Food Additives & Contaminants

ISSN: 0265-203X (Print) (Online) Journal homepage:

Robotic analysis of aflatoxin M1 in milk L. A. Gifford , C. Wright & J. Gilbert To cite this article: L. A. Gifford , C. Wright & J. Gilbert (1990) Robotic analysis of aflatoxin M1 in milk, Food Additives & Contaminants, 7:6, 829-836, DOI: 10.1080/02652039009373945 To link to this article:

Published online: 10 Jan 2009.

Submit your article to this journal

Article views: 10

View related articles

Citing articles: 1 View citing articles

Full Terms & Conditions of access and use can be found at Download by: [University of Florida]

Date: 14 November 2015, At: 23:21

FOOD ADDITIVES AND CONTAMINANTS, 1990, VOL. 7, NO. 6, 8 2 9 - 8 3 6

Robotic analysis of aflatoxin M1 in milk L. A. GIFFORD†, C. WRIGHT† and J. GILBERT‡ †Department of Pharmacy, University of Manchester, Manchester M13 9PL, UK ‡Ministry of Agriculture, Fisheries and Food, Food Science Laboratory, Colney Lane, Norwich NR4 7UQ, UK

Downloaded by [University of Florida] at 23:21 14 November 2015

(Received 7 March 1990; accepted 19 June 1990) An automated aflatoxin M1 assay system capable of performing multiple unattended extractions and chromatographic analyses has been developed. A six-axis anthropomorphic laboratory robot and a flexible computer system are combined to operate a sample turntable, a multisolvent dispensing facility, a solid-phase extraction station, a vacuum manifold, an automatic HPLC sample preconcentration unit, an HPLC, a fluorescence detector and a computing integrator. The system is capable of handling bulk milk samples and can determine aflatoxins at the sub-µg/kg level with an accuracy and precision comparable to those of the manual methods of analysis. Time of analysis is reduced. The system can be run in an unattended mode of operation. Keywords: robotic analysis, aflatoxin M1, milk, automation


Aflatoxin Mi, a hydroxylated metabolite of aflatoxin Bi, is excreted into milk at between 0-3% and 1% of the level of Bi present in the feedstuff (Heathcote and Hibbert 1978). Although the UK in common with other EEC countries has regulations concerning levels of aflatoxin Bi in animal feeds, which are intended to minimize aflatoxin contamination in the milk, there is nevertheless a need for surveillance for aflatoxin Mi in the UK milk supply, as well as monitoring of imported dairy products. Previous surveys have been reported (Gilbert et al. 1984) using conventional analytical approaches, but these are time-consuming for large sample numbers. A need therefore exists for automated analytical procedures to reduce manpower costs and make future monitoring more effective through analysis of larger numbers of samples. Some of the original analytical procedures for determining aflatoxin Mi in milk involved extraction with large volumes of solvents such as chloroform (Stubblefield 1979). A critical assessment has been made of six more recently published extraction and clean-up methods, with comparisons based on cleanliness of the final extract, analysis time, recovery and relative cost of the procedure (Shepherd et al. 1986). The most satisfactory procedure, according to these criteria, was that of Takeda (1984), which employed a disposable pre-packed Cis cartridge onto which a diluted milk sample was applied. Interfering substances were washed from the cartridge with basic 10% acetonitrile and acidic 10% acetonitrile. The aflatoxin Mi was finally eluted with acidic 30% acetonitrile and reduced in volume under vacuum prior to HPLC. Of the methods evaluated, this was found to be the most amenable to automation. 0265-203X/90 $3.00 © 1990 Taylor & Francis Ltd.


L. A. Gifford et al.

The aim of the work reported in this paper was to develop a flexible automatic analysis system capable of handling sample types frequently encountered in trace contaminant analysis of foods. Typical unit operations involved in sample extraction and clean-up were thought to be exemplified by the analysis of aflatoxin Mi in milk. In developing the system, problems of transfer of large sample volumes were addressed and sample concentration prior to analysis had to be resolved. This paper presents a versatile sample preparation system capable of interfacing with conventional laboratory peripherals and its validation in the analysis of aflatoxin Mi in milk.

Downloaded by [University of Florida] at 23:21 14 November 2015


System layout Figure 1 shows a plan view of the robotic system, which is composed of four basic elements: (1) a six-axis laboratory robot, (2) an IBM PC-XT microcomputer, (3) a EuroBeeb microprocessor-based real-time controller and (4) a sample preparation station consisting of sample turntable, multisolvent dispenser and automatic clean-up station. These four components (including the interface for the real-time controller) cost approximately £19 000 at 1990 prices. System components not directly accessed by the robot include an automatic sample injector, precolumn and HPLC with fluorescence detector. The IBM PC acts as the overall system controller with bidirectional communication to both the robot and real-time computers. Robot motion commands are generated by the IBM PC, as are operational commands to the real-time controller for the peripherals. The real-time system controls the sample carousel, six syringe pumps, a variable-position dispensing head, an electric clamp, a vacuum manifold and the HPLC autoinjector/precolumn loader. SYRINGE









a' TUBE RACK U 1 11 1 1 1

•I 111 11 I 1 M 1








Figure 1. Top view of the robotic analysis system.

Microcomputer and real-time laboratory interface The IBM PC (IBM Corporation, Boca Raton, Florida, USA) is equipped with two asynchronous communications channels: coml is connected to the robot microprocessor, com2 to the laboratory interface. The PC delegates control of individual tasks to either processor as appropriate, from a Microsoft BASIC program. Real-

Downloaded by [University of Florida] at 23:21 14 November 2015

Robotic analysis of aflatoxin


time programming is achieved using using a SPIDER package (Paul Fray Ltd, Cambridge, UK) which is very similar to extended BASIC and thus easy to use for this application. Up to 32 channels of opto-isolated I/O may be driven with the CUBE Delegate Driver card (Control Universal, UK) which may be defined in software either as inputs or outputs, but not both simultaneously. Eight software clocks provide real-time event scheduling. Connections to laboratory equipment are achieved through the use of two CUBE Delegate Industrial Interface cards (Control Universal). Each I/O channel has a status LED and two 4 mm sockets. A channel defined as an input recognizes a voltage between 5 V and 30 V as a valid ON state; an OFF state is defined as 0 V, i.e. open circuit. A channel defined as an output can drive a current sourcing load at 5 V to 30 V and switch up to 2 A. Catching diodes permit inductive loads to be driven. All output channels have a common ground. In order to switch mains-powered devices, the output is routed to a through-zero switching relay which activates a mains outlet. Analogue to digital conversion and vice-versa are performed by the CUBAN-12 interface (Control Universal) which provides up to eight multiplexed ADC input channels in the range 0-10 V unipolar and ± 5 V or ± 10 V bipolar. The CUBAN-12 also provides four latched output lines each having an independent adjustable amplifier which can give a maximum output of 10 V and source up to 10 mA.





V Figure 2. Automated solid-phase extraction system.

Downloaded by [University of Florida] at 23:21 14 November 2015


L. A. Gifford et al.

Robot The laboratory robot was designed and built at the University of Manchester (Wynne and Gifford 1984). It is a hybrid design utilizing hydraulic actuators for the lower load-bearing joints (base, shoulder, elbow and wrist) and stepper motors for finer positioning (wrist rotation and gripper actuation). The onboard computer employs an Intel 8088 microprocessor operating at 5 MHz and communication with the IBM PC is by standard RS232 serial interface. The robot is taught required positions in a sequence of operations driven by the IBM. Positional information is stored in RAM on board the robot, and may be downloaded to disc on the IBM for later use. Teaching is in two tiers: first the robot is driven to required positions and these are given three-character names; second the robot is driven to these positions in the sequence demanded to perform a task. These processes are menu-driven, permitting an unfamiliar operator to program the robot after only a short period of 'hands-on' experience. Two types of data are therefore stored: positions—a set of names and their requisite joint coordinates; and sequences—strings of position names in the right order. Only positional information is stored in robot RAM. Sequences are stored and called by the IBM. The IBM passes positional calls from a sequence individually to the robot, which drives to that position and acknowledges it. The IBM then transmits the next position in the sequence, and so on. On completion of a sequence the IBM then proceeds to call the next real-time operation. Real-time commands in SPIDER are preprogrammed into the EuroBeeb using a terminal emulator (Paul Fray) which permits keying in or downloading from disc of software. This is stored in battery backed-up RAM. Control commands are passed to the EuroBeeb via its serial interface. Peripheral devices The rotary turntable (Hook and Tucker Instruments Ltd, New Addington, UK) was modified with a set of posts and sidearms to support a BondElut (Analytichem International Inc., Harbour City, C A) cartridge fitted with a reservoir and 50 mm stainless-steel needle next to each flask. Six syringe pumps, 1 x 20 ml, 4x10 ml, 1 x 5 ml (Hook and Tucker Ltd) were employed to dispense solvents used in the clean-up procedure. Each pump was modified by installing a 24 V operated relay to permit actuation by the laboratory interface. All solvent delivery lines from the syringe pumps were brought to focus at a single dispensing head, which was fabricated from Perspex and drilled to accommodate the delivery lines within a small area. The head was mounted on the end of a 150 mm long silver steel bar (3 mm o.d.) driven through a horizontal arc by a small precision DC servo motor (R.S. Components, Corby, UK). Positional control of the servo was achieved using reference voltages supplied by the DAC of the EuroBeeb. In order for the robot arm to perform most efficiently, it was required to clamp the cartridge above the clean-up stage. This was done using a standard retort clamp (Orme Scientific, Manchester, UK) from which the twin screw was removed and the pivot pin was replaced by a shaft driven by another servomotor. Jaw opening and closing was controlled by a second DAC line. Two vacuum stations were constructed from standard laboratory glassware. The first (clean-up) station consisted of a socket/cone adapter with T connection (Quickfit MF17/1, Fisons Scientific Equipment Ltd, Loughborough, UK) and was capped with a rubber septum and aluminium retaining ring. A large heavy-wall conical flask completed the station

Robotic analysis of aflatoxin









-/ \

Downloaded by [University of Florida] at 23:21 14 November 2015


-z í

-/ \



XČL,VACUUM Figure 3. Vacuum manifold configuration.

and was used as effluent collector. The second (elution) station consisted of another septum-capped T-piece but was left open beneath to receive 15 ml tapered tubes for eluate collection. Solvent could be drawn through the cartridge by pushing the needle through the septum to connect to vacuum. A vacuum-switching manifold was fabricated from four 24 V two-way solenoids (Burkert Controlmatic Ltd, Stroud, UK) operated in pairs and configured as in figure 3. This was used to apply vacuum at one site whilst simultaneously releasing vacuum to atmospheric pressure at the other. Sample preparation Aliquots (50 ml) of the milk samples to be analysed or standards spiked with aflatoxin Mi (Sigma Chemical Co, Poole, UK) are diluted with 50 ml distilled water in 250 ml Erlenmeyer flasks and loaded onto the sample carousel. Full-cream milks are first 'skimmed' by centrifugation to prevent column blockage by fat. The robot selects a flask and transports it to the sampling station, where a probe is placed into the sample and a 20 ml pump is primed with the fresh solution. The rinse solution is pumped to waste. The probe is returned to a wash-bottle and the sample flask to the carousel. A fresh BondElut is selected from its position adjacent to the flask and installed on the clean-up station. The robot arm pushes the needle through the septum to ensure a good connection to vacuum, and withdraws to leave the electric clamp supporting the cartridge assembly. The robot is thus free to perform the next manipulative task, which is to install a 15 ml tapered tube on the bottom of the elution station. Simultaneously the clean-up routine commences, controlled by the SPIDER program. The dispenser head is positioned above the reservoir and 10 ml acetonitrile is dispensed onto the cartridge. A vacuum of ca. 400 mmHg is applied to achieve a sample flow rate of about 10 ml/min through the cartridge bed. A water rinse is then applied, followed by 20 ml sample solution, 10 ml water, 10 ml basic 10% acetonitrile and 10 ml acidic 10% acetonitrile. Each solvent is dispensed as the last of the previous wash is just draining through the cartridge. Vacuum is switched from station 1 to station 2, the dispenser head moves to its 'home' position, the cartridge is removed to the elution site by the robot and the


L. A. Gifford et al.

Downloaded by [University of Florida] at 23:21 14 November 2015

needle pushed through the septum. The dispenser head then moves to the elution station and 5 ml 30% acetonitrile is dispensed onto the cartridge. This is drawn through into the tapered tube and followed by a 10 ml water wash, giving a solution of aflatoxin Mi in 10% acetonitrile. The dispenser head returns to its 'home' position, the cartridge is removed by the robot and is deposited in a waste receptacle. The tapered tube containing the dilute eluate is removed from station 2 upon switching of the vacuum, and is transported to the sample load probe. This complete process takes nineteen minutes. Chromatographie analysis An aliquot of the prepared sample (5 ml) was aspirated into the fill loop of a pneumatically operated Rheodyne valve (model 5703, Rheodyne Inc, Cotati, CA). Bidirectional operation of the Rheodyne was controlled by three-way solenoid valves (Burkert Controlmatic, type 370/A) operated by the laboratory interface. In the inject position the contents of the loop were applied to a Chromsep 1 cm ODS guard column (Chrompack International B.V. Middelburg, Netherlands) connected across a second Rheodyne 5703 valve. Switching of the second valve caused elution of the concentrated aflatoxin Mi from the guard column onto the analytical column. Analysis was performed using an Altex model 110 pump (Beckman Instruments Inc., San Ramon, CA) a 250 x 4-9 mm 5 μχη Spherisorb ODS column and 1% acetic acidracetonitrileimethanol (60:30:10) mobile phase at 0-75ml/min. Fluorescence detection was employed (Perkin Elmer model LS5, Beckonsfield, UK) with the excitation and emission monochromators set at 355 nm and 433 nm, respectively. Peak integration was performed on a Trilab model 2000 integrator (Trivector Scientific Ltd, Sandy, Bedfordshire, UK). Preconcentration and chromatography occupy 22 min per sample. Results and discussion

The results obtained from spiked water samples (from a 2-5 /ig/ml standard in acetonitrile) and samples of fresh semi-skimmed milk spiked at 1·0, 0·5, and 0· 1 μg/kg aflatoxin Mi are shown in table 1. These results are compared with the manual results reported previously by Takeda (1984), and may be seen to be comparable in terms of recovery and repeatability. The automatic system is faster than manual methods, since overlapping of the clean-up and Chromatographie stages reduces the time per sample to below 25 min, but it also has the capacity for uninterrupted operation for extended periods, thereby providing further economies. It is possible to duplicate the peripherals and further increase sample throughput,

Table 1. Aflatoxin Mi recoveries. Method: Sample: Concentration (Ag/kg) 1-0 0-5 0-1

Automated Semi-skim milk

Aqueous Standard Percentagf ; recovery 76 69 48-1

(18) (18) (18)


c.v. 8-1 12-8 22

Percentage recovery 91-4 102-9 89-9

(18) (18) (18)

Semi-skim milk c.\1.

Percentage recovery


•1 9· 0 28

84-1 76-3

9-5 —

! •

Robotic analysis of aflatoxin



Downloaded by [University of Florida] at 23:21 14 November 2015


0.1 ppb EXTRACTS





10 min Figure 4. Typical chromatogram for aflatoxin Mi milk extracts.

since this is limited by the time for clean-up on the BondElut cartridge and Chromatographie analysis. From table 1 it can be seen that recoveries from aqueous spikes are 15-39% lower than from milk samples. This was found to be due to two problems: breakthrough from the BondElut in the basic and acidic acetonitrile washes, and incomplete elution in 30% acetonitrile. Cartridges used for the extraction of l-O/ig/kg aqueous spikes were re-eluted with 30% acetonitrile, Six


L. A. Gifford et al.

cartridges gave recoveries of between 0 and 11 % of the initial amount, with a mean of 3*8%. Cartridges previously used for the extraction of spiked milks yielded no detectable aflatoxin Mi. This may be due to inactivation of strong binding sites by substances present in milk. Figure 4 presents example chromatograms. Aflatoxin Mi elutes after 590 s and is preceded by a milk residue peak.

Downloaded by [University of Florida] at 23:21 14 November 2015


Unattended automated analysis of aflatoxin Mi in milk samples presents an improvement over manual methods of analysis from an economic point of view. The results validate the approach of using commercially available software control packages and custom-designated peripherals as an alternative to dedicated laboratory robotic systems. Flexibility has been achieved and the ability to handle large sample numbers (up to 18 in an 8 h shift) demonstrated. Reliability of the automation equipment is good, typical downtime being less than 2% working a single 8-10-h shift, 5 days per week for 9 months. Overnight working was precluded by university safety regulations. This downtime was for routine maintenance and calibration, rather than due to failure. References GILBERT, J., SHEPHERD, M. J., WALLWORK, Μ. Α., and KNOWLES, M. E., 1984, A survey of the occur-

rence of aflatoxin M1 in UK-produced milk for the period 1981-83. Food Additives and Contaminants, 1, 23-28. HEATHCOTE, J. G., and HIBBERT, J. R., 1978, Aflatoxins: Chemical and Biological Aspects (Amsterdam: Elsevier Applied Science). SHEPHERD, M. J., HOLMES, M., and GILBERT, J., 1986, Comparison and critical evaluation of six

published extraction and clean-up procedures for aflatoxin M1 in liquid milk. Journal of Chromatography,354, 305-315. STUBBLEFIELD, R. D., 1979, The rapid determination of aflatoxin M1 in dairy products. Journal of the American Oil Chemistry Society, 56, 800. TAKEDA, N., 1984, Determination of aflatoxin M1 in milk by reverse phase high performance liquid chromatography. Journal of Chromatography, 288, 484-488. WYNNE, R. J., and GIFTORD, L. Α., 1984, The application of robotics to flexible analysis systems. Conference Proceedings of the Institute of Mechanical Engineers, C460/84, 43-50.

Robotic analysis of aflatoxin M1 in milk.

An automated aflatoxin M1 assay system capable of performing multiple unattended extractions and chromatographic analyses has been developed. A six-ax...
676KB Sizes 0 Downloads 0 Views