Identification and Quantification of Intravenous Therapy DrugsUsing Normal Raman Spectroscopy and Electrochemical Surface-Enhanced Raman SpectroscopyStephanie Zaleski,† Kathleen A. Clark,† Madison M. Smith,† Jan Y. Eilert,‡ Mark Doty,‡

and Richard P. Van Duyne*,†,#,§

†Department of Chemistry, Northwestern University, 2145 Sheridan Road, Evanston, Illinois 60208, United States#Department of Biomedical Engineering, Northwestern University, 2145 Sheridan Road, Evanston, Illinois 60208, United States§Program in Applied Physics, Northwestern University, 2145 Sheridan Road, Evanston, Illinois 60208, United States‡Baxter Healthcare Corporation, 25212 W. Illinois Rt. 120, Round Lake, Illinois 60073, United States

*S Supporting Information

ABSTRACT: Errors in intravenous (IV) drug therapies can cause human harm and even death.There are limited label-free methods that can sensitively monitor the identity and quantity of thedrug being administered. Normal Raman spectroscopy (NRS) provides a modestly sensitive, label-free, and completely noninvasive means of IV drug sensing. In the case that the analyte cannot bedetected within its clinical range with Raman, a label-free surface-enhanced Raman spectroscopy(SERS) approach can be implemented to detect the analyte of interest. In this work, we demonstratetwo individual cases where we use NRS and electrochemical SERS (EC-SERS) to detect IV therapyanalytes within their clinically relevant ranges. We implement NRS to detect gentamicin, acommonly IV-administered antibiotic and EC-SERS to detect dobutamine, a drug commonlyadministered after heart surgery. In particular, dobutamine detection with EC-SERS was found tohave a limit of detection 4 orders of magnitude below its clinical range, highlighting the excellentsensitivity of SERS. We also demonstrate the use of hand-held Raman instrumentation for NRS andEC-SERS, showing that Raman is a highly sensitive technique that is readily applicable in a clinicalsetting.

There is a critical need to accurately monitor drugs toprevent adverse drug events (ADEs), which include errors

such as incorrect drug type prescribed for the patient or drugmislabeling, incorrect concentration, and simultaneous deliveryof incompatible drugs. The number of reported ADEs ininfusion pumps totaled 56 000 in a 4 year period, 710 of whichled to death.1 The primary means of administering drugs ininfusion pump lines is preprogrammed drug libraries thatspecify dose limits of particular medications.2 Although thismethod is useful for controlling infusion rates, it does not havethe ability to accurately identify and quantify the medicationadministered or to determine whether it is correct. While it ispossible to identify and quantify administered drugs usingreporter systems such as functionalized nanoparticles,3−5

reporter systems cannot be introduced into a patient’sintravenous (IV) line due to safety concerns. Therefore, highlysensitive, noninvasive techniques are required to accuratelymonitor administered IV therapy drugs. There is also a need toaccurately and rapidly identify the concentrations of com-pounded solutions or solutions in which a drug composition isspecifically altered for the needs of an individual patient; errorsin drug compounding of a steroid drug led to a majormeningitis outbreak in 2012.6,7 Recent efforts to sensitivelymonitor therapeutic drugs involve the development ofnanoscale biosensors, which relate to the careful monitoring

of an administered drug over time or identifying drug-inducedphysiological changes that occur. Nanoscale biosensors areadvantageous over well-established biosensing techniquesbecause they typically require minimal sample preparation,are relatively low cost, and often provide a higher level ofdetection sensitivity. For example, optical nanoscale biosensorsbased on detection via fluorescence spectroscopy, surfaceplasmon resonance (SPR) spectroscopy, and surface-enhancedRaman spectroscopy (SERS) are attractive candidates fortherapeutic drug monitoring. The aforementioned techniquestypically rely on monitoring a change in optical response uponthe drug of interest binding to the optically active surface. Wedirect the reader’s attention to a recent review highlighting theapplication of nanoscale biosensors for therapeutic drug sensingfor more detail.8

Despite the sensitivity of fluorescence and SPR, thesetechniques are not label-free, typically requiring an indirectreporter molecule or binding event to occur in order to sensethe target analyte. Alternatively, Raman spectroscopy is highlymolecule-specific, sensitive, and label-free. Additionally, the cost

Received: November 22, 2016Accepted: January 17, 2017Published: January 17, 2017

Article

pubs.acs.org/ac

© XXXX American Chemical Society A DOI: 10.1021/acs.analchem.6b04636Anal. Chem. XXXX, XXX, XXX−XXX

of hand-held Raman systems has decreased significantly in thepast few years, making the use of Raman spectroscopy highlyfeasible in a clinical setting. Raman scattering occurs whenscattered photons are shifted in energy from the incidentphotons; the difference in incident and scattered photon energycorresponds to a specific molecular vibration. As noted above,Raman spectroscopy is an ideal technique for detecting drugsbecause of its high molecular specificity and linear correlationof signal intensity to analyte concentration. Additionally, waterdoes not have a strong Raman profile, making the technique anideal candidate for acquiring measurements of solution-phaseanalytes. Although Raman is a powerful technique, clinicalanalyte concentrations are often too low to be detected usingnormal Raman spectroscopy (NRS). In these cases, very lowconcentrations can be detected by implementing a phenomen-on known as surface-enhanced Raman spectroscopy (SERS),first reported by Jeanmaire and Van Duyne in 1977.9 The originof the high enhancements of SERS stems from the excitation ofa localized surface plasmon resonance (LSPR) of a noble metalnanoparticle substrate, which in turn generates a strongelectromagnetic field near the surface. It has been demonstratedthat the SERS signals of ensemble-averaged molecules exhibitenhancements up to 8 orders of magnitude compared to thenormal Raman signal.10,11 In recent years, SERS has establisheditself as a powerful method with unparalleled sensitivity and thelowest known limits of detection. Because of this, SERS hasbeen used in applications such as bacteria sensing,12,13 dyedetection for art conservation,14 and anthrax detection.15,16

There has been a recent push to develop robust, highlyenhancing SERS-active platforms, which range from nano-particle aggregates to lithographically fabricated substrates.11,17

One lithography-based device, coined film over nanosphere(FON), is a highly reproducible, low-cost SERS platform thatwas developed in the Van Duyne group. FONs are produced atrelatively low-cost, have enhancement factors on the order of105−108, and have predictable SERS enhancement with lessthan 10% SERS enhancement factor (EF) variability across a 25mm area.18

In order to detect analytes with SERS, the target analytemust either sufficiently bind to the SERS substrate or be inclose proximity (<3 nm) from the enhancing surface. Recentwork from our group highlights the importance of the distancedependence of SERS for molecular sensing: Masango et al.applied layers of Al2O3 via atomic layer deposition (ALD) onan AgFON substrate. When fitting the SERS response versusAl2O3 thickness, they found that the SERS signal had a two-term distance dependence. The SERS intensity of the CH3bending mode in trimethylaluminum, an Al2O3 ALD precursor,was found to drop by 20% with an Al2O3 thickness 0.7 nmdeposited on the AgFON surface and by 7% with 3 nm Al2O3,which was further corroborated using DFT calculations.19

These findings imply that, in order to obtain the maximumsensitivity in SERS sensing, the target analyte should be bounddirectly to the SERS-active surface, if possible. In most sensingapplications, SERS substrates are tailored to attract specificanalytes of interest, which has been done for detection ofanalytes such as chloride ions20,21 and glucose.22−24 Recentwork has demonstrated the application of SERS to monitortherapeutic drugs such as antibiotics,3,25,26 methotrexate,27 and5-fluoroacil.28 However, the extension of these studies to moregeneral molecular sensing is limited by the use of a specificcapture layer on the surface. The presence of a capture layernot only can limit the generalized use of the SERS substrate for

a spectrum of analytes but also will decrease the SERS signalobtained, ultimately decreasing the sensitivity as dictated by thedistance dependence. Additionally, nanoparticle SERS sub-strates are typically heterogeneous, causing variability in theenhancement, and can be unstable over long periods of time,potentially limiting their commercial viability.Alternatively, electrochemical SERS (EC-SERS) is a label-

free SERS detection method that can be utilized in cases wherethe analyte does not bind to the SERS substrate. In thistechnique, the potential applied to the substrate can becontrolled to electrostatically attract the analyte of interest tothe SERS-active surface. When detection is completed, theapplied potential can be changed to remove the analyte andreuse the substrate. Using EC-SERS for molecular sensing alsomeans that the molecule is in contact with the surface, and themaximum signal enhancement can be achieved. EC-SERS haspreviously been used to detect molecules such as dopamine inmicromolar concentration solutions at neutral pH.29 Recentwork from Brosseau and co-workers has demonstrated the useof a low-cost EC-SERS setup to detect uric acid in syntheticurine samples at clinically relevant concentrations on an Au/Agmultilayer nanoparticle working electrode.30,31

The overarching goal of this work is to develop a sensingplatform to rapidly monitor drug identity and concentration ina clinical setting or for drug compounding using the molecularspecificity and extreme mass sensitivity of Raman spectroscopy.Viability in clinical settings also signifies relevance in drugcompounding settings, which deal with drugs in highervolumes, equal or higher concentrations, and fewer instrumen-tation and financial constraints. In this paper, we demonstratetwo distinct approaches to identify clinically relevant analytes.First, we use NRS with a hand-held device to detect andquantify gentamicin, an antibiotic commonly used to treat awide variety of Gram-negative and Gram-positive bacterialinfections. We then use a label-free EC-SERS approach todetect dobutamine, a catecholamine that is commonlyadministered after heart surgery, which does not havedetectable NRS signals in the clinical range and does notbind strongly to SERS substrates.

■ EXPERIMENTAL DETAILS

Chemicals. Hydrogen peroxide solution 30% (H2O2),ammonium hydroxide solution 28−30% (NH4OH), sodiumhydroxide (NaOH), 1 N hydrochloric acid (HCl), andgentamicin 50 mg/mL standard solution in deionized waterwere purchased from Sigma-Aldrich and used without furthermodification. Gentamicin in 0.9% sodium chloride IV bagsolution (2 mg/mL) and dobutamine IV bag solutions in 5%dextrose and 1% sodium bisulfite (1, 2, and 4 mg/mL, pH =3.5) were received from Baxter; gentamicin dilutions wereprepared in 0.9% sodium chloride, and dobutamine dilutionswere prepared in Milli-Q (MQ) water. Milli-Q water with aresistivity higher than 18.2 MΩ cm was used in all preparations.

FON Fabrication. Twenty-five mm diameter circularpolished Si wafers were purchased from Wafernet, Inc. TheSi wafers were first cleaned with Piranha solution for 30 min(3:1 H2SO4/H2O2), rinsed copiously with MQ H2O, and thentreated with 5:1:1 H2O/H2O2/NH4OH for 45 min to renderthe surface hydrophilic. The wafers were stored in MQ H2Oprior to use. 540 nm SiO2 microspheres (Bangs Laboratories,Indiana) were diluted to 5% with MQ water, and 10−12 μLwas dropcasted onto the Si wafer and allowed to dry. After

Analytical Chemistry Article

DOI: 10.1021/acs.analchem.6b04636Anal. Chem. XXXX, XXX, XXX−XXX

B

drying, 150 nm Au was thermally deposited on the FON masksurface (PVD-75, Kurt J. Lesker).Bulk Electrochemistry. Bulk electrochemical measure-

ments were performed in a capped scintillation vial. A polishedAu disc electrode was utilized as the working electrode and wassubmerged in solution approximately 1 cm above the Pt wirecounter electrode. The reference potential was determined by aleak-free Ag/AgCl reference electrode (Harvard Apparatus).Electrochemical measurements were performed using a CHInstruments potentiostat (CHI660D).EC-SERS Sample Preparation. AuFON working electro-

des were prepared by first cutting the as-deposited FON into 1cm2 pieces with a diamond scribe pen. A 0.25 mm diameter Agwire (Alfa Aesar) was then attached to the FON usingconductive Ag epoxy (Ted Pella) to allow for electrical contactwith the AuFON. A 2 mm diameter leak-free Ag/AgClelectrode (Harvard Apparatus) and a 1 cm length, 0.5 mmdiameter Pt wire (Alfa Aesar) were utilized as the referenceelectrode and counter electrode, respectively. A #1.5 glasscoverslip bottom well plate (Mattek Corporation) was used asthe cell for EC-SERS measurements, as illustrated in theschematic in Figure S1. Approximately 1.5 mL of the solutionof interest was pipetted into a well, and the electrodes weresuspended in the solution of interest using a rubber septum.The potential was controlled using a CH Instrumentspotentiostat (CHI660D).Instrumentation. LSPR measurements were acquired using

a fiber light spectrometer (Ocean Optics), with a flat Au 150nm film deposited on a cleaned glass coverslip as a flat mirrorreference. Hand-held Raman measurements were performedusing a CBEx hand-held Raman spectrometer with 785 nmexcitation, 50 mW power, and various acquisition times.Tabletop normal Raman measurements were performed usinga 785 nm laser (Innovative Photonic Solutions); the Ramanscattered light was collected and redispersed onto a LS785spectrometer (Princeton Instruments) with a 600 groove/mmgrating blazed at 750 nm. EC-SERS measurements wereperformed using an inverted microscope (Nikon Eclipse Ti-U),where the 785 nm laser excitation was focused onto the sampleand the scattered light was collected using a 20× objective

(Plan Fluor, NA = 0.45, Nikon). The laser light was filteredusing a 785 nm long pass filter (Semrock) and focused onto a1/3 m spectrometer (SP2300, Princeton Instruments). Thefocused light was then dispersed (600 groove/mm grating,1000 nm blaze) and focused onto a liquid nitrogen-cooledCCD detector (Spec10:400BR, Princeton Instruments). LSPRand Raman spectra were processed with OriginLab 8.0 andMATLAB.

■ RESULTS AND DISCUSSIONNormal Raman Spectroscopy of Gentamicin. The

primary model drug used for antibiotic detection andquantification via NRS in this work is gentamicin. Gentamicinis a heat-stable protein synthesis inhibitor used to treat Gram-negative and Staphylococcus bacterial infections and inorthopedic surgery. It is typically administered intravenouslyat 2 mg/mL (4.3 mM) and at pH 3.0−5.5.32 We analyzed theNRS spectra for nine reference gentamicin solutions ranging inconcentration from 0.5 to 50 mg/mL (1.12−112 mM) usingboth a macro Raman instrumental setup and the CBEx hand-held Raman spectrometer. Each data point presented is anaverage of five acquired spectra at an acquisition time of 5 seach. The most prominent spectral features of gentamicin werea major mode at 980 cm−1 and a less intense mode at 790 cm−1

(Figure 1), which we tentatively assign to C−O−C stretchingand C−H rocking modes, respectively. We then generated NRSlinear profiles of concentration versus integrated signal intensityusing the 980 and 790 cm−1 modes.Prior to acquiring NRS spectra of gentamicin using the CBEx

hand-held Raman spectrometer, we compared the spectralresolution of the CBEx to the macro Raman instrumental setupused. First, we acquired an NRS spectrum of cyclohexane, aRaman calibration standard, and compared the peak width ofthe 801.3 cm−1 mode. The macro Raman setup had a peak full-width half-maximum (fwhm) of 12 cm−1, and the CBEx had apeak fwhm of 18 cm−1 (Figure S2B). Despite the 6 cm−1

difference in fwhm, we found that the spectral quality of NRSspectra acquired using the CBEx is comparable to that of astandard macro Raman instrument, as displayed with the NRSspectra of the drug propofol in Figure S3.

Figure 1. Normal Raman spectra of gentamicin solutions in MQ H2O at various concentrations compared to the pure gentamicin solid. Acquisitionparameters: λex = 785 nm, Pex = 50 mW, and tacq = 5 s.

Analytical Chemistry Article

DOI: 10.1021/acs.analchem.6b04636Anal. Chem. XXXX, XXX, XXX−XXX

C

The mean integrated peak intensities of the 980 and 790cm−1 modes versus concentration of gentamicin show anexcellent linear relationship with R2 values of 0.997 and 0.994for the standard Raman instrument and 0.999 and 0.999 for theCBEx hand-held Raman spectrometer, respectively (Figure 2).We found that the integrated peak area of each mode shows asimilar linear dependence as a function of concentration withR2 = 0.996 and R2 = 0.986 for the standard macro Ramaninstrument and R2 = 0.999 and R2 = 0.992 for the CBEx hand-held spectrometer, respectively (Figure S4). This strong lineartrend with both the macro Raman setup and the CBEx hand-held Raman spectrometer demonstrates the ability of NRS tosensitively and rapidly quantify antibiotic concentrations andthe utility of hand-held Raman spectrometers for accuratequantitative Raman measurements.We note that we detected Raman signal from lower

concentrations of gentamicin at the clinically relevantconcentration and partial signal below the clinical concentrationin our prepared solutions (Figure 2). In order to verify thecongruence of commercial gentamicin samples with ourprepared solutions, we then analyzed solutions prepared froma 2 mg/mL commercial gentamicin IV bag solution receivedfrom Baxter Healthcare Corporation. This commercialgentamicin solution clearly demonstrated the mode at 980cm−1. The confirmed ability to detect NRS of gentamicin in acommercial solution within its clinical range shows promise forthe use of hand-held Raman to identify antibiotics and otherdrugs in a clinical setting.The linear dependence of Raman signal intensity acquired on

a standard macro Raman setup versus gentamicin concentrationexhibits an interval within 95% confidence of no greater than

0.131 ADU/mW/s aside from 0.234 ADU/mW/s at 50 mg/mL (Table S1). The clinically relevant concentrations tested onthe macro Raman setup, 2 and 4 mg/mL, showed relativelysmall confidence intervals of 0.048 and 0.057 ADU/mW/s,respectively (Table S1). The corresponding experimentperformed on the hand-held Raman CBEx device showedconsiderably increased precision. The linear dependence ofRaman signal intensity acquired on the CBEx hand-held Ramandevice versus gentamicin concentration exhibits an intervalwithin 95% confidence of no greater than 0.014 ADU/mW/s atany tested concentration aside from 50 mg/mL, which showeda relatively small confidence interval of 0.026 ADU/mW/s(Table S2). The concentrations tested within the clinicallyrelevant regime, 2 and 4 mg/mL, showed particularly smallconfidence intervals of 0.009 and 0.010 ADU/mW/s,respectively (Table S2). We also note that each Ramanmeasurement has an acquisition time of 5 s per 5 acquisitions,further demonstrating the rapid quantitative nature of hand-held NRS experiments. This precise and well-defined relation-ship demonstrates the strength of this antibiotic’s Ramanspectrum as an accurate method to quantify drug concen-trations both within and out of a clinically relevantconcentration range using both a tabletop Raman setup and ahand-held Raman spectrometer.In addition to quantification of gentamicin concentration

with NRS, we examined the effect of pH on the Ramanspectrum of gentamicin due to the possible variability of pH inthe commercial IV bag solution. The commercial solution of 2mg/mL gentamicin taken directly from the IV bag was pH =4.73. Solutions were prepared at ten pH levels ranging from 2to 11, and after adjusting the peak intensity at 980 cm−1 for the

Figure 2. Integrated peak intensities at 980 cm−1 (A and C) and 790 cm−1 (B and D) of gentamicin acquired on the macro Raman instrument (Aand B) or the portable CBEx Raman spectrometer (C and D). Each data point is the mean of 5 spectra. Acquisition parameters for all spectra: λex =785 nm, Pex = 50 mW, and tacq = 5 s.

Analytical Chemistry Article

DOI: 10.1021/acs.analchem.6b04636Anal. Chem. XXXX, XXX, XXX−XXX

D

added volume of NaOH or HCl, we found the range of Ramansignal intensity as a function of pH did not vary significantly(Figure S5). The minimal change in pH shows that gentamicinsolutions can be quantified across a wide range of pH values.Overall, using a hand-held Raman instrument is an ideal meansof rapidly quantifying drug concentrations in a clinical settingor for drug compounding.Electrochemical SERS of Dobutamine. In the case that

the Raman signal is not detectable at clinically relevantconcentrations, one can implement SERS to sufficiently amplifythe Raman signal. The target analyte in this study wasdobutamine, a drug used for improving blood flow and relievingsymptoms of heart failure.33 It is most commonly administeredintravenously in concentrations ranging from 0.5 to 4 mg/mL(1.5−12 mM) at pH 3.5−3.7. The most common commercialIV bag concentrations are 1, 2, and 4 mg/mL (3, 6, and 12mM). Additional components of the commercial IV bagsolution are 5% dextrose, edetate disodium dihydrate, and 1%sodium bisulfite. We were not able to detect dobutamine withinthe clinical range using NRS. We also found that dobutaminewas not detectable by SERS using a bare, unfunctionalizedAuFON due to weak binding of dobutamine to the AuFONsurface.In order to reliably detect dobutamine with SERS, we chose

to implement EC-SERS. Previous work has successfullydemonstrated SERS of various catecholamines at neutral pHusing an electrochemically roughened Ag working electrode.29

First, we characterized the solution-phase electrochemistry ofthe dobutamine IV bag solution with an Au working electrode.The solution phase cyclic voltammogram (CV) using apolished Au disc working electrode is displayed in Figure 3A.Dobutamine is a catecholamine and undergoes a reversible 2-electron, 2-proton transfer to form its quinone species (Figure3B).34,35

EC-SERS measurements were performed using an AuFONworking electrode; an AuFON working electrode is a low-cost,highly enhancing SERS-active surface, and making electricalcontact with the AuFON surface is trivial.36,37 The FONelectrode was fabricated by drop casting 540 nm diameter SiO2microspheres on a cleaned 25 mm diameter Si wafer. After thespheres dried in a hexagonal close packed array on the surface,150 nm Au was thermally deposited on the surface (Figure 4A).The LSPR was measured in air and in the dobutamine solution,as the LSPR between a FON in air and in dobutamine solutionchanges due to the change in local refractive index at the SERS-

active surface (Figure 4B).38 The LSPR of the AuFON workingelectrode in dobutamine overlaps well with the 785 nmexcitation wavelength (red dashed line, Figure 4B) and thewavelength region of the Raman scattered light (red bar, Figure4B), ensuring optimal SERS enhancement.39,40

We first characterized the EC-SERS response of dobutamine:peaks at 596, 640, 792, 823, 1203, and 1605 cm−1 are inexcellent agreement with the dobutamine NRS spectrum. Inparticular, the peak at 1605 cm−1 is assigned to the N−Hvibration of the secondary amine. Additionally, there are peaksbetween 1100 and 1500 cm−1 in Figure 5 that do not appear inthe solution phase dobutamine NRS spectrum; these modes arecharacteristic of a catechol moiety bound to Au.29,41,42 We thendetermined the optimal potential to apply to the AuFONworking electrode to yield the strongest EC-SERS signal byapplying potential stepwise from −0.1 to −0.9 V vs Ag/AgCl in

Figure 3. (A) Cyclic voltammogram of 12 mM dobutamine in 0.5% sodium bisulfite at pH = 3.5. Au disc working electrode, Pt wire counterelectrode, and Ag/AgCl reference electrode. (B) Schematic of 2-electron, 2-proton transfer of dobutamine from its catechol to quinone species.

Figure 4. (A) Representative SEM image of 540 nm SiO2 spheresdropcasted on a cleaned Si wafer with 150 nm Au thermally depositedon top of the spheres. (B) LSPR of AuFON in air (black trace) and in4 mg/mL aqueous dobutamine solution (red trace). The red barrepresents the wavelength region of Raman scattered light of interestin this work.

Analytical Chemistry Article

DOI: 10.1021/acs.analchem.6b04636Anal. Chem. XXXX, XXX, XXX−XXX

E

0.1 V intervals. As shown in Figure 6, there is SERS signal at−0.1 V that increased to a maximum at −0.4 V (Figure 6). As

the applied potential is swept to more negative values, there is asignal intensity decrease beginning at −0.5 V and the SERSsignal then increases in intensity at more negative potentials butdoes not surpass the intensity at −0.4 V. The decrease in signalbetween −0.5 and −0.8 V may be due to the oxidation ofdobutamine to its quinone form and its relatively weak bindingaffinity for the AuFON working electrode surface.43 On thebasis of these results, we chose to apply a constant potential of−0.4 V for detection of dobutamine for all of the followingmeasurements in the study, unless otherwise noted. We alsonote that this is the first study to demonstrate EC-SERS ofsecondary catecholamines at acidic pH.Precision and accuracy experiments were then performed to

demonstrate the viability of using an AuFON working electrodeas a sensing platform. Three separate aliquots of a 2 mg/mLcommercial dobutamine IV bag solution were analyzed usingthe same AuFON working electrode; EC-SERS spectra wereacquired from 5 random spots on the AuFON surface. Washingsteps with MQ water were performed in between each aliquotmeasurement. The average peak intensities of the 1605 cm−1

mode for each aliquot step are displayed in Figure 7. Theaverage peak intensity across the three washing steps does not

decrease significantly, which demonstrates the stability andreusability of the AuFON working electrode for multiple EC-SERS measurements. However, the background signal afterwashing increased after the second wash step, indicating thatsome dobutamine may still be bound to the AuFON surface.Lastly, we performed a limit of detection (LOD) study to

determine the sensitivity of the EC-SERS technique. Weprepared serial dilutions of the commercial dobutamine IV bagsolution ranging from 100 ng/mL to 1 mg/mL (300 nM to 3mM) and took SERS spectra with the potential held constant at−0.4 V (vs Ag/AgCl). Each data point in Figure 8 is an average

of 3−5 SERS spectra, where each spectrum is a different spoton the AuFON working electrode surface. We then used theintegrated peak intensity of the 1605 cm−1 peak to fit the EC-SERS data to a Langmuir adsorption isotherm:15,16

θ = =+

II

K DK D

[ ]1 [ ]

1605

1605,max

dobut

dobut (1)

= +I K I D I

1 1 1[ ]

1

1604 dobut 1604 1604,max (2)

where θ is the fractional surface coverage, I1605 is thenormalized peak intensity at 1605 cm−1, [D] is the dobutamineconcentration in mM, and Kdobut is the binding constant. Wedetermined the Kdobut from the Langmuir isotherm fitted data

Figure 5. Representative 12 mM (4 mg/mL) dobutamine EC-SERSspectrum at −0.4 V (black trace) compared to NRS of 0.1 M soliddobutamine in MeOH (red trace). MeOH peaks are starred. SERSdata: λex = 785 nm, Pex = 980 μW, and tacq = 30 s. NRS data: λex = 785nm, Pex = 3.4 mW, and tacq = 120 s. Both data sets were acquired usinga 20× microscope objective.

Figure 6. Dobutamine EC-SERS 1605 cm−1 mode integrated peakintensity as a function of applied potential.

Figure 7. Accuracy measurement of 6 mM (2 mg/mL) dobutaminesolution with water washing steps in between each aliquot. Eachaliquot is measured on the same AuFON substrate. SERS spectraacquisition parameters: λex = 785 nm, Pex = 980 μW, and tacq = 30 s.

Figure 8. Limit of detection determination for dobutamine. Each datapoint is an average of 5 spectra acquired from different spots on theAuFON surface. SERS spectra acquisition parameters: λex = 785 nm,Pex = 980 μW, and tacq = 30 s.

Analytical Chemistry Article

DOI: 10.1021/acs.analchem.6b04636Anal. Chem. XXXX, XXX, XXX−XXX

F

to be 5.7 mM−1 (Figure 8). We then determined the limit ofdetection for dobutamine, which is defined as the peak intensitybeing 3 times greater than the noise level; the LOD ofdobutamine was found to be 3 × 10−7 M, which is 4 orders ofmagnitude below the clinical concentration range and agreeswell with previous LOD studies of catecholamines with EC-SERS.29 This data demonstrates that EC-SERS is an extremelysensitive technique for the label-free detection of clinicallyrelevant drugs that cannot be detected using NRS or that bindweakly to SERS substrates.Lastly, we have demonstrated the feasibility of EC-SERS

detection of dobutamine in a clinical setting by designing a low-cost, disposable EC-SERS device that can be used for either in-line or off-line testing. A photograph of the bare chip and theassembled EC-SERS device is shown in Figure S6. Eachdisposable EC-SERS device is anticipated to cost approximately$3 and can be fabricated to be less than ∼1 in2 using standardprinted circuit board. We used the EC-SERS chip tosuccessfully detect dobutamine taken from a commercial 4mg/mL IV solution, using the CBEx hand-held Ramanspectrometer (Figure 9). The successful detection of dobut-

amine with a portable chip and hand-held Raman spectrometerproves that EC-SERS has the potential to be a low-cost,sensitive sensing technique for detection of clinically relevantanalytes.

■ CONCLUSIONSEither NRS or EC-SERS can be used as a rapid and sensitivetool to monitor drug concentrations in a clinical setting or fordrug compounding. First, we demonstrate the successfuldetection and precise quantification of gentamicin within itsclinically relevant range using both a standard macro and hand-held Raman instrument. In the case that the analyte cannot bedetected within its clinically relevant range with NRS, like inthe case of dobutamine, we can implement SERS. In particular,we implement a label-free EC-SERS detection approach due tothe otherwise weak binding of dobutamine on an AuFONSERS substrate. We demonstrate that EC-SERS can detectdobutamine at clinically relevant concentrations and pH rangewith an LOD of 100 ng/mL (300 nM) and with good accuracyand precision. Additionally, this is the first study to demonstrateEC-SERS of secondary amines at acidic pH. We alsodemonstrate the potential for a low-cost, commercially viableSERS-active chip for performing EC-SERS experiments in aclinical setting. Overall, this work demonstrates that Raman-

based methodologies are a powerful means of facile, rapidmonitoring of drug concentrations in a clinical setting or fordrug compounding applications.

■ ASSOCIATED CONTENT*S Supporting InformationThe Supporting Information is available free of charge on theACS Publications website at DOI: 10.1021/acs.anal-chem.6b04636.

EC-SERS experimental schematic; photo of CBEx andiPhone; additional Raman spectra; integrated peak areasof gentamicin; calculated mean peak intensities; NRSsignal as a function of pH; photographs of the EC-SERSchip device (PDF)

■ AUTHOR INFORMATIONCorresponding Author*E-mail: vanduyne@northwestern.edu.ORCIDRichard P. Van Duyne: 0000-0001-8861-2228Author ContributionsAll authors have given approval to the final version of themanuscript.NotesThe authors declare no competing financial interest.

■ ACKNOWLEDGMENTSThis work made use of the EPIC facility of the NUANCECenter at Northwestern University, which has received supportfrom the Soft and Hybrid Nanotechnology Experimental(SHyNE) Resource (NSF NNCI-1542205); the MRSECprogram (NSF DMR-1121262) at the Materials ResearchCenter; the International Institute for Nanotechnology (IIN);the Keck Foundation; the State of Illinois, through the IIN.

■ REFERENCES(1) Brady, J. L. Association for the Advancement of MedicalInstrumentation 2010, 44, 372.(2) Carayon, P.; Hundt, A. S.; Wetterneck, T. B. Int. J. Med. Inform.2010, 79, 401−411.(3) Zengin, A.; Tamer, U.; Caykara, T. Anal. Chim. Acta 2014, 817,33−41.(4) Rebe Raz, S.; Bremer, M. G. E. G.; Haasnoot, W.; Norde, W.Anal. Chem. 2009, 81, 7743−7749.(5) He, L.; Rodda, T.; Haynes, C. L.; Deschaines, T.; Strother, T.;Diez-Gonzalez, F.; Labuza, T. P. Anal. Chem. 2011, 83, 1510−1513.(6) Wilson, L. E.; Blythe, D.; Sharfstein, J. M. JAMA 2012, 308,2461−2462.(7) Dresser, R. Hastings Center Report 2013, 43, 9−10.(8) McKeating, K. S.; Aube, A.; Masson, J.-F. Analyst 2016, 141,429−449.(9) Jeanmaire, D. L.; Van Duyne, R. P. J. Electroanal. Chem. InterfacialElectrochem. 1977, 84, 1−20.(10) Stiles, P. L.; Dieringer, J. A.; Shah, N. C.; Van Duyne, R. P.Annu. Rev. Anal. Chem. 2008, 1, 601−26.(11) Sharma, B.; Fernanda Cardinal, M.; Kleinman, S. L.; Greeneltch,N. G.; Frontiera, R. R.; Blaber, M. G.; Schatz, G. C.; Van Duyne, R. P.MRS Bull. 2013, 38, 615−624.(12) Premasiri, W. R.; Moir, D. T.; Klempner, M. S.; Krieger, N.;Jones, G.; Ziegler, L. D. J. Phys. Chem. B 2005, 109, 312−320.(13) Shen, H.; Zhou, X.; Zou, N.; Chen, P. J. Phys. Chem. C 2014,118, 26902−26911.(14) Brosseau, C. L.; Casadio, F.; Van Duyne, R. P. J. RamanSpectrosc. 2011, 42, 1305−1310.

Figure 9. EC-SERS of 12 mM (4 mg/mL) dobutamine on an AuFONchip device taken with a CBEx hand-held Raman spectrometer.Acquisition parameters: λex = 785 nm, Pex = 50 mW, and tacq = 1 s.

Analytical Chemistry Article

DOI: 10.1021/acs.analchem.6b04636Anal. Chem. XXXX, XXX, XXX−XXX

G

(15) Zhang, X.; Young, M. A.; Lyandres, O.; Van Duyne, R. P. J. Am.Chem. Soc. 2005, 127, 4484−4489.(16) Zhang, X.; Zhao, J.; Whitney, A. V.; Elam, J. W.; Van Duyne, R.P. J. Am. Chem. Soc. 2006, 128, 10304−10309.(17) Kleinman, S. L.; Frontiera, R. R.; Henry, A.-I.; Dieringer, J. A.;Van Duyne, R. P. Phys. Chem. Chem. Phys. 2013, 15, 21−36.(18) Greeneltch, N. G.; Blaber, M. G.; Henry, A. I.; Schatz, G. C.;Van Duyne, R. P. Anal. Chem. 2013, 85, 2297−303.(19) Masango, S. S.; Hackler, R. A.; Large, N.; Henry, A.-I.;McAnally, M. O.; Schatz, G. C.; Stair, P. C.; Van Duyne, R. P. NanoLett. 2016, 16, 4251−4259.(20) Jin, J.; Ouyang, X.; Li, J.; Jiang, J.; Wang, H.; Wang, Y.; Yang, R.Analyst 2011, 136, 3629−34.(21) Tsoutsi, D.; Montenegro, J. M.; Dommershausen, F.; Koert, U.;Liz-Marzan, L. M.; Parak, W. J.; Alvarez-Puebla, R. A. ACS Nano 2011,5, 7539−7546.(22) Ma, K.; Yuen, J. M.; Shah, N. C.; Walsh, J. T., Jr.; Glucksberg,M. R.; Van Duyne, R. P. Anal. Chem. 2011, 83, 9146−52.(23) Shah, N. C.; Lyandres, O.; Walsh, J. T., Jr.; Glucksberg, M. R.;Van Duyne, R. P. Anal. Chem. 2007, 79, 6927−6932.(24) Yuen, J. M.; Shah, N. C.; Walsh, J. T. J.; Glucksberg, M. R.; VanDuyne, R. P. Anal. Chem. 2010, 82, 8382.(25) Andreou, C.; Mirsafavi, R.; Moskovits, M.; Meinhart, C. D.Analyst 2015, 140, 5003−5005.(26) McKeating, K. S.; Couture, M.; Dinel, M.-P.; Garneau-Tsodikova, S.; Masson, J.-F. Analyst 2016, 141, 5120−5126.(27) Hidi, I. J.; Muhlig, A.; Jahn, M.; Liebold, F.; Cialla, D.; Weber,K.; Popp, J. Anal. Methods 2014, 6, 3943−3947.(28) Farquharson, S.; Shende, C.; Inscore, F. E.; Maksymiuk, P.; Gift,A. J. Raman Spectrosc. 2005, 36, 208−212.(29) Lee, N.-S.; Hsieh, Y.-Z.; Paisley, R. F.; Morris, M. D. Anal.Chem. 1988, 60, 442−446.(30) Robinson, A. M.; Harroun, S. G.; Bergman, J.; Brosseau, C. L.Anal. Chem. 2012, 84, 1760−4.(31) Zhao, L.; Blackburn, J.; Brosseau, C. L. Anal. Chem. 2015, 87,441−7.(32) Salverda, J. M.; Patil, A. V.; Mizzon, G.; Kuznetsova, S.; Zauner,G.; Akkilic, N.; Canters, G. W.; Davis, J. J.; Heering, H. A.; Aartsma, T.J. Angew. Chem., Int. Ed. 2010, 49, 5776−9.(33) Ivanova, O. S.; Zamborini, F. P. Anal. Chem. 2010, 82, 5844−5850.(34) Hawley, M. D.; Tatawawadi, S. V.; Piekarski, S.; Adams, R. N. J.Am. Chem. Soc. 1967, 89, 447−450.(35) Zhang, Z.; Huang, J.; Wu, X.; Zhang, W.; Chen, S. J. Electroanal.Chem. 1998, 444, 169−172.(36) Hulteen, J. C.; Van Duyne, R. P. J. Vac. Sci. Technol., A 1995, 13,1553−1558.(37) Paxton, W. F.; Kleinman, S. L.; Basuray, A. N.; Stoddart, J. F.;Van Duyne, R. P. J. Phys. Chem. Lett. 2011, 2, 1145−1149.(38) Duval Malinsky, M.; Kelly, K. L.; Schatz, G. C.; Van Duyne, R.P. J. Phys. Chem. B 2001, 105, 2343−2350.(39) McFarland, A. D.; Young, M. A.; Dieringer, J. A.; Van Duyne, R.P. J. Phys. Chem. B 2005, 109, 11279−11285.(40) Zhao, J.; Dieringer, J. A.; Zhang, X.; Schatz, G. C.; Van Duyne,R. P. J. Phys. Chem. C 2008, 112, 19302−19310.(41) Salama, S.; Stong, J. D.; Neilands, J. B.; Spiro, T. G. Biochemistry1978, 17, 3781−3785.(42) Wang, P.; Xia, M.; Liang, O.; Sun, K.; Cipriano, A. F.;Schroeder, T.; Liu, H.; Xie, Y.-H. Anal. Chem. 2015, 87, 10255−10261.(43) Puerto, E. d.; Cuesta, A.; Sanchez-Cortes, S.; Garcia-Ramos, J.V.; Domingo, C. Analyst 2013, 138, 4670−4676.

Analytical Chemistry Article

DOI: 10.1021/acs.analchem.6b04636Anal. Chem. XXXX, XXX, XXX−XXX

H