Article pubs.acs.org/ac
Cite This: Anal. Chem. XXXX, XXX, XXX−XXX
Metabolite Biometrics for the Differentiation of Individuals Mindy E. Hair, Adrianna I. Mathis, Erica K. Brunelle, Lenka Halámková, and Jan Halámek* Department of Chemistry, University at Albany, State University of New York, 1400 Washington Avenue, Albany, New York 12222, United States ABSTRACT: Sweat is a biological fluid present on the skin surface of every individual and is known to contain amino acids as well as other low molecular weight compounds.1 Each individual is inherently different from one another based on certain factors including, but not limited to, his/her genetic makeup, environment, and lifestyle. As such, the biochemical composition of each person greatly differs. The concentrations of the biochemical content within an individual’s sweat are largely controlled by metabolic processes within the body that fluctuate regularly based on attributes such as age, sex, and activity level. Therefore, the concentrations of these sweat components are personspecific and can be exploited, as presented here, to differentiate individuals based on trace amounts of sweat. For this concept, we analyzed three model compoundslactate, urea, and glutamate. The average absorbance change from each compound in sweat was determined using three separate bioaffinity-based systems: lactate oxidase coupled with horseradish peroxidase (LOx-HRP), urease coupled with glutamate dehydrogenase (URGlDH), and glutamate dehydrogenase alone (GlDH). After optimization of a linear dependence for each assay to its respective analyte, analysis was performed on 50 mimicked sweat samples. Additionally, a collection and extraction method was developed and optimized by our group to evaluate authentic sweat samples from the skin surface of 25 individuals. A multivariate analysis of variance (MANOVA) test was performed to demonstrate that these three single-analyte enzymatic assays were effectively used to identify each person in both sample sets. This novel sweat analysis approach is capable of differentiating individuals, without the use of DNA, based on the collective responses from the chosen metabolic compounds in sweat. Applications for this newly developed, noninvasive analysis can include the field of forensic science in order to differentiate between individuals as well as the fields of homeland security and cybersecurity for personal authentication via unlocking mechanisms in smart devices that monitor metabolites. Through further development and analysis, this concept also has the potential to be clinically applicable in monitoring the health of individuals based on particular biomarker combinations.
I
of DNA expected from the 4000 to 10 000 leukocytes10 per microliter of blood. However, sweat also consists of compounds such as sodium, chloride, lactate, and urea as well as amino acids and other metabolic products.1,11−16 The method presented in this manuscript focuses on these additional compounds, thus indicating that sweat can serve another purpose to investigators. The purpose of our approach is to introduce sweat analysis as an alternative means of addressing the issues of when biological fluids fail to produce a full DNA profile or when a corresponding DNA profile does not preexist in a database. This concept can also be paralleled in cybersecurity as a way to enhance current security methods when using smart devices or in health applications17 as a method to monitor particular disease biomarkers in sweat. Sweat is a biological fluid that is continuing to attract attention from the scientific community.1,8−16,18 Although more difficult to detect at a crime scene,19 sweat has the advantage of being a noninvasive sample.20,21 The human body contains several million sweat glands spread throughout the
n criminal investigations, biological samples are often essential in discovering valuable information regarding the perpetrator.2,3 Blood is one of the most commonly collected body fluids at crime scenes, because it contains DNA. Since its first appearance4 in court in 1986, DNA analysis has developed into an essential component of modern forensic science with validated protocols that are implemented across the globe to aid in investigations in order to identify criminals and free the wrongfully accused. This increase in awareness of DNA’s ability to help solve criminal cases has led to a surge in the demand for DNA testing nationwide.5,6 Today’s crime laboratories are processing more DNA cases than ever before and struggle to meet the increase in demands. This has led to the growth of backlogged DNA casework and offender samples. The National Institute of Justice (NIJ) defines a backlogged case as one that remains untested for 30 days after submission to a laboratory.5 The purpose of DNA testing is to identify the originator of a biological sample with as much certainty as feasibly possible. In order to test for DNA, biological cells must be present.7−9 Therefore, DNA testing from sweat is available only if there are skin cells swept away with the fluids and oils during the process of sweating. The amount of obtainable DNA from the few skin cells swept away by sweat is significantly less than the amount © XXXX American Chemical Society
Received: January 25, 2018 Accepted: March 21, 2018 Published: March 21, 2018 A
DOI: 10.1021/acs.analchem.8b00414 Anal. Chem. XXXX, XXX, XXX−XXX
Article
Analytical Chemistry epidermis, the outer layer of the skin.22 Individuals are continuously excreting fluids through these glands, because sweating is the main mechanism by which body temperature is regulated.23 Therefore, individuals are leaving a trail of trace amounts of sweat as they travel from location to location and touch various objects and surfaces. The concentrations of the biochemical content within an individual’s sweat are controlled by metabolic pathways24 that fluctuate daily1,25−29 based on factors such as age, sex, diet, and activity level. Since no two individuals will have the same hormone levels or conditions at a given time, the concentrations of these sweat components are specific to each individual. Due to these factors, the biochemical content of sweat can be utilized to obtain valuable information about different persons of interest within a criminal investigation. For example, if multiple sweat samples are found at a crime scene, each sample can be tested to determine the number of different individuals present at the crime scene. The same biochemical information can also be applied to cybersecurity for authentication purposes based on anticipated concentrations of specific metabolites on the surface of the skin.30 This combination of concentrations could potentially be used to unlock smart devices.31−35 Our group aims at using the small molecule content of sweat to distinguish between sweat originators for multiple applications, with a focus on forensic science. This study involves the use of biocatalytic enzyme cascades to focus on the person-specific biochemical content of the sweat, known to contain various amino acids and other low molecular weight compounds.1,13,14 By analyzing multiple metabolites found in trace amounts of sweat, such as lactate, urea, and glutamate,11,12,16 the concentrations of each can be exploited to differentiate individuals. Lactate, a byproduct from the energy producing carbohydrate metabolism of the body, is one of the major components of sweat.12 Lactate as a biomarker has been applied to physiological evaluations for health related purposes, but its potential as a forensic biomarker has yet to be realized in crime laboratories. Urea is the major end product synthesized in the liver from ammonia formed during protein catabolism and amino acid deamination processes.36,37 The elimination of urea in the urine is the main route of excretion for excess nitrogen in the body. Urea is a biomarker of renal hyperuremia, as the concentration of urea in the body directly correlates to kidney function.38 The concentration of urea is dependent on protein intake, protein catabolism, and kidney function. Elevated levels occur with dietary changes, diabetes, congestive heart failure, and diseases of the liver or kidneys.36 Glutamate is an amino acid, formed in the protein catabolism process of metabolism, that is stored and produced in muscle tissue.39 Metabolism is known to cause slight variations in amino acid content based on the physiological state of an individual. Previous research has shown that amino acid levels differ between demographic groups, such as biological sex.40,41 This amino acid contributes to the overall health of the immune and digestive systems through the removal of toxic metabolic waste. Glutamate supplementation is recommended to treat attention deficit disorder (ADD), as it provides energy to the brain, simulates mental alertness, and improves memory function. Other sources of glutamate include protein-rich meats, dairy, and vegetables.39 For the purpose of this experiment, lactate was chosen, because our laboratory has previously demonstrated a method42 for on-site forensic detection of sweat based on the presence of lactate. This method uses a straightforward enzymatic cascade
that exploits the colorimetric properties of the substrates involved to produce a visible color change. Our previous work further established the ability of this technique to distinguish between sweat and saliva. The high specificity and sensitivity of the enzyme cascade allows for lactate detection with as little as 50 nL of sweat. This work served as the basis for this biometric study because of its ability to detect sweat via the presence of low lactate concentrations. Additionally, both urea and glutamate were chosen, as each component possesses an enzymatic assay that is fundamentally understood and applied in clinical diagnostics.36,38,43,44 Furthermore, the expected concentration ranges of lactate, urea, and glutamate are significantly different. This is important because the variation of each metabolite within every individual is proportional to its concentration. For example, the concentration of lactate is the most abundant and the most variable, whereas the concentration of glutamate is the least abundant and least variable of the three metabolites. The combined results of the three metabolite analyses has the ability to differentiate individuals via their unique lactate, urea, and glutamate concentrations.
■
EXPERIMENTAL SECTION Ethics Statement. The Institutional Review Board, Office of Pre-Award and Compliance at the University at Albany has approved the experimental protocols described in this manuscript. For these studies involving the differentiation of an individual based on small molecule content from sweat, we have used mimicked samples based on the physiological concentrations of small molecules known to be present in sweat. Additionally, male and female volunteers provided their sweat for analysis of authentic sweat samples. These volunteers were required to sign a document that stated there will be no risks associated with their participation in this study, as the collection process is noninvasive. This document also states that the volunteers will not receive any benefits such as compensation for their participation. Lastly, the volunteers were informed that no personal information will be used as part of this study with the exception of the small molecule levels of lactate, urea, and glutamate within their sweat. Materials. The following enzymes and reagents were purchased from Sigma-Aldrich: lactate oxidase (LOx, E.C. 1.13.12.4), horseradish peroxidase (HRP, E.C. 1.11.1.7), urease from jack bean type III, (UR, E.C. 3.5.1.5), L-glutamic dehydrogenase from bovine liver type II (GlDH, E.C. 1.4.1.3), 2,2′-azino-bis(3-ethylbenzothiazoline-6-sulfonic acid) (ABTS), α-ketoglutaric acid disodium salt dehydrate (KTG), βnicotinamide adenine dinucleotide reduced dipotassium salt (NADH), β-nicotinamide adenine dinucleotide sodium salt (NAD+), potassium phosphate monobasic, potassium phosphate dibasic, triethanolamine (TEA), Trizma base, sodium Llactate, urea, L-glutamic acid, β-alanine, glycine, L-threonine, Lserine, L-alanine, L-citrulline, L-aspartic acid, L-asparagine, Lglutamine, L-proline, L-valine, L-cystine, L-methionine, Lisoleucine, L-leucine, L-tyrosine, L-phenylalanine, L-ornithine, L-lysine, L-tryptophan, L-histidine, and L-arginine. Hydrochloric acid was purchased from EMD Millipore. An ELGA water purification system, the PURELAB flex, was used to obtain ultrapure (18.2 MΩ·cm) water for each experiment. For the sweat collection process, the BAND-AID brand medium nonstick pads and Coach sport tape were purchased from Johnson and Johnson. The black silicon bands were purchased from Wristband Resources. All analyses were performed at 37 °C using a Molecular Devices SpectraMax Plus 384 UV/vis B
DOI: 10.1021/acs.analchem.8b00414 Anal. Chem. XXXX, XXX, XXX−XXX
Article
Analytical Chemistry
3.5.1.5) and glutamic dehydrogenase (GlDH, E.C. 1.4.1.3). UR is an enzyme specific for hydrolysis of urea to produce ammonium (NH4+). GlDH consumes the ammonium, KTG, and NADH as substrates to produce glutamate and NAD+. This conversion of NADH to NAD+ generates a decrease in absorbance at 340 nm. This two-enzyme cascade was optimized and realized in 1 M Tris-HCl buffer pH 7.8 containing 1 U UR, 1 U GlDH, 0.5 mM KTG, and 0.3 mM NADH. Method for Detection of Glutamate. The assay47 for glutamate detection, depicted in Scheme 3, utilizes GlDH which consumes glutamate and NAD+ as substrates. This reaction results in the production of KTG, NH4+, and NADH. The conversion of NAD+ to NADH generates a visible increase in signal at 340 nm. This single enzyme assay was optimized and realized in 0.2 M TEA buffer pH 7.6 containing 1 U GlDH and 1 mM NAD+. Small Molecule Detection from Mimicked Sweat. Buffer-based solutions were prepared to mimic the physiological concentrations of lactate, urea, and 23 amino acids,40,48−50 including glutamate, found in sweat for a total of 25 metabolites. The distribution of lactate and urea concentrations were determined from previous studies.11,12 The values for glutamate and the remaining amino acids were stated in statistical studies that were positively skewed and consistent with a log-normal distribution.40,41 The parameters of the log-normal distribution were available only for overall metabolite concentrations, while the distribution parameters estimated for each mimicked sample came from logarithmic untransformed data. The existing parameters for a normal distribution were first modified for a log-normal distribution. For each of the metabolic components present in sweat, statistically generated values agreeing with the recalculated parameters of the log-normal distribution were established using R-project software.51−56 As a result, 50 concentrations per metabolite were statistically generated, resulting in 1250 different marker levels, which were then arbitrarily grouped together to create 50 sweat samples representing the population. Each individual sample was evaluated in triplicate using the LOx-HRP, URGlDH, and GlDH biocatalytic assays, separately. This resulted in a total of nine analyses per mimicked sample. After analysis of these mimicked samples, the focus shifted to the analysis of authentic sweat samples. Small Molecule Detection from Authentic Sweat. To determine the viability of these biocatalytic assays, a collection and extraction method was developed and optimized by our group to evaluate real sweat samples from the skin surface using an absorbent pad. Similar sweat collection methods have been used to detect drugs of abuse and have been proven to serve as an alternative to urinalysis for cocaine detection.57−60 Following collection and extraction, the authentic sweat samples were analyzed utilizing the same methods as the mimicked samples. Sweat Collection and Extraction. The sweat pad used for sweat collection consisted of a 7.5 × 1 cm section of a Johnson and Johnson BAND-AID brand medium nonstick pad. A sweat pad was placed on the forearm of each volunteer and held in place with athletic tape. To aid in sweat production, a 1 in. black silicone wristband was placed on top of the taped sweat pad. Participants exercised for 30 min to guarantee sufficient sweat production to fully saturate the sweat pad. Saturation of the sweat pad will ensure a more even distribution of sweat components and decrease sampling error. An extraction method was then used to remove the sweat from the pad.
spectrophotometer/plate reader with 96 well microtiter polystyrene plates purchased from Thermo Scientific. Enzymatic Assays. Three single-analyte enzymatic assays were used for the detection of lactate, urea, and glutamate from sweat. These assays are depicted in Schemes 1, 2, and 3, respectively. Scheme 1. Biocatalytic Cascade for Determination of Lactate Content in Both Mimicked and Authentic Sweat Samples
Scheme 2. Biocatalytic Cascade for Determination of Urea Content in Both Mimicked and Authentic Sweat Samples
Scheme 3. Biocatalytic Cascade for Determination of Glutamate Content in Both Mimicked and Authentic Sweat Samples
Method for Detection of Lactate. Our laboratory has published and developed a method for the detection of low volumes of sweat based on the presence of the bimolecular indicator, lactate.45,46 By utilizing this method, trace amounts of sweat can be detected at the crime scene and can provide the first compound needed for analysis, lactate. The assay for lactate determination, depicted in Scheme 1, consists of lactate oxidase (LOx, E.C. 1.13.12.4) and horseradish peroxidase (HRP, E.C. 1.11.1.7). LOx is an enzyme specific for consumption of lactate and oxygen (O2) to produce pyruvate (Pyr) and hydrogen peroxide (H2O2). HRP uses H2O2 as a substrate in order to oxidize the ABTS dye (2,2′azino-bis(3-ethylbenzothiazoline-6-sulfonic acid)) and generate a visible increase in absorbance at 405 nm. This system is realized in potassium phosphate buffer pH 7.6 containing 0.1 U LOx, 0.2 U HRP, and 1 mM ABTS. Method for Detection of Urea. The assay38 for urea detection, depicted in Scheme 2, utilizes both urease (UR, E.C. C
DOI: 10.1021/acs.analchem.8b00414 Anal. Chem. XXXX, XXX, XXX−XXX
Article
Analytical Chemistry
Figure 1. Calibration curves for (A) detection of lactate via LOx-HRP, (B) detection of urea via UR-GlDH, and (C) detection of glutamate via GlDH each demonstrate the dependence of the assay for the respective analyte.
Figure 2. Mimicked sweat samples consisting of the expected physiological concentrations of (A) lactate, (B) urea, and (C) glutamate.
cascades, and lactate possesses the highest concentration expected in sweat. Figure 1A shows the linear dependence of lactate concentrations on the change in absorbance (ΔAbs). Urea detection required 1 μL of sample, as the UR-GlDH assay is rather sensitive to the presence of urea, and the concentration expected in sweat is lower than that of lactate. Figure 1B shows the linear dependence of urea concentrations on change in absorbance following a 5.5 min incubation at 37 °C, because the initial absorbance values reached the saturation limit of the device. Glutamate detection required 100 μL of sample, as the physiological glutamate concentrations are the lowest of all three observed sweat components. Figure 1C shows the linear dependence of glutamate concentrations on change in absorbance for the GlDH assay. These calibration curves demonstrate the correlation between the absorbance change response and concentration of each analyte. Evaluation of Mimicked and Authentic Samples. R-project software51−56 was used to create 50 mimicked sweat samples containing reported values of lactate, urea, and amino acids in sweat. These samples were analyzed separately in triplicate (N = 3) by each enzymatic cascade to demonstrate that each assay functions properly without cross reactivity with the other compounds present in sweat. Again, due to the sensitivity of each enzymatic assay to its respective analyte, only 300 nL of sample was used for lactate detection, 1 μL of sample was used for urea detection, and 100 μL of sample was used for glutamate detection. This is because less sweat is required for analysis of metabolites that are expected to be found in high concentrations, such as lactate, whereas, more sweat is required for analysis of metabolites that are expected to be found in low concentrations, such as glutamate. The results of these analyses are shown in Figure 2. In order to evaluate the same concentrations in humans, both a collection method using an absorbent pad on the skin surface
Each sweat pad was cut into nine identical pieces for triplicate analysis by each enzymatic assay. Next, each sample was incubated in 120 μL of 10 mM HCl for 20 min at 40 °C. Finally, centrifugation was utilized to obtain 100 μL of sweat for analysis. These samples were used to determine the average absorbance change generated from the analysis of lactate, urea, and glutamate. Statistics. In order to determine if the absorbance values of each individual sample are unique to the sample originator and statistically different from other individuals in this study, a multivariate analysis of variance (MANOVA) test was performed. This analysis was performed on both the mimicked and authentic sample sets. MANOVA is a commonly used technique for simultaneously comparing mean values for multiple dependent variables across two or more groups.61 A low p-value supports that there is a significant statistical difference between each individual when considering all three analytes (lactate, urea, and glutamate) combined. If the MANOVA tests are successful, then an analysis of variance (ANOVA) test is performed. ANOVA is another commonly used technique for comparing mean values in order to evaluate the significance of each dependent variable (metabolite) separately.
■
RESULTS Calibration Curves. The calibration curves for the LOxHRP, UR-GlDH, and GlDH assays, depicted in Figure 1, were generated in a buffered solution using a range of expected physiological concentrations for lactate, urea, and glutamate, respectively. Due to the sensitivity of each enzymatic assay, only small volumes of each mimicked sample were required for analysis. Lactate detection only required 300 nL of sample, as the LOx-HRP assay is the most sensitive of the three enzymatic D
DOI: 10.1021/acs.analchem.8b00414 Anal. Chem. XXXX, XXX, XXX−XXX
Article
Analytical Chemistry
Figure 3. Authentic sweat samples consisting of the actual physiological concentrations of (A) lactate, (B) urea, and (C) glutamate.
and an extraction method were developed and optimized by our group. Authentic sweat samples were then collected from 25 volunteers (12 females and 13 males). As shown in Figure 3, the real sweat samples were extracted, and 100 μL of each extracted sweat sample was analyzed in triplicate (N = 3) for each enzymatic assay to determine the average absorbance change from the presence of lactate, urea, and glutamate for each individual. Figures 2 and 3 demonstrate that every individual, represented by the mimicked and authentic samples, respectively, produces a different response based on the individual’s expected and actual physiological concentrations of lactate, urea, and glutamate in sweat. From the mimicked samples (Figure 2), the change in absorbance values at 12 min for each assay were combined to create the 3D scatter plot depicted in Figure 4A. The absorbance changes at this time were chosen because they provide the greatest separation between individuals for each analysis. Figure 4A consists of each mimicked individual’s average absorbance change from lactate, urea, and glutamate and demonstrates that no two mimicked individuals overlap when considering all three analytes at a given time point. From the authentic samples (Figure 3), the change in absorbance values at 12 min were used for urea and glutamate determination; however, the change in absorbance values at 2.5 min were used for lactate determination. A different time point was chosen for authentic lactate samples due to the high lactate concentrations in authentic samples that caused the reaction to reach completion well before 12 min of analysis. The combination of all three average change in absorbance values at the respective time points were used to create the 3D scatter plot in Figure 4B. This plot demonstrates that no two authentic individuals overlap when considering all three analytes at a given time point. For both 3D plots (Figure 4A,B), the x-axis corresponds to the average absorbance change of glutamate, the y-axis corresponds to the average absorbance change of lactate, and the z-axis corresponds to the average absorbance change of urea. Once the axes were defined, each plot was manually oriented in 3D space to provide the best visual separation between each sample. The average change in absorbance values and standard deviations of the authentic sweat samples (N = 3) for each metabolite were used to generate the bar graphs shown in Figure 5. The standard deviations demonstrate the expected variation in each sample’s response, as the concentrations of each metabolite are anticipated to fluctuate. This error also includes additional error from any loss of sweat during the collection and extraction processes. Furthermore, error is anticipated as each individual provided sweat samples on different days and at different times of day.62,63 Figure 5 further
Figure 4. 3D scatter plot with the average (N = 3) change in absorbance value of lactate, urea, and glutamate at their respective time points for the (A) 50 mimicked samples and (B) 25 authentic samples.
demonstrates that no two individuals possess the same levels of all three markers. Statistics. The MANOVA test was performed on each data set, the mimicked and authentic sweat samples, in order to determine if each individual produced a unique combination of all three metabolites in sweat. The p-values from the MANOVA tests for both the mimicked and authentic samples were calculated to be
