Biosensors and Bioelectronics 64 (2015) 260–268

Contents lists available at ScienceDirect

Biosensors and Bioelectronics journal homepage: www.elsevier.com/locate/bios

Robotics-assisted mass spectrometry assay platform enabled by open-source electronics Shih-Hao Chiu a, Pawel L. Urban a,b,n a b

Department of Applied Chemistry, National Chiao Tung University, 1001 University Road, Hsinchu 300, Taiwan Institute of Molecular Science, National Chiao Tung University, 1001 University Road, Hsinchu 300, Taiwan

art ic l e i nf o

a b s t r a c t

Article history: Received 10 June 2014 Received in revised form 13 August 2014 Accepted 26 August 2014 Available online 6 September 2014

Mass spectrometry (MS) is an important analytical technique with numerous applications in clinical analysis, biochemistry, environmental analysis, geology and physics. Its success builds on the ability of MS to determine molecular weights of analytes, and elucidate their structures. However, sample handling prior to MS requires a lot of attention and labor. In this work we were aiming to automate processing samples for MS so that analyses could be conducted without much supervision of experienced analysts. The goal of this study was to develop a robotics and information technology-oriented platform that could control the whole analysis process including sample delivery, reaction-based assay, data acquisition, and interaction with the analyst. The proposed platform incorporates a robotic arm for handling sample vials delivered to the laboratory, and several auxiliary devices which facilitate and secure the analysis process. They include: multi-relay board, infrared sensors, photo-interrupters, gyroscopes, force sensors, fingerprint scanner, barcode scanner, touch screen panel, and internet interface. The control of all the building blocks is achieved through implementation of open-source electronics (Arduino), and enabled by custom-written programs in C language. The advantages of the proposed system include: low cost, simplicity, small size, as well as facile automation of sample delivery and processing without the intervention of the analyst. It is envisaged that this simple robotic system may be the forerunner of automated laboratories dedicated to mass spectrometric analysis of biological samples. & 2014 Elsevier B.V. All rights reserved.

Keywords: Arduino Automation Mass spectrometry Open-source electronics Robotics Sample handling

1. Introduction In the beginning of the 21st century, bioscience is increasingly relying on chemical information provided by modern instrumental platforms. With the advent of personalized medicine, reliable chemical analysis tools for the assessment of patients' health condition become ever more important (Mancinelli et al., 2000). Mass spectrometry (MS) is a universal tool which enables acquiring comprehensive data on chemical composition of biological and clinical samples (de Hoffmann and Stroobant, 2007; Watson and Sparkman, 2007; Gstaiger and Aebersold, 2009; Gross and Roepstorff, 2011; Kandiah and Urban, 2013). Many innovative MS-based approaches have been proposed in the past few years spearheading developments in the emerging areas of bioscience, including systems and synthetic biology. However, practical n Corresponding author at: National Chiao Tung University, Department of Applied Chemistry, 1001 University Road, Hsinchu 300, Taiwan. Fax: þ886 3 5723764. E-mail address: [email protected] (P.L. Urban).

http://dx.doi.org/10.1016/j.bios.2014.08.087 0956-5663/& 2014 Elsevier B.V. All rights reserved.

implementation of these cutting-edge tools in the real-world laboratories is hindered by the requirement for performing multiple manual operations on biochemical and clinical samples prior to the actual analysis (cf. Granvogl et al., 2007; Gundry et al., 2009; Wiśniewski et al., 2009; Hess, 2013). Such procedures are time and labor-consuming, and prone to human error. While a high level of expertise is required to implement modern mass spectrometric assays, novel analytical methods – which require a number of skills and considerable load of experience – cannot readily enter clinical setting. Along the cost of highly qualified labor, the time spent on sample preparation for MS is a major constraint (cf. Börjesson and Torstensson, 2000; Stalikas and Pilidis, 2000; Kudzin et al., 2002). Automation of production and daily life tasks has been in focus since the beginning of the 20th century. The term “robot” was introduced by Čapek in his musical drama released in 1920 (Čapek, 1920). Joseph Engelberger is regarded as the inventor of one of the first industrial robots (Munson, 2010). In 1956, Engelberger founded the Unimation company in the USA, which utilized a patent by George Devol (Pearce, 2011). This patent file

S.-H. Chiu, P.L. Urban / Biosensors and Bioelectronics 64 (2015) 260–268

covers the design of a “general purpose machine” with a large number of applications enabled by cyclic control (Devol, 1961). Robots started to be used by industry in 1959, and they entered car factories in 1961 (Brooks, 2013). Over time, various improvements were made: 2-axis robots were replaced by multiple-axis devices, the weight decreased, and the fluid-driven servo mechanisms were changed to motor-driven servos to make the device operation less noisy and independent of fluid pumps (Gupta, 2007). The introduction of 6-axis robots in 1973 significantly enhanced production flexibility. The field of robotics reached its pinnacle when Richard Hohn invented the first “minicomputer”-controlled industrial robot – “The Tomorrow Tool” (T3). This greatly contributed to miniaturization of the succeeding robotic systems (RobotWorx Website (2014a)). In 1985, the KUKA company (Germany) commercialized a Z-shaped robot, which enabled threedimensional and rotational movements, further augmenting operation flexibility (RobotWorx Website (2014b)). Finally, in 1994, the Motoman company (Japan) improved the electronic control system, which then enabled 22-axis synchronized control of multiple robots. They also implemented touch screen for easy control of the robotic system by operator (Yaskawa Website, 2014). These milestone inventions and improvements spurred rapid growth of the robotic technology. Over the past few decades, the field of robotics has undergone a remarkable transformation. Nowadays robots find applications in various fields – including car manufacture, food and textile industry. Robots decrease the requirement for excessive human labor. The use of robots by industry also limits the exposure of employees to hazardous conditions (e.g. solvent and acid vapors). In the beginning of the 21st century, robotic systems are being adapted to perform highly specialized tasks, including healthcare procedures (Dahl and Boulos, 2013), surveillance (Paola et al., 2010) as well as scientific

261

research – including areas such as genomics, chemical synthesis and combinatorial chemistry (Hardin and Smietana, 1996; Antonenko, 1999; Maccio et al., 2006; Peplow, 2014; Marquardt et al., 2009). Robots expeditiously carry out routine “wet lab” tasks, fulfilling the stringent high-throughput processing requirements. Automated devices are heavily used in analytical chemistry; for example, to facilitate preparation of samples for analysis (Hu et al., 2014). Autosamplers accompany chromatographs to enable unsupervised sample injection, or even to perform simple derivatization reactions and extractions (e.g. Agilent Website, 2014; Dionex – Technical Note 107). Other examples of operations performed by robotic systems include: sample weighing, transfer of samples from test tubes to homogenization vessels, addition of solvents, sample dispersion, filtration of extracts, dilution and injection to a liquid chromatography system (Höller et al., 2003). Much R&D work has also been directed towards construction and improvement of robotic sample delivery systems for nuclear magnetic resonance spectroscopes (Folmer, 2004; Maccio et al., 2006). The popular robotic autosampler systems – which accompany various analytical instruments – have several limitations: their working range is constrained, and they are customized to particular tasks limiting their application flexibility and upgradeability. The ability to operate in three dimensions without major constraints makes robots versatile automated tools for performing multiple tasks. Importantly, the universal robotic systems, designed for bioanalytic work, may readily be reprogrammed to cope with the varying sample handling requirements. Modular design of robotic platforms can provide quick and cost-effective deployment of solutions (Manley et al., 2008), as needed in various chemistry laboratory applications.

Fig. 1. Illustration of the robotics-assisted mass spectrometry assay (RAMSAY) system – described in this report. (A) The system incorporates an inexpensive robotic arm, an ion-trap mass spectrometer, a number of electronic sensors, and an electronic controller (cf. Fig. 2). (B) Positions of the sensors installed in the robotic arm and in its surroundings (cf. Table 1).

262

S.-H. Chiu, P.L. Urban / Biosensors and Bioelectronics 64 (2015) 260–268

To address the growing demand of clinical and research laboratories for high-quality and high-throughput chemical analyses, here we present the development and application of a robotics-assisted mass spectrometry assay (RAMSAY): an operational robotic platform integrated with mass spectrometry that manages mass spectrometric analysis of samples delivered to the laboratory (Fig. 1A). It was envisioned that a robotic platform for MS analysis should be highly autonomous, and its operation should be intuitive. It should be integrated with the teleinformatic network in order to monitor the analysis progress from distant locations (cf. “chemical teleanalysis” (Ciomartan, 1985)). It should ascertain high reliability of analysis (i.e. prevent confusion of samples and any unauthorized use or tempering with the hardware). At the same time, the prototyping and manufacture cost should be as low as possible. To fulfill these requirements, the authors decided that the platform would be built using opensource electronic components. In fact, open-source electronics has recently become very popular among hobbyists and professionals (Monk, 2013), and it has found various applications in chemistry laboratories (Hu et al., 2014; Pearce, 2014; Ting et al., 2014a; Urban, 2014). It enables construction of inexpensive – yet capable – devices for a wide range of laboratory applications. Examples include: capillary electrophoresis (Mai et al., 2013), digital microfluidics (Choi et al., 2013), extraction (Hu et al., 2014; See and Hauser, 2014), and spectrophotometry (McClain, 2014; Urban, 2014). There are numerous open-source electronic platforms; however, the most popular and universal ones belong to the Arduino family. Therefore, Arduino microcontrollers are used in the current work to drive operations of the robotic system, and to communicate with the outside world. Mass spectrometry is nowadays considered to be a mainstream analytical technology providing high sensitivity (Kandiah and Urban, 2013) and remarkable temporal resolution (Chen and Urban, 2013). Therefore, the described system is interfaced with a commercial mass spectrometer to enable sensitive detection of analytes and reaction products. Digital and analog signals from multiple sensors are sent to the Arduino microcontrollers, and used by the custom programs to guide the operations of a single robotic arm. Output signals from the microcontroller trigger various servos guiding mechanical operations of the entire sample handling line. This includes priming and monitoring of a biochemical reaction.

2. Materials and methods 2.1. Materials LC-grade acetic acid (49–51%), ethanol (99.8%), L-glutathione (reduced form), and LC-MS-grade water were purchased from Sigma-Aldrich (St. Louis, MO, USA). Hydrogen peroxide was from Merck (Darmstadt, Germany). 2.2. Robotic arm The robotic arm used for handling samples prior to analysis by MS was from the Robotic Arm Edge kit (OWI, Carson, CA, USA). The main frame of the arm was assembled from the plastic components. The arm is equipped with a gripper that can open and close. It features wrist motion of 120°, an elbow range of 300°, base rotation of 270°, base motion of 180°, vertical reach of 15 in., horizontal reach of 12.6 in., and lifting capacity of 100 g (OwiRobot Website, 2014). The battery compartment was modified in order to power the robot using two mains-compatible power supplies (3 V, 1 A, Toshiba, Tokyo, Japan). The arm was further customized by attaching 2 force resistive sensors (FS1 and FS2), 1 infrared proximity sensor shield (IRS2), 2 accelerometers/gyroscopes

(ACC1 and ACC2), and 3 photointerrupters (PI1, PI2 and PI3; Fig. 1B). These sensors altogether provide vital information on the position of the arm to the electronic microcontroller unit (see Section 2.3), thus enabling precise control of the arm's motion. The sample vials (∼3 mL, dimensions (w, d, h): 2  2  4.3 cm3) – compatible with the gripper of the robotic arm described above – were designed in the Google SketchUp software (2013), and fabricated by means of 3D printer (UP Plus 2; Beijing TierTime Technology, Beijing, China) using acrylonitrile–butadiene–styrene (ABS) as substrate. The vials have square cross-section which makes it easy for the robot to grab them. To further decrease the possibility of dropping the sample, the gripper fingers of the robotic arm were lined with cushions made of double-sided adhesive tape and modeling clay (Sugru; FormFormForm, London, UK) covering the force resistive sensors (FS1 and FS2). 2.3. Electronic control system In order to operate the single-arm robotic system – as a sample handling tool for MS – we constructed a control unit using opensource electronics (Arduino; Fig. 2). All the functions are controlled by scripts written in a C-derived language, and loaded to the microcontrollers from the Arduino's integrated development environment (IDE; Fig. S1). Since the robotic platform performs multiple functions (identifying the user and the sample, communicating with the user via touch screen and website interfaces, guiding the robotic arm, collecting data from multiple sensors), the control device incorporates several Arduino printed circuit boards (PCBs; Fig. 2): (1) the motherboard of the system (Arduino Mega 2560, Torino, Italy); (2) the internet interface (Arduino Uno); (3) the touch screen interface (Arduino Mega 2560); and (4) the user identification sub-system (Arduino Leonardo). The PCB no. 1 controls all the main functions of the robotic setup including movements of the robotic arm. It is directly linked to a 16-relay extension PCB (eBay; Hong Kong, China). Those relays are used to (i) operate 5 linear direct current (DC) motors (M1-M5) incorporated into the robotic arm, (ii) trigger the MS data acquisition, and (iii) operate a syringe pump (Legato 130, KD Scientific, Holliston, MA, USA). The PCB no. 1 also receives digital and analog signals from multiple sensors: 2 round force-sensitive resistors (FS1 and FS2; model 402, diameter of the sensing region: 0.5 in.; Interlink Electronics, Camarillo, CA, USA); 2 accelerometer shields with gyroscope functionality (ACC1 and ACC2; MMA7361L, Freescale Semiconductor, Austin, TX, USA); 3 photointerrupter shields (PI1, PI2 and PI3; MOCH26A, Longge Dianzi, Guangzhou, China); and 3 infrared proximity sensor shields (IRS1, IRS2 and IRS3; KeyesIR, Shenzhen Wonderwing Electronics, Guangdong, China). The PCB no. 1 sends information about the status of analysis (the number of the operation step, e.g. picking up the sample vial, transferring the vial to the ion source, conducting MS analysis, sample disposal, returning to the original position) to an Arduino-compatible Nixie module (QS30-1, Nixie Clock, Shanghai, China) for display. The PCB no. 3 is stacked onto a 5-in. thin film transistor (TFT) touch screen (ITEAD Studio, Shenzhen, China) using an interface shield (eBay; Hong Kong, China). The information about the status of the device is communicated by the PCB no. 3 to the PCBs Nos. 1 and 2 using a digital interface. The PCB no. 3 also receives information from the barcode reader/scanner module equipped with charge-coupled device (CCD) camera and PS/2 interface (MCR12; Champtek, New Taipei City, Taiwan), positioned in the zone adjacent to the sample drop-off area (Fig. 1B). The PCB no. 3 handles information obtained from the fingerprint scanner (ZFM-20; ZhianTec, Hangzhou, China). The data provided by the barcode scanner and the fingerprint scanner identify the sample and the user, respectively. The entire system is powered with three power supplies: 5 V 3 A (Toshiba, Tokyo, Japan) to sustain operation of the low-voltage shields

S.-H. Chiu, P.L. Urban / Biosensors and Bioelectronics 64 (2015) 260–268

263

Fig. 2. Simplified scheme of the electronic control system developed in this work using Arduino microcontrollers to guide operations of the robotic arm coupled with mass spectrometer.

(accelerometer shields, infrared proximity sensor shields, photointerrupter shields, fingerprint scanner, and barcode scanner); 9 V 3 A (Yi Yang, Hsinchu, Taiwan) to power all the microcontrollers (PCBs Nos. 1–4); and 12 V 12.5 A (Cat. no. NES-150-12; Mean Well, New Taipei City, Taiwan) to power the 16-relay board and the Nixie tube shield. In order to enable communication of the control box with the internet, the Arduino Uno PCB (no. 2) was fitted with an Arduino ethernet shield (W5100; eBay, Hong Kong, China). The website interface was constructed in the HyperText Markup Language (HTML) file loaded onto a secure digital (SD) card (8 GB) inserted to the slot in the ethernet shield. The server program in C-derived language was loaded into the Arduino Uno PCB. The program logs onto the local network using a preset dynamic internet protocol (IP) address. This dynamic IP address is locked in the settings of the network router. When the robot's website is accessed from outside the local network, the console browser calls the static IP of the network, and the router directs the query to the dynamic IP of the Arduino Uno server. (Please note that the robot's website is only available when the robot is on-line, and the Arduino Uno PCB is powered.)

the alternating polarity mode in order to detect as many reactants as possible. However, only the negative-ion spectra are presented in the results section of this report. The electric potential on the inlet of the MS was set to þ4500 V (when collecting negative ions). Flow rate of the dry gas (nitrogen) was set to 10 L min  1. Dry gas temperature was set to 200 °C. When using the Venturi pump for ion generation, the pressure of the nebulizing gas (nitrogen) was set to 20 psi (∼138 kPa). The Venturi pump (Santos et al., 2011) was constructed using a stainless steel T-junction (1/8-in., Swagelok, Supelco, Solon, OH, USA). The outlet end of the Venturi pump was fitted with a 15-mm long section of polytetrafluoroethylene (PTFE) tubing (ID 0.5 mm, OD 1.58 mm; Cat. no. 58701, Supelco, Bellefonte, PA, USA). A 20-cm section of polyimide-coated fused silica capillary (ID 0.150 mm, OD 0.375 mm; 1010-32132, GL Science, Tokyo, Japan) was mounted in the T-junction and slid through the lumen of the PTFE tubing. The inlet of this sampling capillary was also shielded with a 1.2-cm section of PTFE tube (ID 0.3 mm, OD 1.58 mm; Cat. no. 58702, Supelco) to prevent damage of the silica capillary when it hits the wall of the sample vial (during the alignment conducted by the robotic arm). The Venturi pump system was regularly cleaned by passing concentrated ethanol and water through the fused silica capillary.

2.4. Mass spectrometry 3. Results and discussion We utilized the amaZon speed ion-trap mass spectrometer from Bruker Daltonics (Bremen, Germany) as the detector for the RAMSAY system. The controller device (described in Section 2.3) triggers data acquisition by the mass spectrometer via its “auxiliary” interface. The data acquisition was accomplished using the TrapControl software (ver. 7.1; Bruker Daltonics). The mass spectrometer was operated in

3.1. Design of the robotic analysis system In order to automate analysis of liquid samples by mass spectrometry, here we propose a facile robotic arm-aided sample vial delivery system (Fig. 1A). The setup takes advantage of a

264

S.-H. Chiu, P.L. Urban / Biosensors and Bioelectronics 64 (2015) 260–268

Venturi pump (Santos et al., 2011) which enables ionization of analytes present in liquid samples. Importantly, this type of ion source does not require syringe pump because the sample is automatically aspirated as a result of pressure difference between the inlet and the outlet of the sample line. Therefore, the Venturi pump ion source is particularly suitable for the development of automated sample delivery systems (Hu et al., 2014; Ting and Urban, 2014b). In addition, since no syringes are used, the carryover effect can be minimized. The most challenging part in the development of the proposed platform was to assure robust alignment of the robot with the inlet of the sampling capillary. This was related to the fact that we used an inexpensive arm with limited precision (see Section 3.3). This drawback was compensated to some extent by implementing several electronic sensors which sentinel the robotic arm providing feedback information on its momentary orientation in space. Once this obstacle was overcome, the robot could repeatedly transfer sample vials to the inlet of the Venturi pump, thus enabling mass spectrometric analysis. All the movements of the arm were guided by the scripts loaded to the Arduino microcontrollers (see Section 2.3).

relay) to the auxiliary interface of the mass spectrometer, thus initiating data acquisition. After 120 s, the data acquisition is stopped (according to the setting of the mass spectrometer), and the spectra are saved for post-sequence evaluation. If the operation of RAMSAY is stopped by pressing the emergency button or selecting the option “Offline”, the initial position of the robotic arm can normally be restored by selecting the option “Return” (Fig. S2). The operator may also choose to take over manual control of the robotic arm, should the automated positioning of the arm fail. Supplementary material related to this article can be found online at http://dx.doi.org/10.1016/j.bios.2014.08.087. In some experiments (Section 3.4), the syringe pump was triggered to add an aliquot of a reagent to the sample in order to prime a reaction. The robotic arm facilitated mixing of the sample with the reagent by gently moving the sample vial (cf. Movie S1). The incubation of the sample with the reagent was followed by another MS analysis. This process was repeated in order to collect data on the progress of the reaction. Finally, the sample vial was discarded, and the arm returned to the initial upright position. The system was ready for processing another sample. 3.3. Repeatability of autonomous operation

3.2. Operation algorithm The workflow of the analytical procedure guided by the Arduino microcontrollers is illustrated in Fig. S1. The touch screen interface (Fig. S2) of the proposed RAMSAY platform displays three options (“Go Online”, “Go Offline”, and “Return”) which can be selected by the operator. At the beginning of a typical analysis, the operator selects the option “Go Online”. Subsequently, the system requests performing a scan of the operator's fingerprint (Fig. S1). As soon as he/she presses their indicator finger against the scan zone of the fingerprint scanner, the system conducts a search of the recorded fingerprint. If this individual is authorized to perform analysis, their name appears on the screen, and the system proceeds to the next step. The operator places the sample vial in the designated sample drop-off zone – in front of the robotic arm (cf. Fig. 1B). An infrared sensor (IRS1) registers the presence of the sample (Table 1). The barcode scanner attempts reading the barcode attached to the sample vial. When the sample is recognized, the sample number is displayed on the user interface (touch screen; Fig. S2), and posted on the RAMSAY's website (Fig. S3). In addition, the type of the sample is identified based on this unique number, and also displayed on the user interface. Right after that, the robotic arm starts processing the sample according to the sequence illustrated in Fig. 3. In ∼30 s, the sample is delivered to the inlet of the Venturi pump (cf. Movie S1). A small aliquot of sample is aspirated by the Venturi pump, and sprayed in front of the MS orifice. The microcontroller sends a trigger signal (via Table 1 List of the main operation steps in a single RAMSAY sequence (example). No. Step 1 2 3 4 5 6 7 8

Please scan the sample Lowering arm

Sensors involved

Actuators involved

Infrared sensor (IRS1), barcode None scanner Infrared sensor (IRS2) Motor 2, Motor 3, Motor 4 Grabbing sample Force sensors (FS1, FS2) Motor 1 Rotating arm Photointerrupter (PI2) Motor 5 Aligning with Infrared sensor (IRS3) Motor 3 ion source MS analysis None None Sample disposal Force sensors (FS1, FS2) Motor 1, Motor 5 Returning to Photointerrupter (PI3, PI1), Motor1, Motor 2, initial position accelerometers (ACC1, ACC2) Motor 3, Motor 4, Motor 5

It was of paramount importance to make sure that the repeatability of the proposed robotic system was satisfactory for realworld applications. This is because the robotic arm used in this study was much smaller, cheaper (o65 USD) and less robust than regular industrial robots (typically, 4 50,000 USD). In one set of experiments, we verified the precision of positioning the robotic arm in front of the MS ion source. A camera was installed above the robotic arm, recording its movements. Snapshots of the rest position of the arm were taken. The drift of the robotic arm (i.e. run-to-run position repeatability) were determined. The standard deviations of the X and Y positions of the arm were 0.61 and 2.1 mm, respectively (n ¼5). This is worse than the repeatability of expensive industry-standard robots but sufficient for the alignment of the sample vial with the inlet of the Venturi pump aspirating liquid sample for analysis. Replacing plastic parts of the robotic arm (in particular the gear wheels) might possibly reduce micro-vibrations, which would certainly improve the sample handling precision. Implementing an all-metal robotic arm will be justified if the considered analytical application requires a high level of positioning precision. This was not the case in this proof-of-concept study which used relatively large sample vials while low capital cost was a priority. In another experiment – designed to test operation repeatability – the robotic system was already coupled to the mass spectrometer, and analysis of a sample of 10  5 M glutathione (reduced form) in pure water was conducted. In the negative-ion mode, the glutathione peaks were observed at the m/z 306.0 (monomer, [GSH–H]  ) and 613.2 (dimer, [2GSH–H]  ) (Fig. 4). The analyte – aspirated by the Venturi pump – arrived at the ionization zone ∼30 s after the RAMSAY control unit initiated data acquisition. Therefore, the mass spectra obtained after averaging the signal over the period 1.0–2.0 min are considered representative. The PeakFit software (version 4.12; SeaSolve Software, San Jose, CA, USA) was used to fit the highest monoisotopic peaks with the Haarhoff – Van der Linde function curves, and calculate their integrals. The relative standard deviation of the [GSH–H]  peak area was 12% (n ¼ 3) which is considered to be a satisfactory result – considering that the ion-trap mass spectrometer was used as a standalone detector – without chromatograph. In the following tests, we also determined precision of the mass spectrometric detection system equipped with Venturi pump by re-analyzing the same sample (10  5 M glutathione). This time, the ion trap mass spectrometer was switched to the offline mode

S.-H. Chiu, P.L. Urban / Biosensors and Bioelectronics 64 (2015) 260–268

265

Fig. 3. Sequence of positions of the robotic arm while handling samples for analysis by mass spectrometer.

between analyses in order to simulate discontinuous operation of the detection system. The relative standard deviation of the signal was 11% (n¼ 6; based on average extracted ion current from 1-min monitoring). It is suggested that this analytical precision could further be improved by using internal standards. Although the mass spectrometric data obtained here, using the ion trap mass

spectrometer, have some quantitative features, in order to perform truly quantitative analyses, it is advised to implement a triple quadrupole mass analyzer instead. Interestingly, the results of experiments designed to test the mass spectrometric method indicate that varying the voltage of the ion transfer capillary (3000–5500 V) and the temperature of the dry gas (175–215 °C)

266

A

S.-H. Chiu, P.L. Urban / Biosensors and Bioelectronics 64 (2015) 260–268

A

[GSH-H] -

[GSH-H] -

[GSSG-H] -

B

5

10 a.u.

B

10 5 a.u. 0 min

B

7.5 min 28.5 min 200

C

400

500

600

700

800

300

400

500

m/ z

600

700

800

Fig. 4. Three consecutive analyses (A–C) of an artificial sample (900 μL 10  5 M glutathione dissolved in pure water) conducted by the RAMSAY system. Negativeion mass spectra (1.0–2.0 min).

had little influence on the signal intensity (relative standard deviations: 15% and 7%, respectively; n ¼6, based on average extracted ion current from 1-min monitoring, glutathione concentration: 10  5 M). Overall, the above data suggest that the robustness of the MS detection method is sufficient for some of the anticipated applications. 3.4. Reaction monitoring Unlike another robotics-assisted method for sample delivery to time-of-flight mass spectrometer equipped with direct plasma chemical ionization (Bennett et al., 2014), the current system uses an inexpensive robotic arm to feed liquid samples to the ion source of ion trap mass spectrometer fitted with Venturi pump sprayer. Banking on the successful operation of this platform, we intended to further demonstrate its capabilities in the monitoring of a liquid-phase reaction – oxidation of glutathione in the presence of hydrogen peroxide (Deutsch et al., 1999; Kulp et al., 2006; Albrecht et al., 2011): 2GSH þH2O2-GSSG þ2H2O

(1)

The advantage of this reaction as a model is that it is relatively fast while the reactants (GSH and GSSG) can readily be monitored by atmospheric pressure ionization mass spectrometry (e.g. using Venturi pump as the sample emitter). We have chosen such concentrations of the reactants that the reaction reached its completion within ∼30 min. Initially, the sample vial contained 900 μL of 10  5 M glutathione dissolved in pure water. The 0.5 M hydrogen peroxide solution was loaded into 1-mL plastic syringe mounted in the picoliter-volume syringe pump. The program loaded to the Arduino Mega microcontroller (Fig. 2, PCB no. 1) was set to add 100 μL of the hydrogen peroxide solution to start the reaction. After adding this reagent, the arm shook the vial, and promptly aligned it with the inlet of the Venturi pump sample capillary. The experiment was conducted at room temperature (∼25 °C; stabilized by an air conditioning system). A small volume

Scaled relative abundance

m/z

B 200

300

1.2 1.0 0.8 0.6 0.4 0.2 0.0

0

5

10 15 20 25 30 35 40 45 50

Time / min Fig. 5. Kinetic study of the oxidation of glutathione by hydrogen peroxide. In this experiment, RAMSAY carried out temporal monitoring of the redox process by dosing hydrogen peroxide reagent to the sample of glutathione, mechanically mixing the reaction mixture, and subjecting it to mass spectrometric analysis at different time points. (A) Negative-ion mass spectra – collected at different time points – revealing the progress of the oxidation of GSH to GSSG. (Please note the presence of a GSH dimer peak, close to the GSSG peak at 0 min.) (B) Reaction progress curves revealing build-up of GSSG (●) and depletion of GSH (◯).

of the sample was aspirated by the Venturi pump ion source, and nebulized in front of the orifice of the ion-trap mass spectrometer. The system recorded the first spectrum – corresponding to the initial point of the reaction. Subsequently, the robotic arm continued shaking the sample vial, and performed further measurements. The negative-ion mass spectra were collected, and treated after the completion of the analysis. The intensities of the reduced and the oxidized forms of glutathione were determined at the m/z 306.1 and 611.3 u e  1 (Fig. 5A), corresponding to [GSH–H]  and [GSSG–H]  species, respectively (cf. Deutsch et al., 1999). By extracting the peak areas from the spectra collected at different time points of the monitoring routine, it was possible to create kinetic plots showing the progress of the reaction (Fig. 5B) – representing the evolution of the [GSH–H]  /([GSH–H]  þ[GSSG– H]  ) (◯) and [GSSG–H]  /([GSH–H]  þ[GSSG–H]  ) (●) ratios over time (after scaling to the range 0.0–1.0 in order to mitigate the possible influence of contaminant/noise signals). Since the concentration of H2O2 (50  10  3 M) was much higher than the initial concentration of GSH (9  10  6 M), the data points were fitted

S.-H. Chiu, P.L. Urban / Biosensors and Bioelectronics 64 (2015) 260–268

with the pseudo-first order kinetic equations (cf. Kulp et al., 2006): −k′t

GSH = GSH0 e

(2)

GSSG = GSSGmax (1 − e−k′t )

(3)

The kinetic rate constants (k′), obtained from the Eqs. (2) and (3), were 0.1227 0.016 and 0.110 70.017 min  1, respectively. The SPSS software (ver. 19; IBM, Armonk, NY, USA) was used to fit the kinetic functions to experimental data. The obtained values are lower than those obtained in another study – using electrophoretically mediated microanalysis (Kulp et al., 2006) – which can be explained with the fact that different conditions (concentrations of reactants, type and pH of reaction buffer, way of mixing) were used. Another source of bias might be related to small differences in the ionizability of GSH and GSSG molecules. Overall, this experiment shows that the proposed robotic system is capable of executing multiple operations: transferring the sample, dosing a specified amount of a reagent, mixing the reagents, and monitoring the progress of the reaction involving simple biomolecules (GSH, GSSG) by MS over time. While a model reaction (oxidation of glutathione) was used in this proof-of-concept study, it is expected that other reactions can be implemented within the RAMSAY format. It would be appealing to apply this system in the analysis of enzymatic reaction mixtures (for example, proteins digested with proteases), which could facilitate performing routine biochemical assays. Another area of potential applications of RAMSAY relates to the analysis of hazardous (e.g. infectious) samples by mass spectrometry.

4. Conclusions Despite the fact that the idea of implementing robotics in analytical chemistry is at least three-decades old – robotic systems have so far entered analytical laboratories in a reduced form – mainly as autosamplers. Contrary to the applications on Earth, robotic systems for sample handling, combined with mass spectrometers, have already reached other planets. In the 1970s, the Viking probes delivered mass spectrometers with robotic sample preparation stages to Mars (Johnson et al., 2011). They were capable of conducting on-site analysis and sending the data back to Earth. Now is the time to implement a similar “self-sustaining” technology in analytical laboratories equipped with mass spectrometers, so as to facilitate obtaining information on the chemical composition of biomedical samples. It should be borne in mind that robotic arm-based systems can easily be adapted (modified, reprogrammed) for performing different tasks related to sample treatment. This is in contrast with some of the popular 3-axis rail-type sample handling devices. Indeed, the proposed automated analytical system is characterized with some flexibility. Apart from delivering sample for analysis, it can carry out kinetic monitoring of a reaction (as exemplified with oxidation of glutathione), which includes metering of the reagent, mixing, and mass spectrometric analysis of the reaction mixture at pre-defined time points. It also interacts with the user via touch screen panel and has built-in safety features (identification of the person authorized to deliver samples by scanning fingerprint). The network interface enables monitoring of the system status (online/ offline, identification number of the analyzed sample) and environmental conditions in the laboratory. The construction of this appliance could be achieved using modest resources and widely available materials. The system takes advantage of open-source electronics (Arduino) to obtain data from sensors and to trigger actuators. Because of the implementation of inexpensive electronic modules, the cost of the developed device is much lower than

267

the cost of the mass spectrometer. Thus, it is appealing to further combine it with an economical miniature mass spectrometer – in fact, such instruments are currently entering the scientific instrumentation market (Smith et al., 2010; Perkel, 2014; Hendricks et al., 2014). It should also be noted that the proposed inexpensive robotic accessory for MS is not meant to be a substitute for highend robotic sample preparation systems characterized with better precision and range of operation. In future work it would be appealing to equip RAMSAY with various sample preparation functions (e.g. weighing, filtration, extraction, chromatographic separation), which are incorporated to some of the existing commercial platforms. It is envisioned that – in the near future – robotic analytical systems coupled with mass spectrometers may enable fast analysis and screening of various samples, including blood, urine, sputum, sweat, biopsy samples as well as small metazoans and microorganisms. Tight integration of such platforms with teleinformatic networks will be quintessential to their projected functionalities within the “internet-of-things” domain. Down this path, we are currently working on a new version of RAMSAY, which could enable closer interaction between the human and the machine.

Acknowledgments We thank the Ministry of Science and Technology of Taiwan (formerly, National Science Council; Grant no. NSC 102-2113-M009-004-MY2), the National Chiao Tung University as well as the “Aiming for the Top University Program” of the National Chiao Tung University and the Ministry of Education, Taiwan, for the financial support of this work.

Appendix A. Supplementary information Supplementary data associated with this article can be found in the online version at http://dx.doi.org/10.1016/j.bios.2014.08.087.

References Agilent Website, 2014. 7696A Sample Prep Workbench, 〈http://www.chem.agilent. com/en-US/products-services/Instruments-Systems/Gas-Chromatography/ 7696A-Sample-Prep-Workbench/Pages/default.aspx〉 (viewed on 28.03.14). Albrecht, S.C., Barata, A.G., Großhans, J., Teleman, A.A., Dick, T.P., 2011. Cell Metab. 14, 819–829. Antonenko, V.V., 1999. In: Miertus, S., Fassina, G. (Eds.), Combinatorial Chemistry and Technology: Principles, Methods, and Applications. Marcel Dekker, New York. Bennett, R.V., Morzan, E.M., Huckaby, J.O., Monge, M.E., Christensen, H.I., Fernández, F.M., 2014. Analyst 139, 2658–2662. Börjesson, E., Torstensson, L., 2000. J. Chromatogr. A 886, 207–216. Brooks, R., 2013. Futurist 47, 24. Chen, Y.-C., Urban, P.L., 2013. Trends Anal. Chem. 44, 106–120. Choi, K., Ng, A.H.C., Fobel, R., Chang-Yen, D.A., Yarnell, L.E., Pearson, E.L., Oleksak, C. M., Fischer, A.T., Luoma, R.P., Robinson, J.M., Audet, J., Wheeler, A.R., 2013. Anal. Chem. 85, 9638–9646. Ciomartan, D., 1985. Structure elements of the process of teleanalysis of atmospheric pollutants. 〈http://eurekamag.com/research/006/495/structure-elementsprocess-teleanalysis-atmospheric-pollutants.php〉 (viewed 06.06.14). Čapek, K., 1920. Rossum's Universal Robots. Aventinum, Praha. Dahl, T.S., Boulos, M.N.K., 2013. Robotics 3, 1–21. de Hoffmann, E., Stroobant, V., 2007. Mass Spectrometry: Principles and Applications. John Wiley & Sons, Chichester. Deutsch, J.C., Santhosh-Kumar, C.R., Kolhouse, J.F., 1999. J. Chromatogr. A 862, 161–168. Devol, G.C., 1961. Programmed article transfer. Patent No. US2988237. Dionex – Technical Note 107. 〈http://www.dionex.com/en-us/webdocs/110823-TN107HPLC-AminoAcids-In-NeedleDeriv-13Jun2011-LPN2849.pdf〉 (viewed 25.04.14). Folmer R., 2004. Tecan J. 2/2004. p. 6–8. Granvogl, B., Plöscher, M., Eichacker, L.A., 2007. Anal. Bioanal. Chem. 389, 991–1002.

268

S.-H. Chiu, P.L. Urban / Biosensors and Bioelectronics 64 (2015) 260–268

Gross, J.H., Roepstorff, P., 2011. Mass Spectrometry: A Textbook. Springer, Berlin. Gstaiger, M., Aebersold, R., 2009. Nat. Rev. Genet. 10, 617–627. Gundry, R.L., White, M.Y., Murray, C.I., Kane, L.A., Fu, Q., Stanley, B.A., Eyk, J.E.V., 2009. Curr. Protoc. Mol. Biol. 88, 10.25.1–10.25.23. Gupta, A.K., 2007. Industrial Automation and Robotics. Laxmi Publications, New Delhi. Hardin, J.H., Smietana, F.R., 1996. Mol. Divers. 1, 270–274. Hendricks, P.I., Dalgleish, J.K., Shelley, J.T., Kirleis, M.A., McNicholas, M.T., Li, L., Chen, T.C., Chen, C.H., Duncan, J.S., Boudreau, F., Noll, R.J., Denton, J.P., Roach, T.A., Ouyang, Z., Cooks, R.G., 2014. Anal. Chem. 86, 2900–2908. Hess, S., 2013. Curr. Trends Mass Spectrom. 31, 12–17. Höller, U., Brodhag, C., Knöbel, A., Hofmann, P., Spitzer, V., 2003. J. Pharm. Biomed. Anal. 31, 151–158. Hu, J.B., Chen, S.Y., Wu, J.T., Chen, Y.C., Urban, P.L., 2014. RSC Adv. 4, 10693–10701. Johnson, P.V., Hodyss, R., Tang, K., Brinckerhoff, W.B., Smith, R.D., 2011. Planet. Space Sci. 59, 387–393. Kandiah, M., Urban, P.L., 2013. Chem. Soc. Rev. 42, 5299–5322. Kudzin, Z.H., Gralak, D.K., Drabowicz, J., Łuczak, J., 2002. J. Chromatogr. A 947, 129–141. Kulp, M., Urban, P.L., Kaljurand, M., Bergström, E.T., Goodall, D.M, 2006. Anal. Chim. Acta 570, 1–7. Maccio, M.A., Bell, D.H., Davolos, D., Princeton, W., 2006. J. Lab. Autom. 11, 387–398. Mai, T.D., Pham, T.T.T., Pham, H.V., Śaiz, J., Ruiz, C.G., Hauser, P.C., 2013. Anal. Chem. 85, 2333–2339. Mancinelli, L., Cronin, M., Sadée, W., 2000. Am. Assoc. Pharm. Sci. 2, 29–41. Manley, J.D., Smith, T.J., Holden, J., Edwards, R., Liptrot, G., 2008. J. Lab. Autom. 13, 13–23. Marquardt, P., Morelli, G., Lindner, M., Max Planck Institute for Molecular Genetics – brochure, 2009, Berlin, Germany. 〈http://www.molgen.mpg.de/168318/mpim g_image_screen_en.pdf〉 (viewed 25.04.14). McClain, R., 2014. J. Chem. Educ. 91, 747–750. Monk, S., 2013. Hacking Electronics: An Illustrated DIY Guide for Makers and Hobbyists. McGraw-Hill, New York.

Munson, G.E., 2010. Robot Mag. (45) /http://www.botmag.com/the-rise-andfall-of-unimation-inc-story-of-robotics-innovation-triumph-that-changed-theworld/S (viewed 08/09/14). OwiRobot Website, 2014. 〈http://www.owirobot.com/robotic-arm-edge-1〉 (viewed 28.03.14). Paola, D.D., Milella, A., Cicirelli, G., Distante, A., 2010. Int. J. Adv. Robot. Syst. 7, 19–26. Pearce, J., 2011. The New York Times. /http://www.nytimes.com/2011/08/16/business/ george-devol-developer-of-robot-arm-dies-at-99.htmlS, (viewed 08/09/14). Pearce, J.M., 2014. Open-Source Lab: How to Build Your Own Hardware and Reduce Research Costs. Elsevier, Amsterdam. Peplow, M., 2014. Nature 512, 20–22. Perkel, J.M., 2014. Science 343, 928–930. RobotWorx Website, 2014a. 〈http://www.robots.com/education/research-history〉 (viewed on 07.06.14). RobotWorx Website (2014b), 〈http://www.used-robots.com/education/kuka-robothistory〉 (viewed 07.06.14). Santos, V.G., Regiani, T., Dias, F.F.G., Romão, W., Jara, J.L.P., Klitzke, C.F., Coelho, F., Eberlin, M.N., 2011. Anal. Chem. 83, 1375–1380. See, H.H., Hauser, P.C., 2014. Anal. Chem. 86, 8665–8670. http://dx.doi.org//10. 1021/ac5015589S. Stalikas, C.D., Pilidis, G.A., 2000. J. Chromatogr. A 872, 215–225. Smith, J.N., Keil, A., Likens, J., Noll, R.J., Cooks, R.G., 2010. Analyst 135, 994–1003. Ting, H., Hu, J.B., Hsieh, K.T., Urban, P.L., 2014a. Anal. Methods 6, 4652–4660. Ting, H., Urban, P.L., 2014b. RSC Adv. 4, 2103–2108. Urban, P.L., 2014. J. Chem. Educ. 91, 751–752. Watson, J.T., Sparkman, O.D., 2007. Introduction to Mass Spectrometry: Instrumentation, Applications, and Strategies for Data Interpretation. John Wiley & Sons, Chichester. Wiśniewski, J.R., Zougman, A., Nagaraj, N., Mann, M., 2009. Nat. Methods 6, 359–362. Yaskawa Website, 2014. 〈http://www.motoman.co.uk/en/company/about-yaskawa/ our-history〉 (viewed 07.06.14).