On-line computer controlled multiple ion detection ... - ACS Publications

incorporated in the development of the accelerating voltage alternator (AVA) by Sweeleyet al. (3). In this technique, the accelerating voltage is swit...
0 downloads 0 Views 845KB Size
On-Line Computer Controlled Multiple Ion Detection in Combined Gas Chromatography-Mass Spectrometry John F. Holland, Charles C. Sweeley, Ronald E. Thrush, Richard E. Teets, and Mark A. Bieber Department of Biochemistry, Michigan State University, East Lansing, Mich. 48823 A system is described for on-line data collection and automated ion focusing of a single focusing magnetic deflection mass spectrometer equipped with gas chromatographic inlet and accelerating voltage alternator. A special interface consisting of an ion current amplifier, analog -to-d igita1 converter, compu ter, d igital-toanalog converter, and bipolar power supply permits an offset voltage to be added to or subtracted from the accelerating voltage under computer control. These units, together with two status registers, provide a control loop for continuous fine adjustment of the accelerating voltage in order to maintain optimal focusing on each of several ions selected by coarse sequential switching of the accelarating voltage. Tests of the system were performed in a static mode with perfluorokerosene and in a dynamic mode with several compounds introduced by the gas chromatographic inlet, including mixtures of fatty acid methyl esters, the tetra-0-acetyl methyl glycosides of glucose and [6,6-2H2]glucose, and the 0-acetyl methyl esters of prostaglandin PGFz, and 13,3,4,4-*H4]PGF~,. The additional capabilities of continuous ion focusing during the gas chromatographic run and precise selection of integration parameters in the computer-assisted data reduction process enabled stable isotopic abundance measurements to be made with a precision of less than 1% on samples as small as 100 ng.

THEREARE TWO APPROACHES to the problem of increasing the accuracy and sensitivity of the measurement of a small number of specific ions by the technique of combined gas chromatography-mass spectrometry (GC/MS). One approach consists of cyclical, rapid scanning over a limited mass range containing the ions of interest ( I , 2). The second approach is incorporated in the development of the accelerating voltage alternator (AVA) by Sweeley et al. (3). In this technique, the accelerating voltage is switched sequentially from its full value to a series of lesser values, each of which brings into focus an ion of selected mass at the electron multiplier. Applications for the GCjMSjAVA method to data have been varied. It has been used to correct for a chromatographic isotope effect in measurements by GC/MS of stable isotopic abundance in mixtures of protium and di species of glucose (3). Continuous monitoring of several ion intensities during a gas chromatographic separation (mass fragmentography) has been used in several studies of drugs and drug metabolites in extremely low concentrations (4-6). A new approach involving the dilution of biological samples with a stable isotopic-labeled species that serves both as carrier and internal standard has enabled isolation and quanti(1) R. A. Hites and K. Biemann, ANAL.CHEM.42, 855 (1970). (2) L. Baczynskyj, D. J. Duchamp, J. F. Zieserl, Jr., and U. Axen, The Upjohn Co., Kalamazoo, Mich., unpublished work, 1972. (3) C. C. Sweeley, W. H. Elliott, I. Fries, and R. Ryhage, ANAL. CHEM., 38, 1549 (1966). (4) C. G. Hammar. B. Holmstedt. and R. Ryhage, Anal. Biochem., ~25,532 (1968). ( 5 ) T. E. Gaffnev. C. G. Hammar. B. Holmstedt. and R. E. McMahon, ANAL:CHEM.,43, 307 (1971). (6) S . H. Koslow, F. Cattabeni, and E. Costa, Science, 176, 177 (1972). _

I

308

tative analysis of compounds such as prostaglandins (7, 8), insect juvenile hormone (9), and 5-hydroxyindoleacetic acid (IO) in the subnanogram range. Standard isotope dilution methods that involve the measurement of stable isotopic abundance ratios by the AVA technique have been used to determine the pool size of chenodeoxycholic acid in patients with cholelithiasis (11, 12) and the turnover rates of plasma glycosphingolipids have been determined in control subjects and patients with Fabry’s disease (13). In spite of the advantages of specificity and sensitivity, use of the multiple ion detection method has not, to date, been widespread and little of its actual potential has been exploited. One reason involves inherent difficulties with certain instrumental parameters of the measurement. Since this procedure requires prolonged use of the magnet at high current levels, temperature increases affect the magnetic field intensity which, in turn, alters the focus of specific mje values. Other variables can also affect the ion path and gradual changes in any one of them during a GC/MS analysis will defocus the beam and produce relatively large errors in the measured intensities of the ions. To minimize these errors, it has been standard procedure to manually refocus several times during a run. This procedure is usually carried out just before or during the initial part of an eluting peak and is done as rapidly as possible. At best, a small error is introduced and at worst the wrong ion is chosen in haste. A second source of error lies in the manual interpretation of base-line values and integration parameters. Poor reproducibilities by an individual operator coupled with gross variations between different operators produce significant variations in the resulting calculations of heights and areas. To reduce or eliminate these problems, the objective of this study was the development of a computer-controlled AVA for continuous in-line ion beam focusing, on-line, real-time data collection, and rapid objective data reduction and presentation. EXPERIMENTAL

Methods and Materials. Perfluorokerosene (PFK) was obtained from Pierce Chemical Co., Rockford, Ill. A mixture of methyl esters of saturated and monounsaturated fatty acids was prepared by acid-catalyzed methanolysis of free fatty acids extracted from Ivory soap (Procter and Gamble Co., Cincinnati, Ohio). These esters were separated iso~~

(7) B. Samuelsson, M. Hamberg, and C. C. Sweeley, Anal. Biochem., 38,301 (1970). (8) U. Axen, K. Green, D. Horlin, and B. Samuelsson, Biochem. Biophys. Res. Commun., 45, 519 (1971). (9) M. A. Bieber, C. C. Sweeley, D. J. Faulkner, and M. R. Petersen, Anal. Biochem., 41, 264 (1972). (10) L. Bertilsson, A. J. Atkinson, Jr., J. R. Althaus, A. Harfast, J-E. Lindgren, and B. Holmstedt, ANAL.CHEM.,44, 1434 (1972). (11) P. D. Klein, J. R. Haumann, and W. J. Eisler, ibid.,p 490. (12) P. D. Klein, J. R. Haumann, and W. J. Eisler, Clirz. Chem., 17, 735 (1971). (13) D. Vance, Ph.D. Thesis, University of Pittsburgh, Pittsburgh, Pa., 1968.

ANALYTICAL CHEMISTRY, VOL. 45, NO. 2, FEBRUARY 1973

thermally at 170 O C on a 1.4-m x 3-mm (id.) glass column packed with 3 x SE-30 on 80-100 mesh silanized Supelcoport, Supelco, Inc., Bellefonte, Pa. [6,6-*H2]~-G1ucosewas purchased from Merck, Sharpe and Dohme Ltd., Montreal, Canada, and D-glucose was obtained from Mallinckrodt Chemical Works, New York, N.Y. Methyl glycosides of preweighed samples of the sugars were prepared in 0.5 ml of 0.75N methanolic HC1 at 80 O C for 2 hr, after which the solutions were neutralized with solid silver carbonate. The mixtures were filtered and the methanol was removed under nitrogen; the residues were dissolved in 0.5 ml of pyridine: acetic anhydride 1 : 1 and let stand overnight at room temperature (14). The peracetylated methyl glycosides were recovered by evaporation of the solvents with added toluene under a stream of nitrogen, and working solutions of 1 mg/ml prepared in ethyl acetate. The 0-acetyl methyl glycosides were separated a t 185 "C on a 4-m x 2-mm glass column packed with 1 SE-30. Prostaglandin PGF2, and [3,3,4,4-*H4]PGF2, were generous gifts from John E. Pike and Udo Axen of the Upjohn Company, Kalamazoo, Mich. Triacetyl methyl esters were prepared in the following manner. A carefully weighed sample of the protium PGFz, and an estimated weight of the d? compound were treated separately with diazomethane in diethyl ether a t about 0 "C for 20 min, and the ether was then removed with a stream of nitrogen. Pyridine :acetic anhydride 1 : 1 (0.5 ml) was added and the solutions stood overnight at room temperature. After evaporation under nitrogen, a reference solution of the protium form was adjusted t o 1 mg,/ ml with dry ethyl acetate. The reference solution of the d 4 form was adjusted t o that of the protium form by comparison of their gas chromatographic peak areas and serial dilutions were made from these stock solutions. This method was necessary because the d4 species of PGF2, contained minor contaminants and some solvent of crystallization (15). Correction for the presence of small amounts of other deuterated species in the d 4 sample is discussed in greater detail later. The prostaglandin derivatives were separated at 230 " C on a 1.8-m x 3-mm glass column packed with 1 SE-30 on 100200 mesh silanized Supelcoport after prior conditioning for at least 24 hr a t 250 "C. Instrumentation. The LKB 9000 gas chromatograph- mass spectrometer was operated a t 70 eV with a trap current of 60 PA, ion source temperature of 29OoC, and molecular separator temperature of 240 O C . Carrier gas (helium) was a t a flow rate of approximately 30 ml/min. The standard AVA accessory of the LKB was used throughout this investigation. The mass spectrometer was interfaced t o a D E C PDP-811 computer for on-line, real-time data collection and reduction as described by Sweeley et al. (16). This computer system has recently been expanded t o include a Tektronix storage scope with a keyboard terminal, Model 4002A, and a Tektronix Model 4601 hard copy unit. The following modifications were designed and fabricated t o accommodate the problem of stable isotopic abundance measurements on gas chromatographic effluents. The bandpass of the three intensity amplifiers l X , l o x , and lOOx were appreciably narrowed by the addition of RC circuits to obtain optimum filtering for the switching speed and sensitivity desired. The outputs of these filtered amplifiers were attached to separate channels of the A D C multiplexor on the computer.

(14) R. A. Laine, W. J. Esselman, and C. C. Sweeley, "Gas-Liquid Chromatography of Carbohydrates," in "Methods in Enzymology," V. Ginsburg, Ed., Vol. XXVI, Academic Press, New York,

N.Y., 1972. (15) U. Axen, The Upjohn Co., Kalamazoo, Mich., personal com-

munication. 1972. (16) C. C. Sweeley, B. D. Ray, W. I. Wood, J. F. Holland, and M . I. Krichevsky, ANAL.CHEM., 42, 1505 (1970).

--tT

"3 ,,lay

-1

I

1

Ter m i hat ion

4

Retrieve From Tape

I 1I - 1 oT;;peI c

&I PH1 Data Processing and Output

Terminal

Unit

Figure 1. Block diagram of the AVA program

A ten-bit digital-to-analog converter (Pastoriza No. DAC10H) was interfaced t o the computer and its output (0 to -10 V) biased by a n analog circuit such that a voltage excursion from - 5 V t o 5 V was obtained by digital commands ranging from 0000~t o 17778. This analog voltage in turn controlled the output of a low impedance bipolar power supply which served as a n offset for the accelerating voltage of the mass spectrometer. The present version of the unit has a usable range of *lo volts, although a practical upper limit for this could be ten times that amount. The AVA unit itself furnishes the major high voltage offsets (up to 10z of full accelerating voltage) and makes the small range of the computer-controlled offset entirely adequate for fine focusing. This offset was inserted between the accelerating voltage supply and the chassis or reference point. The function of the offset will be described in detail in the next section. Two status flags were interfaced to the negative logic of the computer bus and enabled the computer, under program control, t o identify the position of the AVA switching relay. All other instrumental capabilities of the GClMS computer system were used without further modification. The Program. The software system developed for this investigation is shown in block diagram in Figure 1. The program is operated through a monitor routine to increase the ease and flexibility of the operator-computer interface, and has a number of loops which enable various specific situations t o be accommodated with optimum efficiency. For example, repeated analyses can be run and either outputted or stored on DECtape without redundant entries of experimental parameters or sample bookkeeping. This approach also allows direct access t o specific areas of the program whenever alteration is necessary. All of the keyboard commands must be executed from the monitor subroutine. When preparing the system for experimental operation, the setup routine is used t o enter comments about the sample, cycle speed (rate of AVA switching), and masses in a dialog with the computer via the keyboard. This routine must be re-entered whenever there is a change in compound or other experimental conditions. The focusing routine is entered whenever new masses are t o be focused or whenever the meter on the offset power supply indicates that a n excessive voltage is necessary for the cor-

+

ANALYTICAL CHEMISTRY, VOL. 45, NO. 2, FEBRUARY 1973

* 309

A B

TIME

____

Figure 2. Schematic of a time-dependent output from a single AVA channel Points A , B, and C are described in the text rection of drift and the limits of the supply are being approached. Once entered, the focus subroutine resets the offset supply to zero and permits manual focusing of each selected ion with the controls on the AVA unit. This routine also presents the option of independent focusing for each ion or the slaving of any ion to an offset determined from another ion. Thus, a very small peak can be kept in focus over long periods of time by maintaining a fixed difference between it and a larger ion in its vicinity on which the system will have excellent focusing ability. There is also an option to completely negate the automatic focusing, resulting in the passive use of the computer for data collection and reduction. The output routine can be entered from the data collection routine or from the monitor, either directly or after data retrieval from the magnetic tape system. The interaction between the operator and the accumulated data to establish the various output formats is accommodated in this section, providing flexibility in the scaling of time and intensity axes. Manual or automatic integration limits and background indices can be selected. Weighted data-smoothing manipulations can be applied and ratio calculations between selected mJevalues can be performed and outputted. By far the most important software contribution to this technique lies in the data collection and ion beam focusing algorithm. These portions of the software shall be considered in detail since they constitute the major advantage of the system over manual operation. The standard AVA unit can monitor up to three different ion intensities. These ions are focused manually using the magnetic field control to focus the lowest m/e value when the AVA switch is in the full position. The higher rnje values are brought into focus with the coarse voltage divider range switches and the fine focus controls on the appropriate AVA channels. During this time, the computer-controlled offset is at midscale or zero volts. This makes it possible to use the mass marker to focus the lowest rnje value and readily permits the calculation of the coarse voltage changes necessary to focus the ions of higher mass. These focusing steps are usually carried out on appropriate ions from a reference compound such as PFK, but the specific compounds to be analyzed can also be used. Since the objective of the computer control loop is to establish and maintain focus accurately, the manual adjustments need not be excessively precise. The data collection program is initiated coincident with the injection of the sample into the gas chromatograph. Due to the instrumental necessity of allowing solvents to pass into the inlet pumping system before the analyzer valve is opened to the effluent stream, the actual data input routines are delayed until the valve has been opened, at which time they are manually initiated by a keyboard signal. Prior to this time, the computer keeps a record of the number of switches or cycles of the AVA unit. Elution times of individual gas chromatographic peaks are later calculated from these accumulated cycles. 310

After initiation of the data input routine, the computer becomes the active element in the experiment and continues to function until the routine is terminated at the keyboard. During this time, the computer performs three tasks. It monitors each of the ion beams to determine the presence of peaks, it maintains the focus of each ion beam, and, last, it collects and stores the intensity data from each channel using the amplifier with the most favorable gain for each set of data points obtained. Optimum gain is attained by program-controlled autoranging up and down through the three amplifiers with gains of 1x, lox, and 1OOx. Following each change in the AVA relay, a time delay is initiated that permits the filter amplifiers to settle into the new value for the ion beam brought into focus. Need for this delay is illustrated in Figure 2 , which is a schematic representation of the actual signal received as it relates in time to the switching and focusing events. For clarity, only one of the three channels is shown in the figure, but each of the operations described below will apply in sequence to the other channels. The series of square pulses on the bottom of the figure indicate the status of the flag for the channel being displayed. This digital logic level is in the high or 1 position only when the channel is connected to the accelerating voltage circuit by relay closures on the AVA unit. The time lag of the intensity signal is due to the electronic filtering of the amplifiers, the amount of which was chosen as a compromise of sensitivity versus accuracy. The magnitude of this delay will obviously be dependent on the voltage difference between the two levels involved for the intensities of the two ions. The actual minimum deIay used in this investigation accommodated full scale voltage changes prior to data collection. All of the delays are accomplished by the computer under program control. During the delay, the computer sets the offset voltage to the value where the ion was last focused. This offset is initially at zero but as drift occurs, the computer will change this value to maintain optimum focus. After the delay, the computer offsets the accelerating voltage an amount commensurate with approximately ‘18 peak width at half height and obtains with the ADC an intensity data point after a 6-msec delay. Each data point is the median of 5 successive A/D conversions at a rate of 20 KHz. This action will defocus the beam as shown by the dip in the curve in Figure 2 after point A . The computer then changes the voltage in 11 successive equal steps which have a total traverse of approximately width at half height when scaled to the shape of the peak. Since this value will vary with the m/e value focused, the actual magnitudes of the large step down and the small steps over the summit of the peak are selected by the computer so as to be compatible with the mje value being measured, This provision maintains a fairly linear sensitivity for the shape of the ion peak throughout the mass range. A data point is taken at each step upward in voltage and is stored in a temporary file. During this time, the focus will pass from one side of the ion beam, through the optimum focus, and start down the other side as indicated in Figure 2 between A and B. At the conclusion of this procedure, the temporary file is examined for the presence of a peak and where affirmative, the center of the peak is determined by a modified center of gravity routine and the voltage switched to that value, Fifty data points are then averaged to gain an accurate value for the intensity of the focused ion at that point in the elution of the gas chromatographic peak, shown in Figure 2 as the plateau portion of the curve after B. At the end of this routine, the computer stores the intensity data point and the voltage for optimum ion beam focus for future focusing and then switches the offset to the value for the next ion to be measured. Where appropriate in the cycle, it will also increment the cycle count. The computer then waits for the next status flag and sets up pointers for the next ion. This process is repeated for each ion in succession with three data files being accumulated as a function of the cycle

ANALYTICAL CHEMISTRY, VOL. 45, NO. 2, FEBRUARY 1973

Table I. Precision of AVA Analysesa Difference in ratio, Height Area Expt No. 1 2 3 4 5 6

0.20 0.02 0.26 0.58 0.07 0.35 0.25 =t0.20

12 0

0.23 0.13 0.60 0.96 0.02 0.91 0.48 i~0.41

Average Static determination of the ratios of m/e 269/267 for perfluorokerosene as described in text.

time and stored on the disk. The program is exactly the same for all three channels. Transfers of a small block of pointers and current status values for each channel accompany the channel changes as they are sensed by the IjO flags, an arrangement which makes most efficient use of core memory and allows for easy expansion to accommodate more than three ions if needed in future applications. The sequence of the AVA switching cycle goes from A to full t o B to full, etc. The second full position in each cycle is used to store on the disk the collected data from all three channels. Full positions are sensed by the computer as the disappearance of the A or B flag, indicating the respective switch opening to restore full accelerating voltage. To operate most efficiently, the focusing routine has certain restrictions. Two conditions must be satisfied before any changes are permitted in the offset value representing the center of focus for a given ion. First, the presence of a definite peak, with an intensity maximum in excess of a preselected value, must be confirmed and, second, the rate of change of the intensity for successive data points must fall within a set of lower and upper boundaries. These restrictions preclude the possibility of the focus shifting aimlessly when the ion intensity is very low and not changing, and stops any attempt at focus during the very rapidly changing portions of a gas chromatographic peak. Without these restrictions serious errors can occur from incorrect evaluations by the focusing algorithm. Thus, in effect, focusing occurs only on the gentle sloping initial and final portions of a gas chromatographic peak, an arrangement that has been adequate for accurate and sustained focusing. When the conditions for focus are met, the computer locates the ion beam maximum as a function of the controlled offset voltage and continually updates this value for succeeding measurements. As long as any portion of an ion beam is present and measurable by the electron multiplier, the computer can bring the beam into maximum focus within a few cycles and can still make small changes that are necessary t o accommodate conditions resulting from a slow drift. At C in Figure 2, the focus was manually disturbed and the algorithm immediately responded with changes in the offset voltage until the peak maximum was again located, requiring four cycles to restore focus. Tests of this perturbation indicated that the system will accommodate sudden changes of 1 0 . 2 5 amu from the center of focus on ion peaks of moderate size and will always recover in from three to five cycles. Such gross changes are normally encountered only on the first peak of a sample run when the m/e value has been focused originally on an ion with a sizable difference in mass defect from that of the sample. Little deterioration in accuracy on the first chromatographic peak is obtained, provided the peak is of moderate size. However, if the peak is small, significant data may be lost during the gross focusing process. When observed, the f i s t run can be deleted from the data files and this sample analyzed again. The range of immediate recovery of the focusing routine has been experimentally derived and is sufficient for the highest drift

0

2

4

6

8

T ?ME Figure 3. Specific ion intensity chromatograms of a mixture of fatty acid methyl esters rates encountered in normal use of the multiple ion detection technique. At the conclusion of a gas chromatographic run, the automatic data collection routine is manually aborted via the keyboard, whereupon the data may be processed or stored on magnetic tape for future processing. RESULTS AND DISCUSSION

The method of multiple ion detection was employed on several compounds in order to evaluate the operational parameters of the computer-controlled AVA system. The validity of the data collection and focusing algorithms and the accuracy and stability of the analog amplifiers were determined under a static experimental condition. A constant level of perfluorokerosene was maintained in the ion source via the heated membrane inlet system of the mass spectrometer. Table I presents the precision in measurement between channels A and B, with each independently focusing on the same ion beam. The full position was focused on mje 267 with both A and B focused on mje 269. This should produce equal ratios between A and full and B and full. The average per cent difference in the ratios of peak heights was 0.25z, as opposed to 0.48z for that of the areas. This illustrates an unusual relationship that has often been observed in this investigation. The reproducibility of peak heights exceeded those of the areas, which is in contrast to what would be expected from such experiments, since electronic noise and deviations from gaussian distributions would affect the value obtained for a singular measurement of height more than

ANALYTICAL CHEMISTRY, VOL. 45, NO. 2, FEBRUARY 1973

311

314

Table 11. Precision of AVA Analysis by Combined Gas Chromatography-Mass Spectrometrya Difference in channels, Expt Methyl palmitate Methyl stearate No. Height Area Height Area 1 2

0.70

0.54 0.38 0.86 1.37 2.10 1.40 3 0.59 0.30 0.54 0.52 4 0.96 2.17 2.34 2.52 5 0.92 0.44 1.17 0.42 6 0.30 0.66 0.85 0.46 Average 0.75 i.0.27 0.91 f 0.72 1.23 =k 0.81 1.03 & 0.82

314

1.oo

a Comparison of the intensity of m/e 87 from two channels as described in text.

31 8

i

:

Figure 4. Real-time data display of a mixture of the tri-0-acetyl methyl esters of prostaglandin PGF,, and [3,3,4y4-2H4]PGF,, in a protium/deuterium ratio of 4/1000

that obtained for an area. Because of the persistence of this observation, however, the use of peak heights for quantitative determinations has often been preferred. This steady state test also defined the limit of accuracy that can be expected from the system, since experimental variations introduced by changes of ion intensity during elution of a gas chromatographic peak will certainly not enhance the measurement conditions. A specific ion intensity chromatogram of mje 87 obtained during gas chromatography of a mixture of fatty acid methyl esters is shown in Figure 3. The top curve represents the output of two separate channels, both independently focusing under computer control on the same ion. Their congruence provides evidence for the excellent quality of focusing and the accuracy of the analog amplifiers. The lower curve shows the data as received by the computer from the output of two amplifiers (lox and lOOX) prior t o software scaling and normalization. Table I1 lists the results of a precision test between the two channels from data obtained under experimental conditions similar to those used for Figure 3. Repeated runs indicate a relative precision of 0.91 in area and 0 . 7 5 x in peak height for the completely resolved peak of methyl palmitate, and the less favorable values of 1.03% and 1.23x for the areas and heights, respectively, of the partially resolved peak of methyl stearate. Increased variations for partially resolved GLC peaks can best be accounted for by consideration of the methods used for determination of the base line and selection of the integration limits. The extended duration of a partially resolved peak increases the probability of a change in the base-line value from the beginning to the end of the peak. In routine operation, the points used for the calculation of the base line are determined automatically by the computer after examination of the data file. During this operation, the computer 312

selects the points just prior to and immediately after a specific peak. In the case of a peak for a partially resolved mixture, this method can introduce variations since the actual measured points used as the terminating boundary for the calculation of the base line are taken after the last section of the peak. The base line for each of the various sections of the partially resolved peak must then be interpolated from these initial and final extremes, a situation which can reduce the accuracy of the height that is calculated for each section. Since these points also serve as the integration limits for the peak areas, the accuracy in the areas is likewise affected. The great variety of geometric configurations that can be obtained by partial G L C separation of two components makes an option enabling manual selection of these limits desirable. This option is included in the calculation and output routines and enables the operator to specifically delineate these boundary values after visual examination of the data file. The data listed in Table I1 were obtained by the automatic mode of integration and base-line determination; however, in most of the subsequent experiments, the manual method was employed in order to provide a more initimate operator interaction with the data. The real-time data display as it appears on the face of the storage scope is illustrated in Figure 4. This display was obtained during a GLC separation of one microgram of a mixture of the tri-0-acetyl methyl esters of prostaglandin PGF2, and [3,3,4,4-2H4]PGF2, in a protium/deuterium ratio of 4/1000. The intensity axes are autoranged by the computer under program control in order to collect the data from the appropriate amplifier. Each point is displayed as it is taken subsequent to storage on the disk. The time axis is adjusted to the cycle speed and the anticipated length of the GLC run and each ion is identified and plotted separately so that the operator can review the collection process and immediately ascertain ihe need for any changes in experimental parameters. In the collection of these data channel A and full F were each focusing independently on m/e 314 while channel B was focusing on m/e 318. This illustrates the very useful analytical capability of an internal redundancy in the measurements of the ion intensities. In practice, this duplication is used on the less intense of the two ions selected for study, since optimum focusing becomes increasingly difficult as the limits of sensitivity are approached. Focusing then becomes the limiting factor in the accuracy of a specific measurement. The ability t o obtain two independently focused values for a single ion peak greatly increases the confidence in the value obtained. A large divergence between these redundant Values necessitates a repeated run of the sample.

ANALYTICAL CHEMISTRY, VOL. 45, NO. 2, FEBRUARY 1973

~~~~~

~

~

Table 111. Least Square Analysis of Isotope Dilution Data' Slope Intercept Height Area Height Area Methyl glucoside tetraacetate and methyl [6,6-2H2]gluco1.029 0.989 -0.003 side tetraacetate: deuterium/protein mixtures* +0.015 Tri-0-acetyl methyl esters of prostaglandin PGF?, and [3,3.4,4-2Hl]PGFzo: deuteriumjprotium mixturesc,d 0.943 0.946 $0 ,023 +0.013 protiumjdeuterium mixturescsd 0.922 0.948 $0.018 $0.010 a Each data point represents 4-6 determinations. Dilutions used: l/lOOO, 2/1000, 4/1000, lO/lOOO, 20/1000. Dilutions used: l/l000, 2/1000, 4/1000, lO/loOa. d Calculated from corrected isotopic ratios as described in text.

Correlation coefficient Height Area 0.9998

0,9999

0.9979 0,9999

0.9981 0,9999

20

10

4 2 I

I

I

2

4

10 ng/Hg

I 20

PROTIUM

Figure 6. Working curve for mixtures of the tetra-0-acetyl methyl glycosides of [6,6-*H2]glucoseand glucose

2

4

6

a

TIWE (MIN) Figure 5. Ion intensity plots of prostaglandin isotope dilution experiment Channels A and full F represent m/e 314 at 50X, and channel B represents mje 318 at 1 X . The lower curve has an offset time axis Another benefit from channel duplication lies in the increased statistical accuracy with which a n ion may be measured. An extension of this approach would be the use of additional independently focusing channels for a specific ion, with the objective of an increase in precision and accuracy. One of the actual output formats from the run displayed in Figure 4 is shown in Figure 5. The data were plotted on an incremental plotter after a data smoothing routine was used. The ion intensity curves for m,/e 314 were plotted with a n intensity scale of 50X, while the larger ion at mje 318 was plotted with a scale of 1 x. The output routines will permit

any desired scaling of both time and intensity axes. In the lower curve, the convenient feature of the horizontal offset enables peaks of nearly the same height to be separated while still using the same scale on the vertical axis. A working curve for a dilution series of the tetra-0-acetyl methyl glycosides of glucose and [6,6,*H2]glucose is shown in Figure 6. The ions selected for monitoring these compounds were mje 200 and mje 202, as described by Vance (13). For this derivative, these ions exhibit little cross contribution between the protium and deuterium forms. A 7 % contribution at mje 200, due t o another ionic species in the ds form, can be neglected with mixtures containing small proportions of the dz species. The blank value for the ion pair mje 2021 200 in the pure protium compound was 0.0240 + 0.0002 which compares favorably with the value of 0.0255 Z!Z 0.0005 obtained by Vance on a similar mass spectrometer/AVA system with manual operation (13). F o r isotope dilution studies where there is negligible cross contribution between the two ions, the difference between the observed ratio of a given sample and the ratio of the blank is linear with the concentration of the isotopically labeled species (17). Thus, the blank value was subtracted fiom the actual observed ratios for the various dilutions t o obtain the data points for (17) J. F. Holland and R. E. Teets, Michigan State University, East Lansing, Mich., unpublished work, 1972.

ANALYTICAL CHEMISTRY, VOL. 45, NO. 2 , FEBRUARY 1973

313

Table IV. Determination of Molecular Ion Composition of Methyl Palmitate by AVA Analysis Ratio ( X 100) Height Area Ionpair Expt No. 2711270 1 18.727 18.561 2 19.015 18.795 3 19.103 19,221 4 18,807 18.561 Average 18.913=kOo.146 1 8 . 8 0 5 i 0 . 2 5 4 Predicteda 18.992 2721270 1 2.0822 2.1265 2 2,0785 2.1063 2.1141 2.1625 3 2.0733 2.1148 4 2.0879 rt 0.0135 2.1275 + 0.0175 Average Predicteda 2,1052 2711272 Averageb 9.110 8.839 Predictedu 9.021 Predicted values are those reported for Ci7H3402by Beynon and Williams (18). b Calculated from average values for 271/270 and 272/270. (1

Figure 6. The precision of this blank value serves as the limit of detectability for the dideutero form. In this case, the value would be a ratio of lj5000, and would serve as the absolute value of accuracy. The variation in the blank increases as smaller injections are used, rising to rt0.0005 for amounts as low as 100 ng, thus producing a theoretical optimum of li2000 as the limit of accuracy with this amount. In this investigation, sample sizes of 500 ng were generally used. The linearity of the curve in Figure 6 indicates that the calculated stable isotopic ratios provide a n accurate quantitative determination for the proportion of d2 species in the mixture. The data from the glucose series and two similar experiments involving prostaglandin isotope dilutions are analyzed in Table I11 by least square analysis. The coefficients of correlation are excellent in all cases, while the slopes and intercepts exhibit a greater variation for the prostaglandin series. In these experiments, the acetyl methyl ester derivative exhibited a marked shoulder preceding the main GLC peak. This chromatographic finding was more evident with the protium than the de form, making the setting of integration limits and base-line determination less exact. The ion selected to monitor the protium form was m/e 314 and that for the de form, m/e 318 (8). These fragments represent loss of three acetic acid residues from the molecule and were chosen because of their relatively high intensity and specificity for the individual isotopes. The blank value for the ion pair m/e 318/314 in the reference protium compound was

314

0.00255 i 0.00025 and that for the ion pair m/e 314/318 in the d 4 compound was 0.00394 + 0.00008. The proportionality between the observed intensity of m/e 314 and the amount of the protium form was unity. The concentration found for m/e 318 was corrected for a 5.5 impurity in the d 4 form due to the presence of other deuterated species (6) and for loss of a deuterium atom in a reproducible proportion of ion fragmentations. The determination of natural stable isotopic abundance ratios was performed on methyl palmitate and Table IV lists the results of this application. The three ions M (m/e 270), M 1 (m/e271), and M 2 ( m / e 272) were monitored simultaneously on the three channels F , A , and B. The observed precision of about 1 was maintained throughout many experiments of this type over a relatively long period of time. However, the actual values varied with certain instrumental conditions such as electron multiplier voltage, ionization potential, and the slit settings of the mass spectrometer. Sample sizes of less than 100 ng also produced a variation in the ratios. The data used in the calculations for Table IV were all obtained with 500 ng of sample and under identical instrumental and operational conditions. The conditions selected produce a good agreement with the computer-predicted values of Beynon and Williams (18). The accuracies attained are within an order of magnitude less than that to be desired for determining natural stable isotopic abundance, but are an order of magnitude greater than those calculated from data obtained from the normal scanning mode of operation of the mass spectrometer. This level of' accuracy is sufficient to detect changes in the isotopic ratios of known compounds as they undergo chemical reactions. Thus, this technique could have numerous applications in biochemical and stereochemical fields.

x

+

+

ACKNOWLEDGMENT

The authors are grateful t o Jack Harten for assistance in the data collection and Norman Young for helpful discussions on the program. RECEIVED for review July 25, 1972. Accepted October 6, 1972. This work was supported in part by research grant (RR-00480) and a biochemical training grant (GM-1091) both from the U S . Public Health Services. It is published with the approval of the director of the Michigan Agriculture Experiment Station as Journal Article 6009. (18) J. H. Beynon and A. E. Williams,"Mass and Abundance Tables for Use in Mass Spectrometry," Elsevier Publishing Co., New York, N.Y., 1963, p 139.

ANALYTICAL CHEMISTRY, VOL. 45, NO. 2, FEBRUARY 1973