PAPER

www.rsc.org/analyst | Analyst

Multiplexing of injury codes for the parallel operation of enzyme logic gates Jan Hal amek,a Joshua Ray Windmiller,b Jian Zhou,a Min-Chieh Chuang,b Padmanabhan Santhosh,b Guinevere Strack,a Mary A. Arugula,a Soujanya Chinnapareddy,a Vera Bocharova,a Joseph Wang*b and Evgeny Katz*a Received 27th April 2010, Accepted 15th June 2010 DOI: 10.1039/c0an00270d The development of a highly parallel enzyme logic sensing concept employing a novel encoding scheme for the determination of multiple pathophysiological conditions is reported. The new concept multiplexes a contingent of enzyme-based logic gates to yield a distinct ‘injury code’ corresponding to a unique pathophysiological state as prescribed by a truth table. The new concept is illustrated using an array of NAND and AND gates to assess the biomedical significance of numerous biomarker inputs including creatine kinase, lactate dehydrogenase, norepinephrine, glutamate, alanine transaminase, lactate, glucose, glutathione disulfide, and glutathione reductase to assess soft-tissue injury, traumatic brain injury, liver injury, abdominal trauma, hemorrhagic shock, and oxidative stress. Under the optimal conditions, physiological and pathological levels of these biomarkers were detected through either optical or electrochemical techniques by monitoring the level of the outputs generated by each of the six logic gates. By establishing a pathologically meaningful threshold for each logic gate, the absorbance and amperometric assays tendered the diagnosis in a digitally encoded 6-bit word, defined as an ‘injury code’. This binary ‘injury code’ enabled the effective discrimination of 64 unique pathological conditions to offer a comprehensive high-fidelity diagnosis of multiple injury conditions. Such processing of relevant biomarker inputs and the subsequent multiplexing of the logic gate outputs to yield a comprehensive ‘injury code’ offer significant potential for the rapid and reliable assessment of varied and complex forms of injury in circumstances where access to a clinical laboratory is not viable. While the new concept of parallel and multiplexed enzyme logic gates is illustrated here in connection to multi-injury diagnosis, it could be readily extended to a wide range of practical medical, industrial, security and environmental applications.

Introduction Conventional biomedical, industrial, security and environmental sensing techniques are limited in their capacity to process inputs in the chemical domain in a highly parallel fashion that enables the concurrent realization of high specificity/fidelity and reduced dependence on electronic signal processing. This fundamental limitation presents a challenge when utilizing such chemical sensors for a multitude of diverse applications. Although methods do exist to integrate and process a wide variety of sensor inputs, their operational merits are limited. A concept that seeks to differentiate among various scenarios must be able to address the multitude of possible situational phenomena, only some of which may be of concern. Accordingly, the development of high-fidelity biosensors necessitates the minimization of the electronic backbone and the co-location of the information processing operations within the chemical domain itself, even if multiple situations are assessed simultaneously.

a Department of Chemistry and Biomolecular Science, Clarkson University, Potsdam, NY, 13699-5810, USA. E-mail: [email protected]; Fax: +1 (315) 268 6610; Tel: +1 (315) 268 4421 b Department of NanoEngineering, University of California—San Diego, La Jolla, CA, 92093, USA. E-mail: [email protected]; Fax: +1 (858) 534 9553; Tel: +1 (858) 246 0128

This journal is ª The Royal Society of Chemistry 2010

The reliable and rapid detection of a multitude of common battlefield injuries is a paramount challenge in emergency medicine1 and represents an excellent example of the need to differentiate between various scenarios. Accordingly, there are urgent needs for a novel sensing contingent enabling a comprehensive high-fidelity diagnosis of multiple injury conditions.1 Battlefield injuries can vary greatly in nature, extent, and severity, and thus these injuries manifest a wide array of pathophysiological scenarios.2 This, in turn, presents a fundamental challenge to the identification of different forms of injury in occurrences such as polytrauma, especially in cases arising from blast injury,3 where a rapid determination of the injury/injuries is essential.4 A concept that seeks to differentiate various forms of relevant injuries must be able to mitigate the large number of combinations and permutations of physiological and pathological states in a rapid and reliable manner, only some of which may correspond to bona fide injury. In cases involving polytrauma, the diagnosis for such injury is typically tendered by a medical professional following close physical examination and a comprehensive regimen of conventional laboratory tests.5 Despite the advanced state of diagnostic technologies that assist the physician in assessing individual injuries, very few solutions exist in the hospital setting that can address multiple injuries in a comprehensive, rapid, and high-fidelity manner. Even fewer solutions exist that can yield such results in field settings.6 The Analyst, 2010, 135, 2249–2259 | 2249

realization of advanced, field-based diagnostic devices for reliable injury screening thus represents a crucial challenge in healthcare today. The simultaneous evaluation of multiple injuries is a required core competency of such diagnostic system. When mechanical damage to specific organs or tissues occurs, chemical species (proteins and/or low molecular-weight compounds, normally present only in intravascular compartments) are released into various body fluids (blood, urine, etc.). Fast and sensitive detection of these biomarkers is essential for a proper diagnosis of injury, many of which are already present at relatively high concentrations in blood under normal physiological conditions. Additionally, increased levels of certain freely circulating chemical species following an injury are associated with the metabolic response due to the pathological processes (oxidative stress, metabolic acidosis, etc.) and are not necessarily associated with damage to specific organs or tissues. Biochemical screening of injury/disease biomarkers nowadays relies on optical immunoassays7 performed by well-trained personnel using sophisticated semi-automated hospital analyzers.8 Obviously, such analyzers are not compatible with the requirements of rapid injury diagnostics in field settings. Several approaches have been proposed for the comprehensive decentralized assessment of multiple injuries in a convenient, single-test format.9 These include parallel biosensors/biosensor arrays10,11 and lab-on-a-chip devices,12 as these devices can leverage electronics for further information processing. Although these technologies enable the integration of a number of specific biochemical functionalities for the determination of a multitude of pathophysiological states, they suffer from several limitations.13 Most notably, each sensing element must be read-out and processed by dedicated electronic circuitry, thereby placing an additional burden on the corresponding microelectronic devices and accompanying power sources. Consequently, such schemes are not amenable to ultra-low power decentralized operation and integration into disposable biosensors that is an essential prerequisite of field-based biosensing devices. Moreover, to handle the relatively large amount of information acquired by these systems, the above techniques must rely on multiplexing or encoding operations to be performed in the electronic domain. The limited specificity of assays of single enzyme biomarkers can be rectified by the simultaneous analysis of several biomarkers. Although each biomarker may not allude to a specific injury when evaluated on an individual basis, the integration of multiple biomarkers would enable distinction among various injury conditions. This analysis could be performed with the implementation of Boolean logic in the biocatalytic systems,14 whereby the biochemical input and output signals are considered in the binary format: 0/1 (False/True). In this regard, a threshold separating the normal and pathological concentrations of the biomarkers can be implemented. Logical systems with multiple-input markers can be designed in such a manner that an injury can be detected only in instances when all relevant markers are present. Recent demonstrations of enzyme logic gates14 and their networks15 have shown substantial potential for determining conditions characteristic of hemorrhagic shock (HS) and traumatic brain injury (TBI).16 Despite the current state-of-the-art in this field,17 only single sensors/single gates have been proposed for injury diagnosis.18 An array-like concept employing multiple 2250 | Analyst, 2010, 135, 2249–2259

enzyme logic gates would enable high-fidelity diagnosis while benefiting from the parallelization and inclusion of further diagnostic capabilities that a biosensor array provides. A cornerstone merit of enzyme logic-based sensors lies in the reduced complexity of the electronic backbone required for the determination of a pathophysiological state through the utilization of biocatalytic information processing. Here we introduce a new biocomputing ‘injury coding’ diagnostic technique that can multiplex multiple injuries and assign each pathophysiological state a distinct ‘injury-code’, thereby enabling highly parallelized operation in the digital domain while minimizing the complexity of the analog electronic integration required for multiple-potentiostat electrochemical devices. Due to the Boolean nature of the enzyme logic concept,14 all normal physiological states upon implementation of the AND logic operation can be ascribed a logical ‘0’ value in the biochemical domain, which accounts for most of the injury combinations, prior to the transduction of the signals to the electrical domain. Only pathological conditions causing a change in the outputs relative to a pre-defined threshold level would result in a logical ‘1’ value, thereby alleviating the complex decision routines that must be performed in the electronic domain. It should be noted that the inverted logic values ‘1’ and ‘0’ will be applied for normal and pathological conditions, respectively, upon application of the NAND logic operation. As such, n outputs can be multiplexed into an n-bit word or ‘injury code’ for a comprehensive assessment of health conditions. A unique ‘injury code’ can thus be ascribed to a specific pathophysiological state in accordance with a truth table. A simple look-up table in digital logic circuitry could thus be employed to determine which injuries, if any, have been sustained in accordance with this distinct sequence of bits. In this manner, an array of n individual dual-input enzyme logic gates (each evaluating a separate injury) can assess 2n possible pathological conditions among 22n possible physiological states. The new concept represents the first demonstration of the parallelization of enzyme logic gates applied to diagnostic merits, as well as the simultaneous multiplexing of the outputs of multiple logic gates in the biochemical domain into a binary injury code. To illustrate the new concept and in accordance with the goal of rapid and reliable diagnosis of multiple injury states, an array of two NAND and four AND enzyme logic gates is assembled, as illustrated in Table 1. Six different pathological conditions are assessed including soft tissue injury (STI), traumatic brain injury (TBI), liver injury (LI), abdominal trauma (ABT), hemorrhagic shock (HS), and oxidative stress (OS) using twelve biomarker inputs. Optical absorbance and amperometric characterization of the six-gate system are conducted in order to verify compliance with a truth table as well as to ensure proper differentiation between the logical ‘0’ and ‘1’ output levels. The outputs of the six logic gates were subsequently multiplexed to yield a distinct 6bit injury code representing 64 unique pathological conditions among 4096 possible physiological scenarios. This leads to an additional (comprehensive) level of information on the overall nature of the injury, beyond the assessment of individual injuries performed by the individual gates. The system integration of clinically relevant enzyme logic gates and the subsequent multiplexing of their outputs into a cohesive injury code thus offer great promise for the rapid, reliable, and decentralized assessment of multi-injury and polytrauma conditions that typically This journal is ª The Royal Society of Chemistry 2010

Table 1 Enzyme cascades, equivalent logic gates, and truth tables corresponding to six unique injuries: STI, TBI, LI, ABT, HS, and OS

occur in the battlefield (compared to traditional biosensing schemes). It should be noted that the system is comprised of separate channels individually tailored for specific biomedical needs. The modularity of the system allows for straightforward reconfiguration of the constituents to enable the device to adapt, expand and meet new requirements and applications. The potential of the new modular biocomputing coding concept extends beyond the diagnosis of multiple injuries, as the concept could be readily extended for reliably assessing a wide range of other practical real-world scenarios involving multitude changes.

Experimental section Materials and methods Preparation of chemicals and reagents. Enzymes: glutathione reductase from Saccharomyces cerevisiae (GR, E.C. 1.8.1.7), glucose oxidase type X-S from Aspergillus niger (GOx, E.C. 1.1.3.4), alanine transaminase from porcine heart (ALT, E.C. 2.6.1.2), pyruvate kinase from rabbit muscle (PK, E.C. 2.7.1.40), This journal is ª The Royal Society of Chemistry 2010

creatine kinase from rabbit muscle (CK, E.C. 2.7.3.2), and lactate dehydrogenase from porcine heart (LDH, E.C. 1.1.1.27) were purchased from Sigma-Aldrich and were used as supplied without any pretreatment or purification. L-Glutamate oxidase from Streptomyces sp. (GlOx, E.C. 1.4.3.11) was obtained from Yamasa Corporation, Japan, and used as supplied. All other chemicals (Sigma-Aldrich) were used without purification: bovine serum albumin (BSA), b-D-(+)-glucose (GLC), ()-norepinephrine (NE), microperoxidase (MP-11), b-nicotinamide adenine dinucleotide dipotassium salt (NAD+), b-nicotinamide adenine dinucleotide reduced dipotassium salt (NADH), b-nicotinamide adenine dinucleotide 20 -phosphate reduced tetrasodium salt (NADPH), L-glutamic acid (GLU), L-alanine (L-ALA), a-ketoglutaric acid (KTG), L(+)-lactic acid (LAC), L-glutathione disulfide (GSSG), creatine anhydrous (CRTN), phospho(enol)pyruvate monopotassium salt (PEP), adenosine 50 -triphosphate disodium (ATP, from bacterial source), glycyl-glycine (Gly-Gly), cobalt(II)-phthalocyanine (CoPC), magnesium acetate tetrahydrate (MgAc), sodium hydroxide (NaOH), magnesium chloride (MgCl2), calcium chloride (CaCl2), potassium hydroxide (KOH), pyruvic acid (PYR), Analyst, 2010, 135, 2249–2259 | 2251

methylene green (MG), tris(hydroxymethyl)aminomethane hydrochloride salt (Tris-buffer), potassium phosphate monobasic (PPM), potassium phosphate dibasic (PPD), citric acid anhydrous (CA), sodium citrate dihydrate (SC), and 5,50 -dithiobis-(2-nitrobenzoic acid) (DTNB—Ellman’s reagent). Ultrapure water (18.2 MU cm) from NANOpure Diamond (Barnstead) source was used in all of the experiments. Instrumentation and measurements. A Shimadzu UV-2450 UVVis spectrophotometer (with a TCC-240A temperature-controlled cuvette holder and 1 mL PMMA or 0.5 mL quartz cuvettes) was used for all optical measurements and a CH Instruments model 1232A potentiostat was used for all electrochemical measurements. A Mettler Toledo SevenEasy s20 pH-meter was employed for the pH measurements. A VWR Analog Heatblock was utilized as a temperature-controlled incubator. All optical measurements were performed in temperature-controlled cuvettes/cells at 37  0.2  C mimicking physiological conditions and all reagents were incubated at this temperature prior to experimentation. Measurements for the each combination of input signals were repeated at least 3 times. Electrode design and fabrication. A screen printed three-electrode strip, custom-designed using AutoCAD, consisted of a circular carbon working electrode (geometrical area: 3 mm2) inscribed in hemispherical counter (area: 10 mm2) and reference electrodes (area: 2 mm2). The fabrication of the flexible screenprinted electrode system is detailed: an Ag/AgCl-based ink from Ercon (E2414) was employed to define the conductive underlayer as well as the reference electrode. All potentials are reported vs. this reference electrode. A carbon-based ink (Ercon E3449) was then overlaid on the conductor to define the working and counter electrode geometry. Finally, an insulator ink (Ercon E6165) was overlaid on the Ag/AgCl and carbon layers to insulate all except the contact pads and the upper active segment of the electrodes. A Speedline Technologies MPM-SPM screen printer was used to print the pattern onto a 250 mm thick flexible polyethylene terephthalate substrate (DuPont Melinex 329). Subsequent to the printing process, the patterned substrate was cured in a temperature-controlled convection oven (SalvisLab Thermocenter) at 120  C for 20 min. The substrate was finally cleaved to create single-use test strips possessing overall dimensions of 10 mm  34 mm. For the OS experiments, CoPC microparticles were dispersed in the carbon ink (2% w/w) due to the insolubility of the electrochemical mediator in the solution phase. For the LI experiments, a glassy carbon (GC) disc electrode was used as a working electrode. Composition and operation of channels for the analysis of injuries Among pervasive battlefield injuries, soft tissue injury (STI), traumatic brain injury (TBI), acute liver injury (LI), abdominal trauma (ABT), hemorrhagic shock (HS), and oxidative stress (OS) are the most common sustained by soldiers in combat.19–22 1. Soft tissue injury (STI). Gly-Gly buffer, 50 mM, with 6.7 mM MgAc was titrated by KOH to pH 7.95 and used as a background solution (note that Mg2+ and K+ cations are essential for activation of CK and PK, respectively). The 2252 | Analyst, 2010, 135, 2249–2259

following components were dissolved in this solution to perform the NAND logic operation: NADH (0.3 mM), BSA (0.03% w/v), ATP (2 mM), PEP (0.5 mM), PK (2 U mL1), creatine (15 mM). Logical ‘0’ and ‘1’ levels of CK (0.1 and 0.71 U mL1) and LDH (0.15 and 1 U mL1) input signals were applied to the logic system in order to realize meaningful circulating levels of these biomarkers. Immediately following mixture, optical absorbance measurements were recorded continuously for 300 s at l ¼ 340 nm. The solutions for the electrochemical measurements were incubated for 180 s at 37  C, then an MG redox mediator (0.3 mM) was added to catalyze electrochemical oxidation of NADH on the SPE, and a chronoamperogram was recorded at E ¼ 0.0 V. 2. Traumatic brain injury (TBI). Potassium phosphate buffer, 50 mM, pH 7.4, containing GlOx (1 mU mL1) and MP-11 (0.44 mM for optical and 5 mM for electrochemical measurements) was used to perform the AND logic operation. Logical ‘0’ and ‘1’ levels of NE (2.2 nM and 3.5 mM) and GLU (40 mM and 140 mM) were applied to the logic system in order to realize meaningful circulating levels of these biomarkers. The norepiquinone (NQ) produced in situ was optically measured at l ¼ 487 nm. Chronoamperometric detection of NQ was performed on the SPE at E ¼ 0.4 V. 3. Liver injury (LI). Tris–HCl buffer, 100 mM, pH 7.4, was used as a background solution. L-ALA (200 mM), KTG (10 mM) and NADH (136 mM) were dissolved in this solution to perform the NAND logic operation. Logical ‘0’ and ‘1’ levels of ALT (0.02 and 0.2 U mL1) and LDH (0.15 and 1 U mL1) input signals were applied to the logic system in order to realize meaningful circulating levels of these biomarkers. The output signal corresponding to the decreasing concentration of NADH was measured optically at l ¼ 340 nm. The solutions for the electrochemical measurements were incubated for 180 s at 37  C, chronoamperogram was recorded at bare glassy carbon (GC) electrode at an applied potential of E ¼ 0.8 V. 4. Abdominal trauma (ABT). Potassium phosphate buffer, 50 mM (pH 7.15 for optical or pH 7.40 for electrochemical measurements), containing 0.2 mM MgCl2, 0.01 mM CaCl2, and NAD+ (10 mM for optical or 1 mM for electrochemical measurements), was used to perform the AND logic operation. Logical ‘0’ and ‘1’ levels of LDH (0.15 and 1.0 U mL1) and lactate (1.6 and 6.0 mM) input signals were applied to the logic system in order to realize meaningful circulating levels of these biomarkers. The output signal corresponding to the NADH formation was measured optically at l ¼ 340 nm. The mixture for electrochemical measurements was incubated for 180 s at 37  C, then the MG redox mediator (0.3 mM) was added to catalyze electrochemical oxidation of NADH and a chronoamperogram was recorded at E ¼ 0.1 V. 5. Hemorrhagic shock (HS). Potassium phosphate buffer (50 mM, pH 7.4 for optical or 100 mM, pH 7.0 for electrochemical measurements) containing GOx (2 mU mL1) and MP11 (0.44 mM for optical and 5 mM for the electrochemical measurements) was used to perform the AND logic operation. Logical ‘0’ and ‘1’ levels of NE (2.2 nM and 3.5 mM) and GLC (4 mM and 26 mM) were applied to the logic system in order to This journal is ª The Royal Society of Chemistry 2010

realize meaningful circulating levels of these biomarkers. The norepiquinone (NQ) produced in situ was measured optically at l ¼ 487 nm. Chronoamperometric detection of NQ was performed on the SPE at E ¼ 0.4 V. 6. Oxidative stress (OS). Citrate buffer, 50 mM, pH 5.0, was used as a background solution. DTNB (2 mM) and NADPH (180 mM) were dissolved in this solution to perform the AND logic operation. Logical ‘0’ and ‘1’ levels of GSSG (150 mM and 400 mM) and GR (0.55 U mL1 and 0.65 U mL1) input signals were applied to the logic system in order to realize meaningful circulating levels of these biomarkers. Optical analysis of the reduced glutathione (GSH) produced in situ was performed according to the standard procedure using Ellman’s reagent (DTNB)23 at l ¼ 412 nm for 60 s. Electrochemical analysis of GSH was performed by chronoamperometric measurements at E ¼ 0.5 V on the CoPC-modified SPE in a similar system without DTNB. The concentrations of the biomarker input used for the activation of the logic gates are summarized in Table 2.

Results and discussion The new concept of parallel and multiplexed enzyme logic gates is illustrated here in connection to multi-injury diagnosis. Such biocomputing-based comprehensive diagnosis of multiple injury conditions harnesses the processing capabilities of six individual enzyme logic gates, realizing NAND and AND operations, for the detection of STI, TBI, LI, ABT, HS, and OS using clinically relevant combinations of the biomarkers,24–29 CK, LDH, NE, GLU, ALT, LAC, GLC, GSSG, and GR. The output of each logic gate was subsequently integrated as one of the constituents of the injury code-generating system to enable high-fidelity assessment of multiple injury conditions and hence a comprehensive analysis of the scope of injury. Upon the optimization of the design parameters, the four unique combinations of the two biomarker inputs to each individual logic gate resulted in distinguishable patterns of its respective output. For two of the gates, and in accordance with the NAND gate operational functionality, logical ‘0’ and ‘1’ levels of the input biomarkers, corresponding to normal or anomalous physiological conditions, respectively ((Input 1, Input 2) ¼ (0,0), (0,1), and (1,0)), resulted in an output of logical

‘1’. On the other hand, logical ‘1’ levels of both of the biomarkers (Input 1, Input 2) ¼ (1,1) caused the output state to change from ‘1’ to ‘0’, indicating that an injury has occurred. It should be noted that the logic output signal ‘0’ generated by the NAND gate and corresponding to a positive diagnosis does not imply that the signal is truly at a zero level. Rather, a ‘0’ output implies that the system has transitioned from a state producing a signal of high magnitude to one that yields a low level signal. In this regard, the logic output signal ‘1’ indicates that the output signal is unchanged. The remaining four AND gates functioned under a logical inversion of the above operation. In this manner, logical ‘0’ and ‘1’ levels of the input biomarkers, corresponding to normal or anomalous physiological conditions, respectively ((Input 1, Input 2) ¼ (0,0), (0,1), and (1,0)), resulted in an output of logical ‘0’ whereas logical ‘1’ levels of both of the biomarkers (Input 1, Input 2) ¼ (1,1) caused the output state to change from ‘0’ to ‘1’, thereby signifying that an injury has been sustained. Overall, the digital outputs of the six logic gates can be subsequently multiplexed to yield a comprehensive 6-bit injury code (e.g., 0, 0, 1, 1, 0, 1), corresponding to 64 unique pathological conditions (i.e., various combinations of six unique injuries). With respect to each logic gate, the concentrations of the reagents that served as machinery (and not as inputs) were individually tailored to yield optimal dynamic range between the pathological level (1,1) and normal or anomalous physiological levels not related to injuries (0,0), (0,1), and (1,0). This enabled unambiguous determination of the injury state (when the output signal ‘0’ is generated by the NAND gate and ‘1’ by the AND gate) due to the establishment of a fixed decision threshold. With respect to both gate archetypes, only the simultaneous presence of elevated levels of both inputs would trigger a positive diagnosis. On the other hand, the output signal ‘1’ in the NAND topology and ‘0’ in the AND embodiment imply extraneous pathophysiological states ranging from healthy conditions to various physiological anomalies not related to the injuries under investigation. It should be noted that application of logic ‘0’ and ‘1’ input values corresponding to the physiologically relevant concentrations of the injury biomarkers required substantial optimization of the enzyme-based logic gates. For some of the injury scenarios under study, the concentration differential between logic ‘0’ and ‘1’ values was quite narrow implying strict limitations to the threshold values. Thus, the design of the logic gate required

Table 2 Physiological and pathological levels of clinically relevant biomarkers for each logic gate with the output compound indicated No.

Injury

Biomarkers

Physiological

Pathological

Output

1

Soft tissue injury (STI) Traumatic brain injury (TBI)

3

Liver injury (LI)

4

Abdominal trauma (ABT)

5

Hemorrhagic shock (HS)

6

Oxidative stress (OS)

100 U L1 150 U L1 2.2 nM 40 mM 20 U L1 150 U L1 1.6 mM 150 U L1 2.2 nM 4 mM 150 mM 556 U L1

710 U L1 1000 U L1 3.5 mM 140 mM 200 U L1 1000 U L1 6 mM 1000 U L1 3.5 mM 26 mM 400 mM 650 U L1

NADH decrease

2

Creatine kinase (CK) Lactate dehydrogenase (LDH) Norepinephrine (NE) Glutamate (GLU) Alanine transaminase (ALT) Lactate dehydrogenase (LDH) Lactate (LAC) Lactate dehydrogenase (LDH) Norepinephrine (NE) Glucose (GLC) Glutathione disulfide (GSSG) Glutathione reductase (GR)

This journal is ª The Royal Society of Chemistry 2010

Norepiquinone increase NADH decrease NADH increase Norepiquinone increase GSH increase

Analyst, 2010, 135, 2249–2259 | 2253

a more careful consideration and optimization of the assay parameters compared with logic gates employed in previous studies,14,15 whereby logic ‘0’ and ‘1’ values were represented by truly zero and arbitrarily high concentrations of the input signals, respectively.

Soft tissue injury (STI) Soft tissue injuries are among the most pervasive injuries sustained in combat and can be difficult to identify in numerous circumstances.30 Among clinically established indicators of STI, serum CK and LDH have been routinely employed in the assessment of muscular exertion, fatigue, injury, and trauma.24 Commencing experiments using optical assay methods, Fig. 1A (left) displays the optical absorbance of the NAND gate at l ¼ 340 nm. A large differentiation between pathological and physiological logic levels is observed, reflecting the rapid enzymatic reaction. From the corresponding bar chart (Fig. 2A (left)) constructed using the data at 300 s, the (CK, LDH) ¼ (1,1) logic level was separated by more than 0.52 O.D. from the nearest logic level. An explicit decision threshold could hence be established at 0.49 O.D., leading to highly reliable NAND operation. In accordance with the goal of low-cost decentralized screening of STI and developing compact analytical devices, the aforementioned protocol was subsequently migrated to the amperometric domain using a disposable SPE. Towards this goal, chronoamperometric measurements were performed for each combination of input signals with MG added to the assay, allowing for the low-potential detection of NADH, hence minimizing potential electroactive interferences. The detection potential was established at 0.0 V to maximize the signal-to-noise ratio (SNR) figure of merit. Chronoamperograms are shown in Fig. 1A (right) which were obtained at the carbon SPE by the NAND gate upon application of various input combinations. At 60 s sampling time, the difference in current between the (CK, LDH) ¼ (1,1) logic and (1,0) logic levels was 27 nA, as shown by the bar chart, Fig. 2A (right). As in the optical experiments, the bar chart indicates that a straightforward decision threshold could be instituted to realize high-fidelity NAND gate operation. This threshold was fixed at 135 nA.

Traumatic brain injury (TBI)

Fig. 1 Optical (left) and electrochemical (right) response generated by the (A) STI NAND, (B) TBI AND, (C) LI NAND, (D) ABT AND, (E) HS AND, and (F) OS AND logic gates upon various combinations of the input biomarkers (0,0) (black), (0,1) (blue), (1,0) (green), and (1,1) (red). Optical absorbance measurements were performed at l ¼ 340 nm, 487 nm, 340 nm, 340 nm, 487 nm, and 412 nm for the STI, TBI, LI, ABT, HS, and OS gates, respectively. Electrochemical chronoamperograms were performed at E ¼ 0.0 V, 0.4 V, 0.8 V, 0.1 V, 0.4 V, and 0.5 V (vs. Ag/ AgCl) for the STI, TBI, LI, ABT, HS, and OS gates, respectively.

2254 | Analyst, 2010, 135, 2249–2259

Traumatic brain injuries are among the most debilitating injuries suffered in the battlefield31 and have garnered much recent attention due to the wide range of symptoms and characteristics presented by individuals suffering from the condition.32 NE, a catecholamine hormone neurotransmitter, and GLU, an aminoacid excitatory neurotransmitter, are among the most widely employed indicators of neurological damage and trauma.26 Commencing experiments using optical assay methods, Fig. 1B (left) displays the optical absorbance of the AND gate at l ¼ 487 nm. As with the STI gate, a large differentiation between pathological and physiological logic levels is observed, reflecting favorable enzyme kinetics. From the corresponding bar chart (Fig. 2B (left)) constructed using the data at 200 s, the (NE, GLU) ¼ (1,1) logic level was separated by 6.65 mO.D. from the nearest logic level. An explicit decision threshold could hence be This journal is ª The Royal Society of Chemistry 2010

established at 10.96 mO.D., leading to highly reliable AND operation. The aforementioned protocol was subsequently migrated to the amperometric domain using a disposable SPE. Accordingly, chronoamperometric measurements were performed for each combination of the input signals with the concentration of norepiquinone (NQ) serving as the output indicator. The detection potential was established at 0.4 V for SNR considerations. Chronoamperograms are shown in Fig. 1B (right) which were obtained at the carbon SPE by the AND gate upon application of various input combinations. At 60 s sampling time, the difference in current between the (1,1) logic and (0,1) logic levels was 13 nA, as shown by the bar chart in Fig. 2B (right). As in the optical experiments, the bar chart indicates that a straightforward decision threshold could be instituted to realize high-fidelity AND gate operation, which was established at 30 nA. Liver injury (LI) Screening of liver injury has routinely been employed in the clinical laboratory through enzyme-based assay tests33 to assess sepsis—a secondary, but life-threatening condition arising due to such injuries.34 This injury is especially prevalent in combat situations where damaged organs cause foreign matter to enter and circulate in the bloodstream.35 Serum ALT and LDH have enjoyed widespread use as enzyme biomarkers in such assays27 and are well-suited for integration as input biomarkers and enzyme backbones to the enzyme logic machinery. Commencing experiments using optical assay methods, Fig. 1C (left) displays the optical absorbance of the NAND gate at l ¼ 340 nm. A large differentiation between pathological and physiological logic levels is observed, again a result of enhanced enzyme activity. From the corresponding bar chart (Fig. 2C (left)) constructed using the data at 150 s, the (ALT, LDH) ¼ (1,1) logic level was separated by 0.52 O.D. from the nearest logic level. An explicit decision threshold could hence be established at 0.52 O.D., yielding the expected NAND operation. Electrochemical experiments were subsequently executed by employing a bare GC electrode. Accordingly, chronoamperometric measurements were performed for each combination of the input signals and the level of NADH consumed was monitored at 0.8 V. Chronoamperograms are shown in Fig. 1C (right) obtained at the GC electrode by the NAND gate upon application of various input combinations. At 60 s sampling time, the difference in current between the (1,1) and (1,0) logic levels was 1.3 mA, as shown by the bar chart in Fig. 2C (right). As in the optical experiments, the bar chart validates that a decision threshold could be unambiguously implemented to realize highfidelity NAND gate operation, which was established at 6.0 mA. Fig. 2 Optical (left) and electrochemical (right) bar charts obtained by sampling the output of the (A) STI NAND, (B) TBI AND, (C) LI NAND, (D) ABT AND, (E) HS AND, and (F) OS AND logic gates upon various combinations of the input biomarkers (0,0), (0,1), (1,0), and (1,1). Optical absorbance measurements were extracted at t ¼ 300 s, 200 s, 150 s, 200 s, 100 s, and 60 s for the STI, TBI, LI, ABT, HS, and OS gates, respectively. Electrochemical chronoamperograms were sampled at t ¼ 60 s, 60 s, 60 s, 10 s, 60 s, and 30 s for the STI, TBI, LI, ABT, HS, and OS gates, respectively. Dashed lines indicate the decision threshold for the realization of Boolean logic gate operation.

This journal is ª The Royal Society of Chemistry 2010

Abdominal trauma (ABT) As with liver injury, severe abdominal trauma frequently results in sepsis and must be addressed with little or no delay in order to improve survival.36 In addition, ABT is another example of a common battlefield injury which has been linked to high mortality rates.37 Serum LAC and LDH are well-established biomarkers of such injury24 in addition to serving extensive use in assays.38 Analyst, 2010, 135, 2249–2259 | 2255

Initially, optical experiments were conducted and Fig. 1D (left) displays the optical absorbance of the AND gate at l ¼ 340 nm at each combination of the input signals. Pathological and physiological logic levels were easily differentiated, which confirmed that the assay conditions were favorable for this application. From the corresponding bar chart (Fig. 2D (left)) constructed using the data at 200 s, the (LAC, LDH) ¼ (1,1) logic level was separated by 0.13 O.D. from the logic level in closest proximity. Accordingly, an unambiguous decision threshold could be established at 0.81 O.D. in order to yield AND functionality. Following optical experiments, electrochemical investigations were then performed by employing a disposable SPE. Chronoamperometric measurements were performed for each combination of the input signals with MG serving as the output mediator to reduce the overpotential required for the detection of NADH. The detection potential was established at 0.1 V to maximize SNR. Chronoamperograms are shown in Fig. 1D (right) which were obtained at the carbon SPE by the AND gate upon application of various input combinations. At 10 s sampling time, the difference in current between the (1,1) logic and (1,0) logic levels was 36 nA, as shown by the bar chart in Fig. 2D (right). Upon examination of the bar chart, it is apparent that a decision threshold could be unambiguously implemented to realize high-fidelity AND gate operation (as in the optical case), which was established at 84 nA.

Hemorrhagic shock (HS) Hemorrhagic shock, a condition that arises due to uncontrolled bleeding, is another pervasive example of a high-mortality combat injury and frequently occurs as a result of the infliction of gun-shot wounds and blast injuries.39 This critical condition must be assessed before the individual who has sustained the injury bleeds to death.40 Serum GLC and NE have been identified as biomarkers of such injury25,28 and can increase many-fold upon presentation of this condition.41,42 Optical experiments were conducted and Fig. 1E (left) displays the optical absorbance at l ¼ 487 nm of the AND gate for each combination of the input signals. Pathological and physiological logic levels were well-separated, which again confirmed favorable assay conditions. From the corresponding bar chart (Fig. 2E (left)) constructed using the data at 100 s, the (NE, GLC) ¼ (1,1) logic level was separated by 8.89 mO.D. from the logic level in closest proximity. Accordingly, a straightforward decision threshold could be established at 11.0 mO.D. in order to yield AND functionality. Following optical experiments, electrochemical investigations were then performed by employing a disposable SPE. Chronoamperometric measurements were performed for each combination of the input signals with the concentration of norepiquinone (NQ) serving as the output indicator. The detection potential was established at 0.4 V for SNR considerations. Chronoamperograms are shown in Fig. 1E (right) which were obtained at the carbon SPE by the AND gate upon application of various input combinations. At 60 s sampling time, the difference in current between the (1,1) logic and (1,0) logic levels was 11 nA, as shown by the bar chart in Fig. 2E (right). In accordance with these results, the decision threshold to realize high-fidelity AND gate operation (as in the optical case) was established at 16 nA. 2256 | Analyst, 2010, 135, 2249–2259

Oxidative stress (OS) Oxidative stress refers to a broad scope of pathological states and accompanies nearly all forms of physical stress or strain experienced by the body43 including those acquired in battle.44 GSSG and GR mitigate the body’s biochemical response to oxidative stress and elevations in each compound in serum have been found to be associated with such events. Thus, GSSG and GR are excellent candidates for biomarkers29 that can enable highly reliable logic gate operation. Optical experiments were first conducted and Fig. 1F (left) displays the optical absorbance of the AND gate at l ¼ 412 nm for each combination of the input signals using DTNB for optical analysis of the GSH produced in situ.23 Pathological and physiological logic levels were easily differentiable, thereby validating high-fidelity operation. From the corresponding bar chart (Fig. 2F (left)) constructed using the data at 60 s, the (GSSG, GR) ¼ (1,1) logic level was separated by 0.07 O.D. from the logic level in closest proximity. A decision threshold could hence be set at 0.76 O.D. in order to yield AND functionality. Electrochemical investigations were then performed by employing a disposable SPE. Chronoamperometric measurements were performed for each combination of the input signals with CoPC serving as the output mediator to reduce the overpotential required for the detection of GSH. The detection potential was established at 0.5 V to maximize SNR. Chronoamperograms are shown in Fig. 1F (right) which were obtained at the CoPCmodified carbon SPE by the AND gate upon application of various input combinations. At 30 s sampling time, the difference in current between the (1,1) logic and (1,0) logic levels was 51 nA, as shown by the bar chart in Fig. 2F (right). As this figure illustrates, a decision threshold could be implemented to realize highfidelity AND gate operation (as in the optical case), and was accordingly established at 412 nA.

Systems integration Accordingly, good correlation was observed between the optical and electrochemical data in each of the six separate experiments, as indicated from a comparison of the bar charts presented in Fig. 2, thereby confirming the validity of the transition of the experimental procedure from the optical to the electrochemical domain. With the electrochemical protocol in functional order, the output of each logic gate could be integrated as one of the constituents of the injury code-generating system. With the above six gates optimized to enable high-fidelity detection, a 6-bit injury code was concatenated from the outputs of each of the individual gates. Accordingly, 26 ¼ 64 unique injury combinations could be ascertained, as shown in Table 3, among 212 ¼ 4096 possible pathophysiological scenarios. A comprehensive injury code could thus be constructed to account for various combinations of six unique injuries. Such an encoding scheme enables a reduced dependence on analog circuitry required for the synthesis of high-integrity waveforms required for digital signal processing, as the outputs are generated with substantial differentiation between the digitally defined physiological and pathological levels in the biochemical domain. Moreover, the binary nature of the output indicators of injury enabled multiplexing operations to be This journal is ª The Royal Society of Chemistry 2010

Table 3 Truth table and corresponding injury codes for all 64 possible combinations of STI, TBI, LI, ABT, HS, and OS No.

STI

TBI

LI

ABT

HS

OS

Injury code

Biomedical conclusions

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64

0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1

0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1

0 0 0 0 0 0 0 0 1 1 1 1 1 1 1 1 0 0 0 0 0 0 0 0 1 1 1 1 1 1 1 1 0 0 0 0 0 0 0 0 1 1 1 1 1 1 1 1 0 0 0 0 0 0 0 0 1 1 1 1 1 1 1 1

0 0 0 0 1 1 1 1 0 0 0 0 1 1 1 1 0 0 0 0 1 1 1 1 0 0 0 0 1 1 1 1 0 0 0 0 1 1 1 1 0 0 0 0 1 1 1 1 0 0 0 0 1 1 1 1 0 0 0 0 1 1 1 1

0 0 1 1 0 0 1 1 0 0 1 1 0 0 1 1 0 0 1 1 0 0 1 1 0 0 1 1 0 0 1 1 0 0 1 1 0 0 1 1 0 0 1 1 0 0 1 1 0 0 1 1 0 0 1 1 0 0 1 1 0 0 1 1

0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1

000000 000001 000010 000011 000100 000101 000110 000111 001000 001001 001010 001011 001100 001101 001110 001111 010000 010001 010010 010011 010100 010101 010110 010111 011000 011001 011010 011011 011100 011101 011110 011111 100000 100001 100010 100011 100100 100101 100110 100111 101000 101001 101010 101011 101100 101101 101110 101111 110000 110001 110010 110011 110100 110101 110110 110111 111000 111001 111010 111011 111100 111101 111110 111111

STI + LI STI + LI + OS STI + LI + HS STI + LI + HS + OS STI + LI + ABT STI + LI + ABT + OS STI + LI + ABT + HS STI + LI + ABT + HS + OS STI STI + OS STI + HS STI + HS + OS STI + ABT STI + ABT + OS STI + ABT + HS STI + ABT + HS + OS STI + TBI + LI STI + TBI + LI + OS STI + TBI + LI + HS STI + TBI + + LI + HS + OS STI + TBI + LI + ABT STI + TBI + LI + ABT + OS STI + TBI + LI + ABT + HS STI + TBI + LI + ABT + HS + OS STI + TBI STI + TBI + OS STI + TBI + HS STI + TBI + HS + OS STI + TBI + ABT STI + TBI + ABT + OS STI + TBI + ABT + HS STI + TBI + ABT + HS + OS LI LI + OS LI + HS LI + HS + OS LI+ ABT LI + ABT + OS LI + ABT + HS LI + ABT + HS + OS Normal OS HS HS + OS ABT ABT + OS ABT + HS ABT + HS + OS TBI + LI TBI + LI + OS TBI + LI + HS TBI + LI + HS + OS TBI + LI + ABT TBI + LI + ABT + OS TBI + LI + ABT + HS TBI + LI + ABT + HS + OS TBI TBI + OS TBI + HS TBI + HS + OS TBI + ABT TBI + ABT + OS TBI + ABT + HS TBI + ABT + HS + OS

performed directly in the digital domain and could be appraised against a lookup table for reliable evaluation of pathophysiological state. The sizeable dynamic range and high noise margins This journal is ª The Royal Society of Chemistry 2010

of the NAND and AND logic gates facilitate such high-fidelity operation and serve to further underscore the advantages of the concept when contrasted with traditional biosensor and Analyst, 2010, 135, 2249–2259 | 2257

lab-on-a-chip approaches. Without the use of Boolean processing and the establishment of a concomitant decision threshold for digital diagnosis, a reliable assessment of injury could not be tendered nor could an injury code be constructed. In the absence of these merits, the approach would encounter serious obstacles in discarding the noise arising from extraneous interferents and anomalous pathophysiological conditions and extracting the signal of interest. Therefore, the operational merits of the logic gate architecture enabled the establishment of an unambiguous decision threshold and digital manipulation of the biomarker signals, thereby enabling the concatenation of an inclusive injury code to assess multiple injury conditions, which allowed the system an enhanced ability to detect injury when compared with traditional biosensing concepts. Noise reduction, fault tolerance, robustness and scalability are all factors to be considered when attempting to improve the performance of enzyme-based information processing systems. Experimental optimization was implemented in order to produce a reliable threshold between output ¼ 0 and output ¼ 1. Although noise can be suppressed by logic network design, other intrinsic experimental parameters will dictate the precision of each logic level. When assessing the optical and electrochemical experimental results, the standard deviation for most of the values does not exceed 5% of the mean value, demonstrating reproducibility within the chosen physiological and pathological input ranges. The chosen threshold lines separating logic output ¼ 0 and logic output ¼ 1 varied between 11% and 60% of the mean value; hence we can consider that the logic gates serve as reliable injury predictors for at least 90% of the repetitions.

Conclusions We have introduced a new modular biocomputing coding concept based on parallel and multiplexed enzyme logic gates. This enzyme logic coding approach has been demonstrated towards a multi-injury diagnosis in connection to twelve biomarker inputs. This system is able to assess 64 individual pathological conditions among 4096 possible pathophysiological states through multiplexing the outputs into a 6-bit ‘injury code’. The new system, consisting of AND and NAND gates, is able to evaluate a greater number of injuries than permitted by single enzyme logic gates while leveraging an electronic backbone of similar complexity. Moreover, due to the concept’s Boolean biochemical signal processing architecture, the system is able to infer pathological conditions with a greater degree of reliability while enabling lower power consumption operation due to the reduced dependence on electronic operations than conventional biosensor arrays are currently able to offer. The concept represents the first demonstration of the parallelization of enzyme logic gates applied to diagnostic merits as well as the simultaneous multiplexing of the outputs of multiple logic gates in the biochemical domain into a binary injury code. Such devices hold considerable potential for future work in the advancement of low-cost, disposable biosensors. Further development is required to realize an on-body biosensing paradigm, albeit the injury code concept is well-positioned to enable the rapid and reliable assessment of multiple life-threatening injuries away from the hospital setting. Moreover, in addition to its connection to multiinjury diagnosis, the concept could be extended to the reliable 2258 | Analyst, 2010, 135, 2249–2259

assessment of a wide range of other practical medical, industrial, security and environmental scenarios.

Acknowledgements This work was supported by the Office of Naval Research (Award #N00014-08-1-1202).

References 1 G. S. F. Ling, K. Day, P. Rhee and J. M. Ecklund, Proc. SPIE-Int. Soc. Opt. Eng., 1999, 3712, 1–8. 2 R. J. Winterborn and T. A. Cook, Surgery (Oxford), 2003, 21, 240a– 240e. 3 S. G. Scott, H. G. Belanger, R. D. Vanderploeg, J. Massengale and J. Scholten, J. Am. Osteopath. Assoc., 2006, 106, 265–270. 4 M. Keel and O. Trentz, Injury, 2005, 36, 691–709. 5 P. V. Giannoudis, J. Bone Jt. Surg., Br. Vol., 2003, 85-B, 478–483. 6 I. Moser, G. Jobst and G. A. Urban, Biosens. Bioelectron., 2002, 17, 297–302. 7 A. Ettinger, A. B. Laumark, R. M. Ostroff, J. Brundell, W. A. Baumgartner and A. Y. Razumovsky, Ann. Thorac. Surg., 1999, 68, 2196–2201. 8 G. Bourner, J. Dhaliwal and J. Sumner, Lab. Hematol., 2005, 11, 285– 297. 9 J. Wang and Q. Chen, Anal. Chem., 1994, 66, 1007–1011. 10 F. S. Ligler, K. E. Sapsford, J. P. Golden, L. C. Shriver-Lake, C. R. Taitt, M. A. Dyer, S. Barone and C. J. Myatt, Anal. Sci., 2007, 23, 5–10. 11 R. Bashir, Adv. Drug Delivery Rev., 2004, 56, 1565–1586. 12 C. H. Ahn, J. W. Choi, G. Beaucage, J. H. Nevin, J. B. Lee, A. Puntambekar and J. Y. Lee, Proc. IEEE, 2004, 92, 154–173. 13 D. R. Thevenot, K. Toth, R. A. Durst and G. S. Wilson, Anal. Lett., 2001, 34, 635–659. 14 G. Strack, M. Pita, M. Ornatska and E. Katz, ChemBioChem, 2008, 9, 1260–1266; J. Zhou, M. A. Arugula, J. Halamek, M. Pita and E. Katz, J. Phys. Chem. B, 2009, 113, 16065–16070; V. Privman, V. Pedrosa, D. Melnikov, M. Pita, A. Simonian and E. Katz, Biosens. Bioelectron., 2009, 25, 695–701; D. Melnikov, G. Strack, M. Pita, V. Privman and E. Katz, J. Phys. Chem. B, 2009, 113, 10472–10479; V. Privman, G. Strack, D. Solenov, M. Pita and E. Katz, J. Phys. Chem. B, 2008, 112, 11777–11784. 15 M. A. Arugula, J. Halamek, E. Katz, D. Melnikov, M. Pita, V. Privman and G. Strack, IEEE Computer Society Publications (Los Alamitos, California), 2009, 1–7; T. K. Tam, M. Pita and E. Katz, Sens. Actuators, B, 2009, 140, 1–4; V. Privman, M. A. Arugula, J. Halamek, M. Pita and E. Katz, J. Phys. Chem. B, 2009, 113, 5301–5310; M. Privman, T. K. Tam, M. Pita and E. Katz, J. Am. Chem. Soc., 2009, 131, 1314–1321. 16 K. M. Manesh, J. Halamek, M. Pita, J. Zhou, T. K. Tam, P. Santhosh, M.-C. Chuang, J. R. Windmiller, D. Abidin, E. Katz and J. Wang, Biosens. Bioelectron., 2009, 24, 3569–3574; M. Pita, J. Zhou, K. M. Manesh, J. Halamek, E. Katz and J. Wang, Sens. Actuators, B, 2009, 139, 631–636. 17 E. Katz and V. Privman, Chem. Soc. Rev., 2010, 39, 1835–1857; E. Katz, V. Privman and J. Wang, Proceedings of the 4th International Conference on Quantum, Nano and Micro Technologies (ICQNM 2010), St. Maarten, Netherlands, 2010, pp. 1–9. 18 J. Wang and E. Katz, Anal. Bioanal. Chem., 2010, DOI: 10.1007/ s00216-010-3746-0. 19 D. Michaeli, J. R. Soc. Med., 1979, 72, 370–373. 20 M. Wilson, D. P. Davis and R. Coimbra, J. Emerg. Med., 2003, 24, 413–422. 21 H. R. Champion, R. F. Bellamy, P. Roberts and A. Leppaniemi, J. Trauma, 2003, 54, 13–19. 22 M. E. Clark, M. J. Bair, C. C. Buckenmaier, R. J. Gironda and R. L. Walker, J. Rehabil. Res. Dev., 2007, 44, 179–194. 23 P. Eyer and D. Podhradsky, Anal. Biochem., 1986, 153, 57–66. 24 J. E. Olerud, L. D. Homer and H. W. Carroll, Arch. Intern. Med., 1976, 136, 692–697; A. Kratz, M. Ferraro, P. M. Sluss and K. B. Lewandrowski, N. Engl. J. Med., 2004, 351, 1548–1563. 25 B. J. Zink, C. H. Schultz, X. Wang, M. Mertz, S. A. Stern and A. L. Betz, Brain Res., 1999, 837, 1–7.

This journal is ª The Royal Society of Chemistry 2010

26 J. C. Rosenberg, R. C. Lillehei, J. Longerbean and B. Zini-Nierinann, Ann. Surg., 1961, 154, 611–627; M. R. Prasad, C. Ramaiah, T. K. McIntosh, R. J. Dempsey, S. Hipkeos and D. Yurek, J. Neurochem., 1994, 63, 1086–1094. 27 K.-K. Ker-Kan Tan, S.-L. Bang, A. Vijayan and M.-T. Chiu, Injury, 2009, 40, 978–983. 28 J. A. Kline, P. C. Maiorano, J. D. Schroeder, R. M. Grattan, T. C. Vary and J. A. Watts, J. Mol. Cell. Cardiol., 1997, 29, 2465–2474. 29 T. Weinstein, A. Chagnac, A. Korzets, M. Boaz, Y. Ori, M. Herman, T. Malachi and U. Gafter, Nephrol., Dial., Transplant., 2000, 15, 883– 887; R. Rossi, A. Milzani, I. Dalle-Donne, D. Giustarini, L. Lusini, R. Colombo and P. Di Simplicio, Clin. Chem. (Washington, DC, U. S.), 2002, 48, 742–753. 30 A. Bhasale, Fam. Pract., 1998, 15, 308–318. 31 J. A. Pineda, K. K. W. Wang and R. L. Hayes, Brain Pathol., 2004, 14, 202–209. 32 L. Papa, G. Robinson, M. Oli, J. Pineda, J. Demery, G. Brophy, S. A. Robicsek, A. Gabrielli, C. S. Robertson, K. K. Wang and R. L. Hayes, Expert Opin. Med. Diagn., 2008, 2, 937–945. 33 I. Mosser, G. Jobst, P. Svasek, M. Varahram and G. Urban, Sens. Actuators, B, 1997, 44, 377–380.

This journal is ª The Royal Society of Chemistry 2010

34 A. P. Wheeler and G. R. Bernard, N. Engl. J. Med., 1999, 340, 207– 214. 35 N. D. Maynard, D. J. Bihari, R. N. Dalton, R. Beale, M. N. Smithies and R. C. Mason, Chest, 1997, 111, 180–187. 36 K. A. Kudsk, M. A. Croce, T. C. Fabian, G. Minard, E. A. Tolley, H. A. Poret, M. R. Kuhl and R. O. Brown, Ann. Surg., 1992, 215, 503–513. 37 D. P. Rignault, World J. Surg., 1992, 16, 940–946. 38 S. D. Sprules, J. P. Hart, S. A. Wring and R. Pittson, Anal. Chim. Acta, 1995, 304, 17–24. 39 I. Wedmore, J. G. McManus, A. E. Pusateri and J. B. Holcomb, J. Trauma, 2006, 60, 655–658. 40 D. Lakstein, A. Blumenfeld, T. Sokolov, G. Lin, R. Bssorai, M. Lynn and R. Ben-Abraham, J. Trauma, 2003, 54, 221–225. 41 G. Lanzino, N. F. Kassell, T. Germanson, L. Truskowski and W. Alves, J. Neurosurg., 1993, 79, 885–891. 42 W. M. Manger, J. L. Bollman, F. T. Maher and J. Berkson, Am. J. Physiol., 1957, 190, 310–316. 43 H. Sies, Exp. Physiol., 1997, 82, 291–295. 44 K. H. Taber, D. L. Warden and R. A. Hurley, J. Neuropsychiatry Clin. Neurosci., 2006, 18, 141–145.

Analyst, 2010, 135, 2249–2259 | 2259