Mais J.
Jebrail‡
a,
Hao
Yang‡
a,
Jared M.
Mudrik
a,
Nelson M.
Lafrenière
a,
Christine
McRoberts
b,
Osama Y.
Al-Dirbashi
bc,
Lawrence
Fisher
b,
Pranesh
Chakraborty
bc and
Aaron R.
Wheeler
*ad
aDepartment of Chemistry, University of Toronto, 80 St. George Street, Toronto, Ontario M5S 3H6, Canada. E-mail: aaron.wheeler@utoronto.ca; Fax: +416-946-3865; Tel: +416-946-3864
bNewborn Screening Ontario, Children's Hospital of Eastern Ontario, 401 Smyth Road, Ottawa, Ontario K1H 8L1, Canada
cDepartment of Pediatrics, Faculty of Medicine, University of Ottawa, 451 Smyth Road, Ottawa, Ontario K1H 8L1, Canada
dInstitute of Biomaterials and Biomedical Engineering, University of Toronto, 164 College Street, Toronto, Ontario M5S 3G9, Canada
First published on 25th August 2011
Blood samples stored as dried blood spots (DBSs) are emerging as a useful sampling and storage vehicle for a wide range of applications. Unfortunately, the surging popularity of DBS samples has not yet been accompanied by an improvement in automated techniques for extraction and analysis. As a first step towards overcoming this challenge, we have developed a prototype microfluidic system for quantification of amino acids in dried blood spots, in which analytes are extracted, mixed with internal standards, derivatized, and reconstituted for analysis by (off-line and in-line) tandem mass spectrometry. The new method is fast, robust, precise, and most importantly, compatible with automation. We propose that the new method can potentially contribute to a new generation of analytical techniques for quantifying analytes in DBS samples for a wide range of applications.
We report here a new set of techniques useful for extracting and quantifying analytes for DBS samples. The methods are fast, use small amounts of reagents, and as far as we are aware, represent the first application of microfluidics to DBS samples. (Note that analytes manually extracted from DBS samples have been analyzed by microfluidic techniques,5 but the methods reported here are the first to accept a DBS directly as sample input.) The new methods are powered by digital microfluidics (DMF), a fluid-handling technique in which discrete droplets of samples and reagents are manipulated (i.e., dispensed from reservoirs, split, merged and mixed) on an open surface by applying a series of electrical potentials to an array of electrodes.6,7 Droplet actuation in such systems is driven by electromechanical forces8,9 generated on free charges in the droplet meniscus (in case of conductive liquids) or on dipoles inside of the droplet (in case of dielectric liquids).
The microfluidic DBS analysis methods described here were validated by application to quantifying analytes that are commonly measured as biomarkers for amino acid metabolism disorders in newborn patients: methionine (as a marker for homocystinuria10,11), phenylalanine (as a marker for phenylketonuria12,13), and tyrosine (as a marker for tyrosinemia11,14). These analytes are typically quantified using tandem mass spectrometry (MS/MS), and comprehensive testing represents a large public health undertaking around the world. For example, the Newborn Screening Ontario (NSO) facility at the Children's Hospital of Eastern Ontario10 uses MS/MS to evaluate DBS samples from approximately 140,000 patients each year, which necessitates a large laboratory that employs a large number of technologists and other staff-members.
In a typical screen for amino acid metabolism disorders in a newborn patient, a 3.2 mm diameter disc is punched from a DBS sample on filter paper, and the analytes are extracted, mixed with isotope-labelled internal standards, derivatized, and then reconstituted for analysis by tandem mass spectrometry (MS/MS). As shown in Fig. 1a, the derivatization step transforms each amino acid (AA) to its corresponding butyl ester (derivatized AA) that allows for a characteristic fragmentation pattern (neutral loss of 102) viacollision induced dissociation (CID). Fig. 1b–c contains representative primary (MS1) and secondary (MS2) mass spectra for phenylalanine, with peaks at m/z 222 and 120.
Fig. 1 Processing dried blood spot samples for quantification of amino acids (AAs) by tandem mass spectrometry. (a) Reaction scheme involving derivatization of the extracted AA, followed by derivatization with n-butanol, followed by the formation of a product ion by collision induced dissociation (CID) in the mass spectrometer. (b)Mass spectrum generated from primary analysis (MS1) of derivatized phenylalanine (Phe). (c)Mass spectrum generated from the secondary analysis (MS2) of derivatized Phe showing the loss of 102 amu as a result of CID. |
The process flow shown in Fig. 1 is specific for newborn screening, but many different kinds of applications of DBS samples follow a similar scheme of extraction, derivatization, mixing with standards, purification, and analysis.1,4,15,16 We propose that the new methods reported here may represent a useful new tool for a wide range of applications of dried blood spots.
Fig. 2 Two digital microfluidic methods designed to quantify amino acids in dried blood spot samples. In method 1 (a), a 5 μL droplet of blood is spotted directly onto the device surface and allowed to dry. In method 2 (b), a 3.2-mm diameter punch from filter paper bearing dried blood is positioned on the device surface. |
An experiment using method 1 is depicted in Fig. 3a. As shown, a blood sample is spotted onto the device, dried, extracted into methanol containing isotope-labelled standards, and the solvent is allowed to evaporate. The extract and standards are then derivatized, and the products are isolated by allowing the solvent to evaporate. The entire process requires 50 min to complete. Samples processed by method 1 were collected and analyzed off-line by nanoelectrospray ionization tandem mass spectrometry (nESI-MS/MS) to quantify AAs. Calibration curves with R2 greater than 0.996 (Fig. S1 in the online Supplementary Information†) were generated for Met, Phe, and Tyr by analyzing standards processed by DMF at known concentrations from the abundance ratio of each AA to its deuterated standard peak in the secondary (MS2) spectra.
Fig. 3 Analysis of amino acids in dried blood spots by digital microfluidic method 1. (a) Sequence of frames from a movie depicting several stages in sample processing by DMF including: (1) a dried blood sample prior to processing; (2) mixing and incubating an extractant droplet with the sample; (3) a droplet containing sample extract after translation away from the dried sample; (4) a dried extract; (5) mixing and incubating a derivatization reagent droplet with the dried extract; and (6) the dried, derivatized product. (b) Comparison of Met, Phe, and Tyr concentrations in normal (green) and spiked (red) blood samples as biomarkers for homocystinuria, phenylketonuria, and tyrosinemia, respectively. The dashed lines indicate the upper levels for normal concentrations in newborn blood samples. Each data point represents at least four replicate measurements, and error bars represent ±1 S.D. |
To evaluate the potential of DMF method 1 as a platform for discriminating between disease and healthy states, spiked blood samples (mimicking diseased states) and non-spiked blood samples (mimicking healthy state) were analyzed by mass spectrometry. Fig. 3b shows a comparison of measured concentration of AAs in normal and spiked blood samples. The dashed line indicates typical threshold values that correlate with homocysteinuria (100 μM Met)10,11phenylketonuria (120 μM Phe)12,13 and tyrosinemia (150 μM Tyr).11,14 As shown, the method is useful for distinguishing between these concentrations.
The new method was evaluated for a number of analytical performance parameters, including extraction efficiency, linear dynamic range, and detection limits. Two orthogonal tests were used to evaluate extraction efficiency: fluorescence and MS/MS. In the former, a fluorogenic dye was used to label standard samples dissolved in solvent before and after processing by DMF to determine the concentrations. In the latter, blood samples were spiked with AAs and the recovery efficiency was determined by comparing the measured AA concentration (endogenous plus spiked AA) vs. known concentration. As listed in Table 1, the two orthogonal methods (fluorescence and MS/MS) agree and reveal the new DMF technique to be highly efficient, with recoveries of ≥80% for each AA. More importantly, the precision of these measurements was high, with CVs ranging from 1 to 10%. And as shown in Figure S1 in the online supplementary information,† each measurement had a linear dynamic range of three orders of magnitude, and detection limits were 14.6, 6.0, and 15.2 ng for Met, Phe, and Tyr, respectively.
Amino Acid | % Recovery by Fluorescence ± 1 S.D. (standards in solvent) | % Recovery by MS/MS ± 1 S.D. (standards in blood) |
---|---|---|
Methionine | 98 ± 10 | 100 ± 1 |
Phenylalanine | 86 ± 9 | 85 ± 5 |
Tyrosine | 82 ± 10 | 84 ± 7 |
Digital microfluidic method 2 (Fig. 2b) was developed to analyze punched discs from DBS samples on filter paper. A portion of an experiment is depicted in Fig. 4a. As shown, a droplet of extraction solvent was dispensed and driven to a 3.2 mm diameter filter paper punch, and the extract was then moved away for further processing (i.e., derivatization and solvent exchange, similar to Fig. 3a). After wetting, the filter paper punch remains adhered to the surface through friction and capillary forces, and fluids could be moved onto, through, and off-of the filter paper smoothly. The capacity to easily handle mesoscale solid samples like punched filter paper is a unique advantage of digital microfluidics relative to other miniaturization modalities. For example, 3.2-mm dia. filter paper punches are much larger than the dimensions of standard microchannels, which are moreover susceptible to clogging in applications involving biological samples.17 This is likely the reason why microfluidic devices reported previously for DBS sample analysis5 were used to analyze liquid extract (after off-chip sample processing), rather than the DBS samples, themselves.
Fig. 4 Analysis of amino acids in dried blood spots by digital microfluidic method 2. (a) Frames from a movie depicting sample processing of 3.2 mm diameter punch of a dried blood spot (DBS) on filter paper by digital microfluidics (DMF). (b) Graph of Phe concentrations measured by the digital microfluidic method in DBS punches from three NSO patients. As shown, patients 1 and 3 were correctly diagnosed as having phenylketonuria, and patient 2 was correctly identified as being unaffected. The dashed lines indicates the upper level for normal concentrations of Phe in newborn blood samples. |
To evaluate digital microfluidic method 2 relative to gold standard practices, a series of punches from blood samples containing various concentrations of Phe were processed by DMF method 2, and punches from the same samples were evaluated using the conventional newborn screening technique at NSO. As listed in Table 2, a paired t-test revealed no significant difference between the two data sets at a 95% confidence level. This result is notable, as the two sets of data were generated using different extraction techniques as well as different mass spectrometers (from different manufacturers) in different locations run by unique operators. To validate the new technique for application to clinical samples, DBS punches from three newborn patients of NSO were evaluated by the DMF method. As shown in Fig. 4b, the new technique correctly identified patients 1 and 3 as suffering from phenylketonuria, and patient 2 as being unaffected.
Sample | Measured Phe concentration (μM) using DMF method 2 | Measured Phe concentration (μM) using standard NSO technique |
---|---|---|
1 | 70 | 70 |
2 | 550 | 548 |
3 | 93 | 88 |
4 | 93 | 92 |
5 | 368 | 302 |
6 | 534 | 539 |
7 | 735 | 871 |
While the data described above were generated with microfluidic sample processing and off-line analysis, there may be some cases in which a fully integrated (processing + analysis) method is useful. To accommodate this potential need, a digital microfluidic platform was coupled directly to a nanoelectrospray emitter for in-line analysis by mass spectrometry. As shown in Figure S3 in the online Supplementary Information,† the central feature of this method (building on recent work with hybrid microfluidics18) is a vertical intersection between an array of DMF electrodes on top of a device, and a microchannel on the bottom of the device. A portion of an experiment is depicted in Fig. 5a. A blood sample was first spotted on the device as in DMF method 1 (note that the technique is also compatible with filter paper punches as in DMF method 2) and the AAs were extracted and derivatized as described above. The dried, derivatized sample was then resuspended and the droplet was actuated to the access hole such that it filled the channel below by capillary action. A nanoelectrospray was generated at a corner of the device19,20 (Fig. 5b) by applying a high voltage to the DMF top-plate electrode. Representative mass spectra generated from samples processed and analyzed on-chip are shown in Fig. 5c. The entire process requires ∼1 h from sampling to analysis, and requires only the hybrid DMF device and a mass spectrometer (i.e., no nanoflow pumps, capillary connections, or samplers). While digital microfluidics has been used to handle liquid samples for eventual analysis off-line by electrospray21–24 and MALDI25–31mass spectrometry in previous work, the device shown in Fig. 5 is the first reported to integrate digital microfluidics with in-line mass spectrometry.
Fig. 5 Processing and in-line analysis of amino acids in dried blood spots. (a) Frames from a movie (left-to-right) demonstrating derivatization and extraction of AA from dried blood, resolubilization in solvent, and analyte solution in spray microchannel. (b) Image of sample spraying from an integrated emitter. (c) MS2 spectra of Phe (left) and d5-Phe (right) generated from blood samples. |
The new microfluidic methods reported here have several advantages relative to conventional techniques. An obvious advantage is the potential for automation -- a single technologist using an instrument powered by digital microfluidics could likely do the work of several technologists using manual techniques. Furthermore, for the applications demonstrated here (for newborn screening), the digital microfluidic methods facilitate reduction in reagent use (20 μL vs. 170–450 μL14,32,33) and a reduction in analysis time (∼1 h vs. >3.5 h1,34). Another advantage of the new technique is the potential for direct integration with mass spectrometry (Fig. 5). Nanoflow electrospray ionization (nESI) is a finicky technique requiring operator expertise and vigilance to achieve reproducible results, which is, in part, the reason why all newborn DBS samples in Ontario are mailed to a single facility.10 Indeed, a recent editorial on this subject decried the inherent limitations of systems relying on “conventional electrospray plumbing including capillary tubes etc.”1 In the new method, there is no plumbing or tubes -- an integrated device is simply positioned in front of the mass spectrometer to generate nESI-MS/MS spectra.
Over the past two decades, MS/MS has revolutionized newborn screening, replacing the tests used historically for diagnosing amino acid disorders (i.e., Guthrie's bacterial inhibition35 and fluorometric36 and enzymatic37 tests). This is mainly because MS/MS facilitates the analysis of multiple analytes simultaneously in a single run in comparison to the traditional single-analyte assays, which improves efficiency and reduces costs.1,38 In this work, we present the first method to combine microfluidic sample handling and preparation with MS/MS for analysis of amino acids. We note that there are other metabolic and non-metabolic disorders (e.g., cystic fibrosis, congenital hypothyroidism, biotinidase deficiency, galactosemiaetc.) that are commonly screened in newborn patients that are not routinely detected by MS/MS. Instead, these diseases are evaluated by enzyme-, DNA-, or immunoassay-based tests38,39 and a variety of new technologies39,40 (including digital microfluidics41) are being developed to implement them, as well.
The current work demonstrates proof-of-concept for quantification of amino acids that are often measured for early diagnosis of diseases in newborn screening, but we propose that similar techniques might be useful for a wide range of applications in the future. For clinical analysis,1,2 blood samples could be spotted onto filter paper (Fig. 2b) or onto inexpensive DMF devices42,43 (or removable device coverings27) (Fig. 2a), after which they would be mailed or delivered to a laboratory for analysis. For pharmaceutical applications,3 built-in emitters (Fig. 5) for in-line analysis with mass spectrometry might be appropriate. We propose that the methods described here offer a flexible range of analysis modes that may be useful for the growing number of laboratories that are adopting dried blood spot samples for routine analysis.
Working solutions of all amino acids (AAs) (25, 50, 100 and 500 μM ea.) were prepared in DI water. For derivatization of extracted AAs, a 3 N HCl in n-butanol solution was used. For analysis of AAs in blood samples, the extracting solvent (MeOH) contained 50 μM of the appropriate deuterated AA (d3-Met, d5-Phe or d4-Tyr). For quantitative analysis of AA recovery from blood and for experiments mimicking diseased/healthy infant blood, samples were spiked with 200 μM of the appropriate AA (Met, Phe or Tyr). A series of 3.2 mm dia. punches of dried blood samples containing different concentrations of phenylalanine (see Table 2) were prepared by NSO staff.16
For samples processed by digital microfluidic method 1 (Fig. 2a), calibration plots (Fig. S1 in the Supplementary Information†) were generated by plotting the MS/MS intensity ratio of product ions from the extracted AAs from standards in methanol relative to the those of the internal standards (i.e., Met m/z 104:107, Phe m/z 120:125, and Tyr m/z 136:140) as a function of AA concentration in standard solutions (25–500 μM in DI water). Data points included in the calibration plots represent an average of at least 4 replicate measurements, and the data in each plot were fit with a linear regression. Blood samples were then evaluated (with on-chip derivatization and extraction, and measurement by MS/MS relative to internal standards, as above), and the values were compared to the calibration plots to determine the AA concentrations.
For samples processed by digital microfluidic method 2 (Fig. 2b), quantification was similar to that for method 1, except that standards were formed from adult male blood spiked with Phe to different concentrations (25–900 μM). 3.1 μL aliquots were pipetted onto pre-punched (3.2 mm dia.) discs of Whatman 903 filter paper (Whatman, NJ) and were allowed to dry. Calibration plots (Fig. S2 in the online Supplementary Information†) were formed with the assumption that endogenous blood samples contained 38 μM Phe (the measured concentration in the sample prior to spiking). Newborn blood samples (as DBS filter paper punches) from NSO were then analyzed as above and compared with the calibration plot to determine the Phe concentrations.
Footnotes |
† Electronic supplementary information (ESI) available. See DOI: 10.1039/c1lc20524b |
‡ These authors contributed equally to this work. |
This journal is © The Royal Society of Chemistry 2011 |