Self-Assembled TNT Biosensor Based on Modular ... - ACS Publications

We demonstrate a self-assembled reagentless biosensor based on a modular design strategy that functions in the detection of TNT and related explosive ...
1 downloads 0 Views 373KB Size
Download PDF
Anal. Chem. 2005, 77, 365-372

Accelerated Articles

Self-Assembled TNT Biosensor Based on Modular Multifunctional Surface-Tethered Components Igor L. Medintz,*,† Ellen R. Goldman,† Michael E. Lassman,† Andrew Hayhurst,‡ Anne W. Kusterbeck,† and Jeffrey R. Deschamps§

Center for Bio/Molecular Science and Engineering, Code 6900, and Laboratory for the Structure of Matter, Code 6030, U.S. Naval Research Laboratory, Washington D.C. 20375, and Department of Virology and Immunology, Southwest Foundation for Biomedical Research, 7620 N. W. Loop 410, San Antonio, Texas 78227-5301

We demonstrate a self-assembled reagentless biosensor based on a modular design strategy that functions in the detection of TNT and related explosive compounds. The sensor consists of a dye-labeled anti-TNT antibody fragment that interacts with a cofunctional surface-tethered DNA arm. The arm consists of a flexible biotinylated DNA oligonucleotide base specifically modified with a dye and terminating in a TNB recognition element, which is an analogue of TNT. Both of these elements are tethered to a Neutravidin surface with the TNB recognition element bound in the antibody fragment binding site, bringing the two dyes into proximity and establishing a baseline level of fluorescence resonance energy transfer (FRET). Addition of TNT, or related explosive compounds, to the sensor environment alters FRET in a concentration-dependent manner. The sensor can be regenerated repeatedly through washing away of analyte and specific reformation of the sensor assembly, allowing for subsequent detection events. Sensor dynamic range can be usefully altered through the addition of a DNA oligonucleotide that hybridizes to a portion of the cofunctional arm. The modular design of the sensor demonstrates that it can be easily adapted to detect a variety of different analytes.

Increasing demand is being placed on the biosensor community to deliver sensors for a wide range of applications including health care, military and security purposes, food and water * Corresponding author. Phone: 202-404-6046. Fax: 202-767-9594. E-mail: [email protected]. † Center for Bio/Molecular Science and Engineering, U.S. Naval Research Laboratory. ‡ Southwest Foundation for Biomedical Research. § Laboratory for the Structure of Matter, U.S. Naval Research Laboratory. 10.1021/ac048485n CCC: $30.25 Published on Web 12/21/2004

© 2005 American Chemical Society

assurance, and environmental and industrial process monitoring to name but a few.1-8 Rather than developing a unique sensor design for each analyte of interest, generalized strategies are needed that are readily adaptable to a variety of small analytes in many challenging environments. Optically addressable biosensors have been developed, in part, to address many of these needs. One family of these biosensors utilizes the allosteric coupling of fluorescence and binding events engineered into bacterial periplasmic binding proteins (bPBPs) for reagentless signal transduction.9-12 Use of these proteins has even been extended from solution phase to surface tethering.13 bPBP-based sensors have also undergone computationally intensive redesign of the binding sites to adapt recognition to alternate substrates14 or have been fused between fluorescent proteins to create elegant intracellular fluorescence resonance energy-transfer (FRET)-based analyte (1) Sadik, O. A.; Wanekaya, A. K.; Andreescu, S. J. Environ. Monit. 2004, 6, 513-22. (2) Wolfbeis, O. S. Anal. Chem. 2004, 76, 3269-83. (3) Monk, D. J.; Walt, D. R. Anal. Bioanal. Chem. 2004, 379, 931-45. (4) Scheller, F. W.; Wollenberger, U.; Warsinke, A.; Lisdat, F. Curr. Opin. Biotechnol. 2001, 12, 35-40. (5) Iqbal, S. S.; Mayo, M. W.; Bruno, J. G.; Bronk, B. V.; Batt, C. A.; Chambers, J. P. Biosens. Bioelectron. 2000, 15, 549-78. (6) Baeumner, A, J. Anal. Bioanal. Chem. 2003, 377, 434-45. (7) Nakamura, H.; Karube, I. Anal. Bioanal. Chem. 2003, 377, 446-68. (8) D’Amico, E. Chem. Week 2002, 164, 25-6. (9) Marvin, J. S.; Hellinga, H. W. Nat. Struct. Biol. 2001, 8, 795-8. (10) De Lorimier, R. M.; Smith, J. J.; Dwyer, M. A.; Looger, L. L.; Sali, K. M.; Paavola, C. D.; Rizk, S. S.; Sadigov, S.; Conrad, D. W.; Loew, L.; Hellinga, H. W. Protein Sci. 2002, 11, 2655-75. (11) Benson, D. E.; Conrad, D. W.; De Lorimier, R. M.; Trammell, S. A.; Hellinga, H. W. Science. 2001, 293, 1641-4. (12) Dwyer, M. A.; Hellinga, H. W. Curr. Opin. Struct. Biol. 2004, 14, 495504. (13) Wada, A.; Mie, M.; Aizawa. M.; Lahoud, P.; Cass. A. E.; Kobatake, E. J. Am. Chem. Soc. 2003, 52, 16228-34. (14) Looger, L. L.; Dwyer, M. A.; Smith, J. J.; Hellinga, H. W. Nature 2003, 423, 185-90.

Analytical Chemistry, Vol. 77, No. 2, January 15, 2005 365

Figure 1. (A) Schematic of the modular biosensor consisting of two modules: the biorecognition module and the modular arm. Both modules are specifically attached to a surface in a particular orientation. Choices for surface attachment include biotin-avidin chemistry, metal-affinity coordination, thiol bonding, hydrophobic interactions, DNA-directed immobilization, etc.13,18 The biorecognition module can consist of proteins (enzymes, receptors, bPBPs, antibody fragments, peptides), aptamers, carbohydrates, DNA, PNA, RNA, etc. This module is site-specifically dye-labeled in the current configuration. The modular arm may consist of flexible moieties such as DNA, PNA, RNA, peptides, polymers, etc. The modular arm is also site-specifically dye-labeled. An analogue of the primary analyte is attached to the distal end of the flexible arm to act as the recognition analogue. Binding of this recognition element in the binding pocket of the biorecognition element assembles the sensor into the ground state by bringing both dyes into proximity, which establishes FRET. Addition of analyte competitively displaces the analogue and signal transduction is designed to be sensitive to this displacement. FRET donor/acceptor can be placed on either module. Mechanisms of controlling binding affinity include stiffening the flexible arm or switching in of different affinity biorecognition elements. (B) Schematic of the TNT targeting biosensor. The dye-labeled anti-TNT scFv fragment (1, the biorecognition module) is attached to the surface with Bio-X-NTA coordinating the 12 histidines (12-HIS) and orienting the protein on the NA. The dye-labeled TNB DNA arm (2, modular arm) is attached to the NA via complementary hybridization to a biotinylated (B) flexible DNA linker (see Figure 2A). Both are added in equimolar amounts. ScFv binding of the TNB analogue brings the protein located dye and DNA located dye into proximity establishing FRET. Addition of TNT displaces the TNB analogue and DNA arm disrupting FRET in a concentration-dependent manner. The sensor is washed and regenerated for subsequent detection events. Rigidifying the DNA arm modulates sensing.

monitors.15-17 However, the number of protein bioreceptors that undergo the necessary ligand-dependent conformational changes (15) Fehr, M.; Frommer, W. B.; Lalonde, S. Proc. Natl. Acad. Sci. U.S.A. 2002, 99, 9846-51. (16) Fehr, M.; Lalonde, S.; Lager, I.; Wolff, M. W.; Frommer, W. B. J. Biol. Chem. 2003, 278, 19127-33. (17) Fehr, M.; Ehrhardt, D. W.; Lalonde, S.; Frommer, W. B. Curr. Opin. Plant Biol. 2004, 7, 345-51.

366 Analytical Chemistry, Vol. 77, No. 2, January 15, 2005

needed for this type of signal transduction, or are amenable to redesign, is limited; therefore, the need remains for biosensors that are readily adaptable to a variety of small analyte targets. We have previously described a general biosensing strategy that is based on a modular design (see Figure 1A for description) and employs multifunctional surface-tethered components. Biosensors built on this strategy are designed to be fully reversible

and reagentless and self-assemble on surfaces.18 The first prototype developed to test the sensor design consisted of dye-labeled biotin-linked Escherichia coli maltose binding protein (MBP) bound to a Neutravidin (NA) surface in a specific orientation (biorecognition module; see Figure 1). This bioreceptor interacted with a modular tether arm, also bound to the NA surface, which consisted of a flexible biotinylated DNA oligonucleotide, a fluorescent dye, and a distally affixed β-cyclodextrin maltose analogue. After surface tethering of both elements, the sugar analogue is bound in the MBPs’ sugar binding site, bringing both dyes into proximity and establishing a baseline level of FRET. FRET is the signal transduction of choice in our sensor assembly due to its exquisite sensitivity to molecular rearrangements on the scale of most biological macromolecules such as proteins.4,5,7,9,10,18,19 Addition of target analyte, maltose, displaced the tethered sugar altering FRET in a concentration-dependent manner. The sensor could be regenerated several times by washing away analyte. MBP solution-phase sensitivity and specificity was retained in this tethered assembly. Sensor dynamic range could be altered by adding a complementary strand of DNA that “stiffened” a portion of the cofunctional modular arm.18 The overall design of the sensor is modular with the idea of facile adaptation to target other analytes and is designed to be amenable to multiple chemistries and different types of components; see Figure 1A. In the current work, we demonstrate that this self-assembled modular sensing strategy, used in the initial MBP-based prototype, can be applied to the detection of an analyte completely unrelated to maltose, namely, 2,4,6-trinitrotoluene (TNT), by altering two modules of the sensor assembly. Using the same sensor platform, we first substitute a dye-labeled anti-TNT single-chain Fv antibody fragment (R-TNTscFv) for the MBP bioreceptor portion of the sensor and modify the multifunctional tether arm to contain a dye attached to a modified internal base. The DNA arm is further modified to terminate with the TNT analogue 1,3,5-trinitrobenzene (TNB) (see Figures 1 and 2) The R-TNTscFv portion is allowed to bind the TNB analogue, bringing both dyes into proximity, establishing a baseline level of FRET, and this complex is then self-assembled on a NA surface. Addition of TNT to the sensor solution results in a concentration-dependent change in FRET. In this configuration, the sensor retains its analyte specificity yet can still have its dynamic sensing range usefully adjusted. The sensor can be washed free of analyte and reforms for subsequent detection events. The demonstration that this sensor can be readily adapted to a completely different analyte highlights the utility of its modular construction and its general applicability to a variety of targets. MATERIALS AND METHODS Reagents. Explosive standards including TNT, TNB, hexahydro-1,3,5-trinitro-1,3,5-triazine (RDX), 2-amino-4,6-dinitrotoluene (2A-4,6-DNT), and 2,4-dinitrotoluene (2,4-DNT) were obtained from Cerilliant Corp. (Austin, TX). Reactibind Neutravidin-coated plates were obtained from Pierce (Rockford, IL). (18) Medintz, I. L.; Anderson, G. P.; Lassman, M. E.; Goldman, E. R.; Bettencourt, L. A.; Mauro, J. M. Anal. Chem. 2004, 76, 5620-9. (19) Lakowicz, J. R. Principles of Fluorescence Spectroscopy; Kluwer Academic/ Plenum Publishers: New York, 1999.

Mutagenesis and Protein Expression. The R-TNTscFv fragment, a derivative of TNB2 previously described,20 was engineered to express a (His)6GlyGlySerGlyGly(His)6 carboxy terminus, where (His)6 ) 6-histidine residue sequence.21 Selected cysteine mutations were introduced into the DNA coding sequence using the Quickchange Site-Directed Mutagenesis Kit (Stratagene, La Jolla, CA). For mutagenesis, polyacrylamide gelpurified oligonucleotides were obtained from Sigma Genosys (Woodlands, TX). Transformants were selected, and correct mutational residue changes were verified by DNA sequencing. R-TNTscFv-Cys mutants were expressed in the E. coli Tuner strain (Novagen, San Diego CA) and purified from the periplasm as described.20 R-TNTscFv-Cys mutant samples were labeled with maleimide-activated AlexaFluor 532 dye (AFF 532: quantum yield 0.8, molar extinction coefficient 81 000 M-1 cm-1; Molecular Probes, Eugene OR, see Supporting Information for structure) as described,22 with the modification of using 100× less of the dithiothreitol reducing agent.23,24 Dye-to-protein ratios of ∼1 were obtained. Construction and purification of (His)6-appended myoglobin, with a cysteine mutation at position 64, and its subsequent labeling with Cy3-maleimide (Amersham Bioscience, Piscataway, NJ), was as described.21,22 ELISAs for determining TNB binding were performed on the R-TNTscFv-Cys mutants as described.20 ScFv Antibody Fragment Modeling. In the absence of direct structural information on the R-TNTscFv and its polyhistidine tail, a homology model was formulated. A starting model was selected based on the degree of sequence homology between R-TNTscFv and scFv crystal structures in the Protein Data Bank. An alignment of the sequence of 1DZB,25 a phage-derived ScFv fragment complexed with lysozyme, and R-TNTscFv was prepared, and the conserved (invariant) regions of the latter were identified. Locations of point mutations were selected based upon solvent exposure and proximity to the antigen binding site. The structure of the histidine-rich tail was constructed using Chem-3D Ultra and incorporated into the PDB-derived structure using MidasPlus, which was also used for final rendering.26,27 After energy minimization using the MOPAC module included with Chem-3D Ultra, the histidine-rich tail was folded to allow interaction of the histidine side chains with the surface in accordance with prior studies.21 Dye-Labeled TNB DNA Arm. The precursor of the dyelabeled TNB DNA arm (Figure 2A, in red) was purchased from Qiagen (Alameda, CA) already modified with a teramethylrhodamine (TAMRA) dye-labeled internal T* base (quantum yield 0.7, molar extinction coefficient 91 000 M-1 cm-1) and a 5′-amino modification on a C-6 linker. See Figure 2B for AFF 532 spectral overlap with TAMRA. The DNA precursor was subsequently (20) Goldman, E. R.; Hayhurst, A.; Lingerfelt, B. M.; Iverson, B. L.; Georgiou, G.; Anderson, G. P. J. Environ. Monit. 2003, 5, 380-3. (21) Goldman, E. R.; Medintz, I. L.; Whitley,J. L.; Hayhurst, A.; Uyeda, H. T.; Deschamps, J. R.; Lassman, M. E.; Mattoussi. J. Am. Chem. Soc. Submitted. (22) Medintz, I. L.; Goldman, E. R.; Lassman, M. E.; Mauro, J. M. Bioconjugate Chem. 2003, 14, 909-18. (23) Albrecht, H.; Burke, P. A.; Natarajan, A.; Xiong, C. Y.; Kalicinsky, M.; DeNardo, G. L.; DeNardo, S. J. Bioconjugate Chem. 2004, 15, 16-26. (24) Schmiedl, A.; Breitling, F.; Winter, C. H.; Queitsch, I.; Dubel, S. J. Immunol. Methods 2000, 242, 101-14. (25) Ay, J.; Keitel, T.; Kuttner, G.; Wessner, H.; Scholz, C.; Hahn, M.; Hohne, W. J. Mol. Biol. 2000, 301, 239-46. (26) Ferrin, T. E.; Huang, C. C.; Jarvis, L. E.; Langridge, R. J. Mol. Graphics 1988, 6, 13-27. (27) Huang, C. C.; Pettersen, E. F.; Klein, T. E.; Ferrin, T. E.; Langridge, R. J. Mol. Graphics 1991, 9, 230-6.

Analytical Chemistry, Vol. 77, No. 2, January 15, 2005

367

Figure 2. (A) Structure of the dye-labeled TNB DNA arm. The precursor DNA modified with a TAMRA dye on an internal T* and the 5′-amino modifier C6 is shown in red. The amine was subsequently modified with TNB (blue). The biotinylated hybridizable flexible DNA linker and modulator DNA aligned with its complement are shown in black. (B) Dye absorption and emission spectra (R0, distance for 50% energy-transfer efficiency, for the AlexaFluor 532 (AFF 532) donor and TAMRA acceptor is ∼63 Å).19

reacted with a 1000-fold molar excess of 2,4,6-trinitrobenzenesulfonic acid (5% solution, Sigma, St. Louis, MO), Figure 2A blue, in 1 mL of 0.136 M sodium tetraborate buffer pH 8.5 supplemented with 20 µL of 2 M NaOH. The resulting solution was reacted overnight at room temperature under continuous agitation, then loaded on a Supelclean LC-18 SPE column (Supelco, Bellefonte PA), washed with 0.1× borate buffer, and eluted with an increasing concentration of methanol in borate buffer. The mass of the final product (MW ∼5878) was verified on an Applied Biosystems API QSTAR Pulsar mass spectrometer by positive electrospray ionization.18,22 Sensor Assembly. For each well of the Reactibind NA-coated plates (binding capacity ∼60 pmol of biotin), 30 pmol of dyelabeled TNB DNA arm (Figure 2A red and blue) was added to 30 pmol of hybridizable flexible DNA linker (Figure 2A black), heated to 90 °C for 5 min, and then cooled to room temperature slowly to preclude secondary structure formation.18 The DNA melting temperature is ∼62 °C. Concurrently, 30 pmol of AlexaFluor 532labeled R-TNTscFv was mixed with 30 pmol of biotin-X nitrilo368

Analytical Chemistry, Vol. 77, No. 2, January 15, 2005

triacetic acid tripotassium salt precharged with nickel (Bio-X NTA, Biotium, Hayward CA, Bio-biotin, X-aminomethoxy spacer, NTAnitrilotriacetic acid chelator, see SupporingInformation for structure). After 2 h, DNA and protein solutions were mixed together (now equimolar), diluted with phosphate-buffered saline pH 7.4 (PBS) to a final volume equivalent to 50 µL/well. and incubated for 4 h at room temperature. A 50-µL aliquot of sensor solution was added to each well of the NA plates and incubated at 4 °C overnight. For regeneration, sensor-coated plates were washed 10× with 200 µL of PBS and reconstituted in 50 µL of PBS for subsequent retitration. For experiments utilizing modulator DNA, a 1:1 molar ratio of modulator DNA (Figure 2A, black) was added to the primary DNA solution (DNA melting temperature is ∼55 °C). Fluorometry and Titration. Fluorometric analysis was performed on a Safire dual monochromator multifunction microtiter plate reader (Tecan, Boston, MA). Samples were excited at 510 nm, and emission was recorded at 600 nm. A 5-µL aliquot of 2× solution containing diluted explosive was added to each well at

Figure 3. Model structure of the R-TNT scFv fragment. The (His)6GlyGlySerGlyGly(His)6 carboxy terminus is shown in green. The sites of cysteine mutation closest to the carboxy terminus are Ala241 in yellow, Ala210 in red, and Leu145 in magenta. The Gln13 site in red is shown at the bottom of the structure. The “TNT binding region” as defined by the hypervariable regions of the scFv is cyan. Inset: cartoon of the R-TNT scFv fragment schematically demonstrating critical intrachain disulfide bonds, which stabilize the VH and VL domains. The fragment is attached to the NA surface via the C-terminus by the Bio-X-NTA, suggesting that the fragment is oriented with the binding site face up.

binding constants Kapp were estimated from the second derivative of titration curves and rounded to whole integers.

Figure 4. Repetitive titrations of the sensor assembly with TNT. The sensor was assembled as described in Materials and Methods, titrated against TNT, washed, regenerated, and titrated again. Samples were excited at 510 nm, and emission was monitored at 600 nm. The change in fluorescence is plotted on the vertical axis versus TNT concentration in the horizontal axis. TNT concentrations are in mg/L which corresponds to ppm. In this configuration, a maximum of 6-8 consecutive titrations were attained.

room temperature for 5 min to give a final concentration in 100 µL as indicated in Figures 4 and 5. Each point of a titration was performed in triplicate (three separate wells), and FRET efficiency at 600 nm was monitored during experiments. Titrations are plotted against the change (loss) in fluorescence compared to the 0 concentration point. In the figures, error bars appear where appropriate; if not, the error is within the point value. Approximate

RESULTS ScFv Antibody Fragment Cysteine Mutants. A preliminary sensor was assembled using a heterogeneously labeled nonmutated R-TNTscFv protein. The protein was labeled nonspecifically with a NHS-ester activated QSY9 dark quencher dye (Molecular Probes), which targets the multiple amine functionalities. Although a clear change in fluorescence occurred upon TNT addition, the results indicated a great deal of heterogeneity was present due to the high background with large error bars and well-to-well inconsistencies (data not shown). Thus, a homogeneously labeled R-TNTscFv protein and sensor was desirable. Site-specific labeling of proteins is easily accomplished on unique cysteine residues.28 The current R-TNTscFv protein already contains four cysteines, which are required to form intrachain disulfide bonds to stabilize the variable heavy (VH) and variable light (VL) domains and help maintain the proteins’ integrity and recognition function.23,24 Since the introduction of additional cysteines could result in detrimental disulfide bond formation during the expression and folding process and a nonfunctional “disulfide scrambled” protein, an intuitive approach based on modeling and design was undertaken. A model was constructed of the R-TNTscFv protein (see Materials and Methods) for the determination of optimum mutational sites. Selection criteria included surface-exposed residues to facilitate labeling, proximity to peptide turns, location distal from the binding site, location distal (28) Hermanson, G. T. Bioconjugate Techniques; Academic Press: New York, 1996.

Analytical Chemistry, Vol. 77, No. 2, January 15, 2005

369

Figure 5. (A) Sensor specificity. Sensor assemblies were titrated against the indicated explosives and analogues as described in Materials and Methods. See Supporting Information for explosive structures. (B) Modulated sensor titration. The sensor was assembled with the addition of the modulator DNA (see Figures 1 and 2) and titrated against TNT resulting in a lower limit of detection of 0.1 mg/L or 100 ppb.

from the already present cysteines, and selection of residues, which, if changed, would have minimal affect on overall structure. The four sites selected were Gln13, Leu145, Ala210, and Ala241; see Figure 3. Transformants were obtained for the first three mutational sites. The Ala241 site is proximal to the (His)6GlyGlySerGlyGly(His)6, region and this may have interfered with mutagenic oligonucleotide hybridization. After DNA sequence verification of the three mutants obtained, growth and induction revealed that only the strains expressing the R-TNTscFv Ala210Cys and Leu145Cys mutants produced any protein expression. TNB binding was assayed by ELISA as described,20 and results indicted that only the R-TNTscFv-Leu145Cys mutant protein had any appreciable TNB binding capacity. This mutant was subsequently dye-labeled with AFF 532 and still retained its TNB binding capacity when reassayed using ELISA. Sensor Assembly, Titration versus TNT, and Regeneration. The sensor was assembled as described in Materials and Methods using the AFF 532-labeled R-TNTscFv-Leu145Cys protein. Attempts at tethering the dye-labeled TNB DNA arm and the labeled protein separately resulted in low amounts of functional sensor assembly; thus, both elements were prebound prior to tethering to the NA surface. This is in contrast to the maltose targeting prototype, which was assembled sequentially. By pre370

Analytical Chemistry, Vol. 77, No. 2, January 15, 2005

binding prior to tethering, more active sensor sites consisting of both elements in functional orientations were attached to the surface and available for regeneration after washing. Following sensor self-assembly, TNT in solution was titrated at the concentrations indicated (see Figure 4). The resulting loss of FRET-based fluorescence at 600 nm was plotted against the concentration. TNT concentrations higher than 50 mg/L were not used, as TNT solubility in aqueous solutions saturates at concentrations approaching 100 mg/L.29 A lower limit of detection of 1 mg/L TNT (1 part per million/ppm) was noted for this sensing assembly. The binding curve appears to be a relatively linear function over this series of concentrations tested. The sensor assembly was washed using 10 volumes of 200 µL PBS and then regenerated in 50 µL of PBS for 24 h at 4 °C. A second TNT titration was performed, the sensor washed and then regenerated for 1 h at room temp, and then a third titration was performed (See Figure 4). Similar control experiments performed using sensor assemblies incorporating a dye-labeled DNA arm lacking the TNB moiety or substituting Cy3-labeled myoglobin for the R-TNTscFv resulted in no change in fluorescence upon TNT addition (data not shown). Additionally, the 600-nm TAMRA emission for these control assemblies was substantially lower, also indicative of inefficient FRET. These results confirm that the same sensor binding of the terminal DNA arm recognition analogue and sensor regeneration performance seen in the maltose sensing prototype is present in the current TNT sensing assembly. Sensor Specificity and Modulation. To determine specificity and cross-reactivity, sensor assemblies were tested against a variety of TNT structural analogues as described in Figure 5 (see Supporting Information for chemical structures). Not surprisingly, TNB elicited the highest sensor response in terms of absolute fluorescence change, followed closely by TNT. This result can be attributed to the fact that the antibody fragment was originally selected against TNB and the analogue attached to the DNA arm is also TNB.20 Thus, TNB will be the most effective competitor for binding sites. RDX and 2-A-4,6-DNT caused significantly less response, with 2,4-DNT eliciting the least response. Since 2-A4,6-DNT elicits all of its response between 5 and 20 mg/L, the result is almost a complete “classical” binding curve in comparison to the results elicited from the other explosive compounds. It should be noted that, in this format, TNT cannot be distinguished from the structurally related competitors until higher concentrations. As mentioned, issues of solubility limit the concentration range that can be tested for these compounds, limiting them to the “linear” portion of the curve. Nevertheless, these results very closely parallel the competition ELISA format used to test the precursor anti-TNB scFv selected.20 This demonstrates that the original R-TNTscFv specificity is retained even after multiple modifications including the following: (1) appending a (His)6spacer-(His)6, (2) point mutation to a cysteine, (3) dye labeling of the cysteine, (4) surface tethering, and (5) interacting with the dye-labeled TNB DNA arm as part of sensing. This also parallels the specificity retained by the maltose sensing prototype after similar modifications. The dynamic sensing range of the maltose sensing prototype could be modulated by “stiffening” of the tether arm through the (29) Adrian, N.; Campbell, E. U.S. Army Corps of Engineers, Engineer Research and Development Center, CERL Technical Report 99/102, 1999; p 24.

Table 1. Properties of the Sensors Used in This Studya

sensor tested against

Kapp, µM

lower limit of detection, mg/L (ppm, S/N >3)

% Fl increase upon sensor saturationb

TNT TNB 2-amino-4,6-DNT RDX 2,4-DNT TNT(+ modulator DNA)

11 ( 2 15 ( 1 16 ( 3 15 ( 2 13 ( 1 12 ( 1

1 3.5 1 5 7.5 0.1

15 25 11 12 8 21

a All readings are performed at 20-25 °C. See Figure 1 for schematic. b Calculated as (Flsat - Fl0)/Fl0 × 100, where Flsat is the fluorescence at saturation and Fl0 is the fluorescence before the addition of analyte.18

addition of a DNA complementary to the DNA linker.18 Although the binding constant essentially remained the same, the binding curve of the modulated prototype became broader, significantly increasing the useful sensing range. Similar experiments were performed with the current TNT sensing assembly. During selfassembly, modulator DNA was added to the hybridizable flexible DNA linker (Figure 2A black), significantly increasing the amount of double-stranded DNA from 16 of 44 bases (∼35%) in the unmodulated sensor to 36/44 (∼80%) when modulated. The lower limit of TNT detection of the modulated sensor dropped ∼10-fold from 1 to 0.1 mg/mL (1 ppm to 100 parts per billion/ppb); see Table 1 and Figure 5B. In parallel to the prototype, these results show that the modulation mechanism can be exploited to adjust useful sensor properties. The choice of C6 chain, 5′-amino-C6 and 15-carbon spacer in the DNA arm was driven by availability for incorporation during automated DNA synthesis. It should be noted that adjusting the length of the single-stranded DNA arm (as opposed to rigidity) in the maltose sensing prototype did not effect binding properties.18 Thus, adjusting these alkane moieties in terms of length were not considered critical in the current sensor as well. “Switching in” of a more sensitive scFv antibody fragment would be another mechanism of adjusting sensor properties.18 DISCUSSION AND CONCLUSIONS The TNT sensing assembly we demonstrate retains almost all of the desirable properties of the MBP-based prototype while targeting a different analyte. For discussion of current methods of detecting TNT, the interested reader is directed to refs 30-34. Among the useful features of this sensor are the use of robust avidin-biotin technology for surface tethering and the ease of self-assembly in a microtiter plate format, which also facilitates analysis by readily accessible fluorescent plate readers. The prototype and current sensor use dye-based FRET; however, signal transduction may be expanded to other proximity-sensitive methods such as electrochemical detection.11 Although not (30) Steinfeld, J. I.; Wormhoudt, J. Annu. Rev. Phys. Chem. 1998, 49, 203-32. (31) Marazuela, M. D.; Moreno-Bondi, M. C. Anal. Bioanal. Chem. 2002, 372, 664-82. (32) Ambayah, M.; Quickenden, T. I. Talanta 2004, 63, 461-467. (33) Yinon, J. TrAC, Trends Anal. Chem. 2002, 21, 292-301. (34) Charles, P. T.; Shriver-Lake, L. C.; Francesconi, S. C.; Churilla, A. M.; Rangasammy, J. G.; Patterson, C. H.; Deschamps, J. R.; Kusterbeck, A. W.; J. Immun. Methods 2004, 284, 15-26.

investigated here, the choices of FRET dyes and their locations on either module are also interchangeable, which can allow optimization to address any potential distance requirements.18 Furthermore, the sensor can be regenerated for subsequent reuse, in these experiments, an upper limit of 6-8 regenerations was obtained. From the results presented here, regeneration in less than 1 h is feasible (Figure 4). Sensor response in the current microtiter well format is limited by technician handling time, which suggests that real-time sensing in a flow cell may be feasible. The major focus of the current study was to demonstrate the modularity of the sensor design by readily adapting it to target another analyte. As such, the sensitivity of the biorecognition element was not a major criterion. However, this does not preclude the current sensor from actual use as a solution-phase TNT sensor. More sensitive (His)6-appended TNT recognition elements are available14 that demonstrate nanomolar sensitivity. The current R-TNTscFv fragment could undergo another round of evolution to select for higher affinity mutants. Paradoxically, increasing the affinity to increase sensitivity may be a potential liability of this format. If the affinity of the biorecognition entity for the analogue/ analyte is too high, displacement, regeneration, or both may take far too long for useful sensing (seconds/minutes vs hours/days). Although the modulated behavior of the sensor was modest, ∼10fold decrease in limit of detection, the conservation of this feature seen in the prototype is important. In parallel to the prototype, we believe that the addition of DNA to stiffen the arm creates sensors that are more sensitive to perturbation, thus increasing sensitivity and yielding a lower limit of detection.18 Due to the issues of TNT solubility, the same broadening of the dynamic sensing range with DNA stiffening could not be fully tested. However, this result suggests that modulation through careful control of arm stiffness/kinetics may be a method that can be further exploited to yield sensing assemblies with variable control over the desired affinities. The combination of “switching in” of mutant proteins with different affinities and sensor arm modulation may also potentiate the sensitivity attainable. The most important result of the current study is that it validates the modular nature and adaptability of this sensor design (see Figure 1A). As demonstrated, the current assembly can easily be adapted to target another analyte and yet remain functionally robust. Choices of biorecognition elements that can fit directly into the current format include (His)6-appended proteins such as bPBPs, scFv fragments, or even soluble fragments of cloned cellular receptors. The (His)6 motif is commonly engineered into many cloned proteins for facile purification over Ni-NTA media.35 Additionally, an analogue of the primary analyte of interest must be attached to the distal end of the DNA arm and many covalent and noncovalent chemical linking strategies are available to address this.28 A variety of choices are available for both the biorecognition module and the modular arm, as described in the Figure 1 legend. However, analyte size may be limited to smaller molecules due to the need to tether an analogue, as well as intrinsic FRET constraints. Current research focuses on adapting this sensor for testing in a drug discovery assay where a variety of compounds with differing affinities for the biorecognition module can be screened in parallel. Multianalyte or “multiplex” (35) Drees, J.; Smith, J.; Schafer, F.; Steinert, K. Methods Mol. Med. 2004, 94, 179-90.

Analytical Chemistry, Vol. 77, No. 2, January 15, 2005

371

sensing is another possibility. Other applications may include providing a variety of sensors in a microtiter well format for screening or quantitation of many different analytes such as in a clinical setting. Experiments with the maltose sensing prototype showed that sensor assemblies could be dried, stored, and reconstituted for later use.18 This suggested that sensors could be assembled on microtiter plates and dried/stored for later use in other locations. Although demonstrated in a microtiter well plate format, assembly of these sensors in flow cells and on nanoparticle surfaces also remains an interesting avenue of future research. ACKNOWLEDGMENT This research was funded by the Office of Naval Research. I.L.M. was partially supported by a National Research Council

372

Analytical Chemistry, Vol. 77, No. 2, January 15, 2005

fellowship through the Naval Research Laboratory. M.E.L. was supported by an American Society for Engineering Education (ASEE) fellowship. A.H. acknowledges support from the USAA Foundation. SUPPORTING INFORMATION AVAILABLE Additional information as noted in the text. This material is available free of charge via the Internet at http://pubs.acs.org.

Received for review November 18, 2004. AC048485N

October

13,

2004.

Accepted