbance value becomes nonlinear due to substrate limitation. Enzyme activity in the solution is greater in all cases, which is attributed to a decrease in immobilized enzyme activity resulting from steric hindrance of the active site or microenvironment partition effects (H+ ion or substrate gradients) that alter kinetic parameters.42 After characterizing the enzyme activity of each microcavity chip, the devices were rinsed with water and dried under nitrogen, and the optical response was recorded. Changes in the optical response following each surface functionalization step were observed as illustrated in Figure 6 for the microcavity chip immobilized with 1 nmol of GST. Because of the low Q factor and the asymmetry following GST exposure, the wavelength position of the Fabry-Perot resonance dip is difficult to determine precisely. Hence, the wavelength shifts reported were determined by averaging values measured at various percent reflection across the wavelength spectrum ranging between 575 and 635 nm. Surface derivatization decreases the porosity (increases the index of refraction) causing red shifts in the optical response following each surface functionalization step. A red shift of ∼10-11 nm was induced following 1 nmol of GST contributing to a cumulative red shift of ∼24 nm from the starting oxide. The magnitude of the optical shift as a function of the immobilized GST was cross-correlated against the immobilized enzyme activity as shown in Figure 7. The microcavity response appears linear, increasing ∼10 nm/nmol of GST. This is in contrast to the nonlinear enzymatic response observed at the higher GST 3228
Analytical Chemistry, Vol. 77, No. 10, May 15, 2005
concentrations. The data also show the optical response detection limit falls between 0.05 and 0.1 nmol whereas the activity assay indicates enzyme is present. Subsequent attempts to measure the optical response following immobilization of 0.1 and 0.2 nmol of GST (50 µL of ∼2 and 4 µM GST solutions) in microcavities of similar construction proved negative. Hence, we conclude from this study that the detection limit for the particular microcavity architecture investigated here requires immobilization of a minimum of 0.2-0.5 nmol of GST (10-25 µg of protein) within the matrix of the device to transduce a signal. This is nearly 1 order of magnitude worse than the enzyme assay, which is most likely due to incomplete penetration of the GST into the porous matrix.49 If small quantities of GST bind in the top 1-2 layers of the microcavity, a much attenuated optical response will result. Nonetheless, assuming for analysis purposes, that GST is homogeneously dispersed throughout the matrix (7 µm thick with an internal surface area of ∼1000 cm2 ), this result corresponds to an areal mass sensitivity on the order of ∼100-250 pg/mm2, which is only 1-2 orders less sensitive than benchmark detection limits claimed by SPR17,25 but 10-100 times better than that reported for streptavidin binding to a biotinylated porous silicon single-layer Fabry Perot interference filter.50 Given that the reflection probe measures the optical response from a volume of ∼ 2.2 × 10-2 mm3 containing an approximate internal surface area of 24 cm2, this threshold signal results from the binding 100600 ng of protein. DISCUSSION The sensitivity dependence of a p+ mesoporous silicon λ/2 optical microcavity to changes in refractive index has been investigated in two ways. First we studied sensitivity as defined by the magnitude of the wavelength shift caused by changes in the bulk refractive index unit (nm/RIU). Second, studies were conducted to determine sensitivity defined as the minimum amount of protein required to transduce a threshold optical response. Results show that wavelength shift sensitivity depends on the resonance wavelength ranging between ∼400 and 900 nm/ RIU for devices tuned to operate between the visible and IR. While an order of magnitude lower than the 3000-8800 nm/RIU benchmark values claimed by the planar SPR technique,17 the microcavity remains competitive with the wavelength shift sensitivity recently reported for the LSPR (190 nm/RIU)19,25 and nanoSPR techniques (10-13 nm/RIU)20 based on supported nanoparticles. Given the advantages of p+ PSi microcavities discussed in terms of the large internal surface area in which a high concentration of receptor can be immobilized, the microcavity device could potentially offer unique sensitivity advantages for detecting weaker Kd targets, which are common in biological systems. This is particularly important as device dimensions miniaturize to accommodate arrayed formats where less capture probe is immobilized per spot. Recently, Laurell and co-workers36 demonstrated the possibility of using hydrophobic macroporous silicon as a substrate for protein microarrays using fluorescent labeled detection. While this work did not seek to leverage the (49) Karlsson, L. M.; Tengvall, P.; Lundstrom, I.; Arwin, H. J. Colloid Interface Sci. 2003, 266, 40-47. (50) Janshoff, A.; Dancil, K. P. S.; Steinem, C.; Greiner, D. P.; Lin, V. S. Y.; Gurtner, C.; Motesharei, K.; Sailor, M. J.; Ghadiri, M. R. J. Am. Chem. Soc. 1998, 120, 12108-12116.
Figure 6. Reflection spectra for microcavity chip derivatized with 1 nmol of GST. Red shift values reported are averaged over several points measured between 575 and 675 nm: aminosilane +6 nm; glutaraldehyde +8 nm; 1 nmol of GST +10 nm, yielding a total shift of +24 nm from the oxide.
Figure 7. Correlation of sensor response with enzyme activity. Results show the following red shifts; 0.05 nmol + 0 nm; 0.1 nmol + 0 nm; 0.5 nmol + 4.4 nm; 1.0 nmol +10.4 nm; 1.9 nmol +18.9 nm. Optical response is nearly linear changing ∼10 nm/mol of GST. The microcavity detection limit appears to lie between 0.1 and 0.5 nmol of GST. Standard deviation error on the microcavity red shift values was determined from measurements made at various percent reflection across the wavelength spectrum ranging between 575 and 635 nm.
advantages of the high internal surface area or the optical response properties of PSi, they did show how tailoring surface chemistry can enable a high spot density (4400 spots/cm2) and a low consumption of capture antibodies (0.14 fmol/spot) to yield picomolar detection sensitivity in complex biological solutions. To characterize the p+ mesoporous microcavity sensitivity, defined as the minimum amount of protein required to transduce a threshold optical response, a unique approach was taken by exploiting the intrinsic activity of immobilized GST enzyme. Following immobilization of 50 µL of GST (0.05-2.0 mg/mL, ∼1 to ∼40 µM) the magnitude of the optical response was cross correlated to enzyme activity. A optical detection limit in the lowmicromolar range (2-10 µM) was measured; however, the
enzyme assay proved to be at least 1 order of magnitude more sensitive (1 mm thick with an average pore diameter of 5.5 nm. Campbell and co-workers51 also showed from mass transport studies on planar surfaces that the time required for target binding is proportional to the inverse square of the solution concentration. This situation is expected to be exacerbated for diffusion and binding in a porous matrix particularly when the biomolecular size is of a dimension similar to that of the pore diameter.52 We (51) Jung, L. S.; Nelson, K. E.; Stayton, P. S.; Campbell, C. T. Langmuir 2000, 16, 9421-9432. (52) Charles, P. T.; Goldman, E. R.; Rangasammy, J. G.; Schauer, C. L.; Chen, M. S.; Taitt, C. R. Biosens. Bioelectron. 2004, 20 (4), 753-764.
Analytical Chemistry, Vol. 77, No. 10, May 15, 2005
3229
hypothesize that given the wide pore size distribution in our microcavities (5-45 nm) there may be ample space for GST (∼10nm diameter) to infiltrate homogeneously, but at low GST concentrations molecules may preferentially bind near the microcavity surface in the top mirror layers. Optical response simulations show the microcavity resonance peak shifts in a layer-bylayer fashion as the index of refraction is altered in sequential layers. Because of the resonance property of the microcavity, reflected light is concentrated in the active central layer and a proportionally larger shift results when the refractive index of this layer is changed. Three strategies can be taken to improve pore infiltration, increase pore diameter, use a flow cell, and design microcavities to operate in the IR. A new device architecture that could permit the active flow of bioreagents parallel to pore channels through a free-standing membrane is preferred. While this architectural improvement is forthcoming, studies are currently underway investigating sensitivity limits employing Escherichia coli Initimin/Tir (Kd ∼ 0.3 µM)53,54 streptavidin/biotin (Kd∼ 1.0 pM), and IgG/anti-IgG (Kd ∼1.0 nM)55 systems using n+ microcavity macroporous (>50 nm) microcavity devices. Initial efforts to study these systems using mesoporous microcavities have proved problematic as the pore diameters (∼20 nm) do not permit facile diffusion of proteins exceeding ∼60 kDa. These studies will enable us to assess the effect of nonspecific binding on affinity detection sensitivity and to draw conclusions regarding the dependence of Kd on detection sensitivity. The present work has demonstrated that progress is being made in gaining a fundamental insight into the factors that (53) Gauthier, A.; de Grado, M.; Finlay, B. B., Infect. Immun. 2000, 68 (7), 43444348. (54) Luo, Y.; Frey, E. A.; Pfuetzner, R. A.; Creagh, A. L.; Knoechel, D. G.; Haynes, C. A.; Finlay, B. B.; Strynadka, N. C. J. Nature 2000, 405, 1073-1077. (55) Welschof, M.; Terness, P.; Kipriyanov, S. M.; Stanescu, D.; Breitling, F.; Do ¨rsam, H.; Du ¨ bel, S.; Little, M.; Opelz, G. Proc. Natl. Acad. Sci. U.S.A. 1997 94 (5), 1902-1907.
3230
Analytical Chemistry, Vol. 77, No. 10, May 15, 2005
influence the detection limits of porous silicon microcavities, which have emerged as promising new optical biosensing technology. Much research remains, however, to understand whether the apparent requirement that 0.1-0.6-µg of target must bind the device to transduce a signal is in fact limiting. The ability to tune the wavelength of operation, percent porosity, and pore diameter are suggested by our data to be significant parameters for impacting sensitivity. In conclusion, we believe that the number of grams of target/transducer area (internal for a porous silicon device) required to generate a threshold signal is an appropriate metric to define sensitivity for all chip-based detection modalities as it eliminates ambiguity associated with molarity, stemming from lack of standardized protocols. ACKNOWLEDGMENT The authors acknowledge Huimin Ouyang and Dr. Alice Pentland for helpful discussions. L.A.D. acknowledges financial support from the NIH Dermatology training grant (5T32AR07472). Equipment used in this research was purchased with support from the Infotonics Technology Center (524051). SUPPORTING INFORMATION AVAILABLE SEM images are presented and analyzed using NIH ImageJ to calculate the average pore diameters for the microcavity structures utilized in this study. Also, an equilibrium simulation is presented illustrating how binding affinity (Kd) influences detection sensitivity and how sensitivity can be improved by raising the receptor concentration. This material is available free of charge via the Internet at http://pubs.acs.org.
Received for review December 15, 2004. Accepted March 4, 2005. AC048144+
