Anal. Chem. 2000, 72, 5881-5885

Internal Calibration on Adjacent Samples (InCAS) with Fourier Transform Mass Spectrometry

Peter B. O’Connor* and Catherine E. Costello

Mass Spectrometry Resource, Department of Biochemistry, Boston University School of Medicine, 715 Albany Street, R806 Boston, Massachusetts 02118

Using matrix-assisted laser desorption/ionization (MAL- ability to isolate and fragment selected ions for MSn, and high DI) on a trapped ion mass spectrometer such as a Fourier mass accuracy and resolution. This last property, high mass transform mass spectrometer (FTMS) allows accumula- resolution/accuracy, is limited primarily by the ion abundance- tion of ions in the cell from multiple laser shots prior to dependent space charge frequency shifts that are commonly in detection. If ions from separate MALDI samples are the ∼10-100 ppm range depending on severity, although space accumulated simultaneously in the cell, ions from one charge errors can usually be minimized by using a calibration sample can be used to calibrate ions from the other spectrum with total ion abundance similar to the analyte spectrum. sample. Since the ions are detected simultaneously in the Several interesting new methods have been proposed to correct cell, this is, in effect, internal calibration, but there are these frequency shift parameters including Easterling’s linear no selective desorption effects in the MALDI source. This frequency adjustment13 and Bruce’s DeCal method.14 These method of internal calibration with adjacent samples is methods propose to correct the frequencies of the ions by a demonstrated here on cesium iodide clusters, peptides, known, constant space charge factor applied before mass calibra- oligosaccharides, poly(propylene glycol), and fullerenes tion and show significant promise in this regard. The alternative and provides typical FTMS internal calibration mass is internal calibration. accuracy of <1 ppm. Internal calibration is common throughout mass spectrometry and consists of mixing several compounds of known molecular Matrix-assisted laser desorption/ionization (MALDI)1-4 has not weight with the analyte prior to mass analysis and then using the only become a routine ionization mode in mass spectrometry known masses to calculate the unknown ones. The improvement ∼ laboratories but it also enables use of a crucial method for in mass accuracy is significant (usually 10-fold) and very reliable; biochemical research. The ability to accurately measure molecular however, generating a mass spectrum of the mixture is sometimes weights of a wide variety of large biomolecules including pro- problematic as ionization efficiencies vary significantly among teins,2,5,6 oligosaccharides,7,8 and oligonucleotides4 up to several standards and maintaining signal parity between the calibrant and the analyte can be difficult. million daltons in molecular weight has revolutionized the field. When MALDI mass spectrometry is performed on a time-of- Also, the continuing evolution of methods to perform tandem mass flight instrument, it is common to place a spot containing a spectrometry9 allows biochemists to obtain more information than calibrant sample on the target near to the spot having the analyte ever before from their samples. of interest. Then, one calibrates the mass spectrometer im- Fourier transform mass spectrometry (FTMS) incorporating mediately prior to acquiring the spectrum of the analyte, thus a MALDI ion source allows some unique advantages such as the minimizing the calibration shifts due to slow temperature changes ability to store and accumulate ions10,11 in the Penning trap,12 the in the length of the flight tube, surface aberrations in the MALDI target such as pits or scratches, or field effects that vary with * Corresponding author: (phone) (617) 638-6705; (fax) (617) 638-6761; (e-mail) [email protected]. sample position on the target that can cause mass errors in the (1) Karas, M. I.; Bachmann, D.; Bahr, U.; Hillenkamp, F. Int. J. Mass Spectrom. 0.1% range. This method is, however, an external mass calibration - Ion Processes 1987, 78,53 68. and is subject to the attendant shot-to-shot mass accuracy shifts. (2) Hillenkamp, F.; Karas, M.; Beavis, R. C.; Chait, B. T. Anal. Chem. 1991, 63, 1193A-1203A. However, the same method can be used in an FTMS, wherein (3) Karas, M.; Bahr, U.; Giessmann, U. Mass Spectrom. Rev. 1991, 10, 335- the ionization and detection processes are almost completely 357. decoupled, and the ions from the calibrant and the analyte spots (4) Berkenkamp, S.; Kirpekar, F.; Hillenkamp, F. Science 1998, 281, 260- 262. can be accumulated together in the Penning trap prior to analysis. (5) Castoro, J. A.; Wilkins, C. L. Anal. Chem. 1993, 65, 2621-2627. As the trapped ions are simultaneously analyzed, any space charge - (6) Fenselau, C. Anal. Chem. 1997, 69, 9, A661 A665. shifts or other drifts in the mass calibration are canceled by doing (7) Okamoto, M.; Takahashi, K.; Doi, T.; Takimoto, Y. Anal. Chem. 1997, 69, 2919-2926. internal calibration. This technique is shown here and in a related (8) Harvey, D. J. Mass Spectrom. Rev. 1999, 18, 349-450. (9) McLafferty, F. W. Tandem Mass Spectrometry; Wiley: New York, 1983. (13) Easterling, M. L.; Mize, T. H.; Amster, I. J. Anal. Chem. 1999, 71, 624- (10) O’Connor, P. B.; Costello, C. E. Anal. Chem. 2000, 72, 5125-5130. 632. (11) Mize, T.; Amster, J. Anal. Chem., submitted. (14) Bruce, J. E.; Anderson, G. A.; Brands, M. D.; Pasa-Tolic, L.; Smith, R. D. J. (12) Penning, F. M. Physica 1936, 3, 873-894. Am. Soc. Mass Spectrom. 2000, 11, 416-421.

10.1021/ac000770t CCC: $19.00 © 2000 American Chemical Society Analytical Chemistry, Vol. 72, No. 24, December 15, 2000 5881 Published on Web 11/08/2000 paper11 with a wide range of samples selected to be extreme cases ground potential to the trapping voltage of 20 V after a delay of of differences in ionization efficiencies. The samples range from approximately 500-1000 µs from the laser pulse varied according cesium iodide clusters, fullerenes, and poly(propylene glycol) to to the mass range expected. A gas pulse is then injected into the oligosaccharides and peptides. Mass accuracy of 0.1-1 ppm is mass spectrometer through a pulse valve system (pressure in the typical. analyzer stage of the vacuum system rises to ∼1 × 10-6 Torr for 2 s) for cooling the ions to the center of the cell. The ions are EXPERIMENTAL SECTION then held in the cell for 5-10 s while the pressure pumps back ∼ × -10 These experiments were performed using an IonSpec HiRes- down to a base pressure of 1 10 Torr prior to excitation MALDI Fourier transform mass spectrometer with a 7-T active (frequency sweep from 100 to 2000 Da in 4 ms at 130 V amplitude) shielded superconducting electromagnet. This instrument has a and detection (acquisition rate of 2 MHz into 512k 12-bit data differentially pumped vacuum system with three turbomolecular points). The analyzer tube pressures are measured with an ion ∼ pumps (Pfeiffer TMU 260), analog electronics (24 DAC voltages, gauge positioned 50 cm from the cell and above the turbopump s 16 TTL triggers, rf quadrupole circuit, high-power excitation flange so that these pressures are only approximate to the actual amplifier, mixer/preamplifier for detection), dual pulsed valve inlet pressures in the cell. The signal is then apodized (quarter-sine), system, two ionization guages (Granville Phillips model 274 zerofilled, and magnitude mode Fast-Fourier transformed. Peak Bayard-Alpert type), quadrupole ion guide, and capacitively picking is done by a standard method, and the peaks are coupled closed cylindrical cell (similar to the design by Beu and centroided by fitting the upper half of the peak to a second-order Laude,15 but with additional external trapping plates). The ion polynomial and finding the point of zero gradient (note that this requires a minimum of three points on the top half of the peak). source is described below. The system shares the magnet, data The sample spots are deposited on the facets of the probe by system, and computer with a modified IonSpec HiResESI mass carefully depositing ∼1 µL of saturated 6-azo-2-thiothymine (ATT) spectrometer. The magnet is a Cryomagnetics 7-T active shielded matrix in methanol in a thick layer. The layer of matrix is created superconducting magnet with a 5-in. warm bore; the 5-G line is by using an Eppendorf gel-loader tip (Brinkmann Instruments, ∼10 cm from the face of the magnet. The data system consists of Inc. Westbury, NY) dipped into the matrix solution so that a small a 40-MHz frequency synthesizer, an Arbitrary waveform generator quantity of the matrix enters the tip by capillary action and then (40 MHz, 1 megapoint), voltage sequence programmer (for spotting this droplet onto the probe. This is repeated 20-50 times controlling DACs as a function of time through the pulse until a completely white layer of matrix is created. Then, 1 µLof sequence), pulse sequence generator (1 µs timing resolution), and analyte solution, typically dissolved at 10 µM in water with 0.1% transient digitizer (2 megapoints in acquisition and accumulation trifluoroacetic acid, is dropped onto the surface and allowed to buffers). The computer is a Pentium III 600 computer with 128- dry. The cesium iodide spots were created using a similar MB RAM, 8-GB hard drive. The instrument control software is technique; however, no matrix was used and only a saturated IonSpec’s Explorer version 6.72 running in MS-DOS. solution of CsI in water was pipetted onto the probe. The cesium The MALDI mass spectrometer uses an IonSpec 10-faceted iodide laser desorption/ionization (LDI) spectra required ∼30% sample probe, a 377-nm N laser (Laser Sciences, Franklin, MA) 2 more laser power than the MALDI spectra of peptides or with a motor-driven beam steering mirror (Newport Optics, Irvine, oligosaccharides. The fullerene sample (saturated solution of C CA), a motor driven focusing lens (Newport Optics, Irvine, CA), 70 in benzene) was deposited on the probe in a thick layer, and the an iris, and a manually adjusted gradient filter for adjusting the benzene was allowed to evaporate, leaving a fullerene layer on laser spot power. The spot is imaged with a CCD camera the probe. The laser desorption of the fullerene required much (Panasonic model GP-KR222), and the signal is displayed on a less power than either the CsI (∼50%) or the MALDI (∼75%) color monitor (Sony model PVM-14N2M). Sample spot preparation samples. To select among the various spots on the probe, the is described below. The plume generated by the laser desorption probe is manually rotated while the laser beam’s position remains erupts normal to the target surface and at 45° to the entrance fixed. The laser power was varied by manually rotating a gradient orifice to the mass spectrometer. The ions are then extracted filter to reduce the laser beam’s intensity prior to focusing it onto through the orifice (the sample surface is physically grounded, the probe. Unfortunately, we do not have an appropriate method the “right” and “left” electrodes are typically biased at +50 V dc to quantify the laser power and power density other than to use for positive ions, and the “extraction” electrode is biased at -10 the laser’s power specifications (300 µJ/pulse, focused to a ∼0.5- V ), into the rf-only quadrupole ion guide16-18 (biased at -40 V , dc dc mm spot) and reduce that by estimating the transmission ratio 100-300 V , ∼1 MHz) where they traverse the magnetic field bp through the currently used region of the gradient filter. However, gradient and enter the high magnetic field region of the mass as long as the laser is working well, the relative power differences spectrometer. The ions are trapped using gated trapping with the quoted above are adequate for day-to-day laboratory usage. outer trapping plates of the capacitively coupled closed cylindrical The internal calibration with adjacent samples (InCAS) experi- cell. For gated trapping, the trap plates typically are pulsed from ments also employed gated trapping with pulsed gas cooling, but did so by dropping the front trap plate to 1 V instead of dropping (15) Beu, S. C.; Laude, D. A., Jr. Int. J. Mass Spectrom. Ion Processes 1992, 112, 215-230. it to ground potential so that previously trapped ions would not (16) McIver, R. T., Jr. Int. J. Mass Spectrom. Ion Processes 1990, 98,35-50. escape. The stitched pulse sequence method previously de- (17) McIver, R. T., Jr.; Hunter, R. L.; Bowers, W. D. Int. J. Mass Spectrom. Ion scribed10 was used to allow easy tuning of the number of laser Processes 1985, 64,67-77. (18) Huang, Y. L.; Guan, S. H.; Kim, H. S.; Marshall, A. G. Int. J. Mass Spectrom. shots of analyte to the number of laser shots of calibrant added Ion Processes 1996, 152, 121-133. to the cell prior to detection.

5882 Analytical Chemistry, Vol. 72, No. 24, December 15, 2000 Figure 2. InCAS (A) cesium iodide with C70 fullerenes and (B) Figure 1. InCAS with cesium iodide and the oligosaccharide polypropylene glycol with the peptide substance P. Calibrant peaks stachyose: (A) one shot of CsI followed by one shot of stachyose; are marked with “@”. (B) one shot of CsI followed by two shots of stachyose; (C) one shot of CsI followed by three shots of stachyose. calibration of a peptide (substance P with its metastable water loss fragments) using poly(propylene glycol). Again, in both cases, DISCUSSION the ionization efficiencies of the calibrant and the analyte are vastly The InCAS method is demonstrated in Figure 1. Figure 1A different so that selective ionization would be a serious concern shows one shot of cesium iodide accumulated with one shot of in typical internal calibration. the oligosaccharide stachyose (â-D-fructofuranosyl-O-R-D-galacto- The method of accumulating ions over several laser shots in pyranosyl-(1f6)-O-R-D-galactopyranosyl-(1f6)-R-D-glucopyrano- the Penning trap has the interesting effect that trapping efficiency side) in the Penning trap prior to FT detection. The cesium iodide for the later shots in the cell is much higher than the trapping + + 10,11 primarily yielded three peaks corresponding to the Cs2I ,Cs3I2 , efficiency for the first shot. This is also demonstrated nicely + and Cs4I3 clusters. The stachyose yielded the sodiated molecular in Figure 3. In this case, two similar experiments are compared. ion at m/z 689.149 and the metastable fragment corresponding In the first (Figure 3A), one shot of cesium iodide is loaded into to the loss of a single hexose unit. There is also a minor abundance the Penning trap followed by two shots of stachyose. In the second of the potassiated peak. Figure 1B shows accumulation of one (Figure 3B), two shots of stachyose are loaded followed by one shot of cesium iodide with two shots of stachyose. Figure 1C shot of cesium iodide. Although there is some difference in the shows accumulation of one shot of cesium iodode with three shots ionization efficiencies from spot to spot, the important thing here of stachyose. is the relative amplitudes of cesium iodide to stachyose. If the Figure 1 demonstrates the calibration of one type of ion (low stachyose follows the cesium iodide, then the two shots of laser power MALDI ions of an oligosaccharide) with a vastly stachyose accumulate more ions in the cell than the cesium iodide, different type of ion (cesium iodide clusters formed from high while if the stachyose precedes the cesium iodide, the cesium laser power LDIswithout matrix). If these two molecules are iodide predominates. Similarly, as the fullerenes in Figure 2A had mixed together in solution and deposited on the target together, a much higher ionization efficiency than the cesium iodide, it was as is typically done with internal calibration, the selective ionization crucial to accumulate the fullerenes in the cell prior to accumula- yields quite different spectra (not shown) with no CsI ions tion of the cesium iodide to improve the trapping efficiency for apparent in an otherwise standard stachyose spectrum. Through the cesium iodide to genereate sufficient abundance for internal the three example spectra in Figure 1, the number of oligosac- calibration. This improved trapping efficiency effect is presumed charide ions were varied relative to the number of cesium iodide to depend on the phenomenon that trapped ions have Coulombic ions simply by changing the number of laser shots accumulated collisions with ions entering from the source and lose much of for each. With InCAS, one can control the relative abundance of their translational energy to improve their trapping efficiency. the calibrant to analyte without concern for selective ionization. Table 1 shows the mass accuracy of the spectra used in Figures Figure 2 shows two more examples of InCAS. The first 1 and 3. On average, the masses have an error of ∼0.1 ppm with example (Figure 2A) is accumulation of three shots of CsI clusters a standard deviation (σ) of 1.3 ppm. Table 2 shows the mass

(using maximum laser power) after one shot of the fullerene C70 accuracy of the spectra in Figure 2. In all of these measurements, (at half laser power). The second example (Figure 2B) shows low-intensity peaks (signal/noise <5) were rejected as the noise

Analytical Chemistry, Vol. 72, No. 24, December 15, 2000 5883 Table 1. Internal Calibration with Adjacent Spots with Cesium Iodide and the Oligosaccharide Stachyosea

Figure 1A Figures 1B and 3A Figure 1C Figure 3B average predicted observed error observed error observed error observed error observed error mass mass (ppm) mass (ppm) mass (ppm) mass (ppm) mass (ppm) 132.904 85 132.904 80 0.38* 132.904 90 -0.38* 132.904 85 0.00* 392.714 78 392.715 10 -0.81* 392.714 50 0.71* 392.714 80 -0.05* 392.714 60 0.46* 392.714 75 0.08* 527.158 25 527.158 90 -1.22 527.157 80 0.86 527.158 10 0.29 527.158 10 0.29 527.158 23 0.06 528.161 65 528.164 80 528.162 70 -1.98 528.162 10 -0.84 528.162 20 -1.03 528.162 95 -2.45 543.132 19 543.132 10 0.17 543.131 80 0.73 543.131 90 0.54 543.131 93 0.48 652.524 68 652.524 70 -0.03* 652.525 30 -0.95* 652.527 20 -3.86* 652.526 10 -2.18* 652.525 83 -1.75* 689.211 08 689.209 90 1.71 689.211 00 0.11 689.210 00 1.56 689.211 80 -1.05 689.210 68 0.58 690.214 48 690.214 50 -0.03 690.213 60 1.27 690.214 05 0.62 705.185 02 705.184 80 0.31 705.186 20 -1.68 705.185 20 -0.26 705.185 40 912.334 59 912.333 60 1.09* 912.332 80 1.96* 912.333 20 1.52* average 0.20 -0.35 -0.15 -0.14 -0.10 st dev 1.02 1.00 1.70 1.37 1.24

a Asterisks indicate calibrant peaks.

Table 2. Internal Calibration with Adjacent Spots with Fullerenes with Cesium Iodide (2A) and Peptides with Poly(propylene glycol) (2B)a

Figure 2A Figure 2B predicted observed ∆ predicted observed ∆ mass mass (ppm) mass mass (ppm) 392.71 478 392.714 75 0.08* 853.585 89 853.585 10 -0.93* 419.999 55 419.999 45 -0.23 911.627 76 911.627 30 -0.50 652.524 68 652.524 84 -0.25* 969.669 62 969.669 60 -0.02 719.999 45 719.999 33 0.17 1027.711 49 1027.712 00 0.50* 721.002 85 721.002 62 0.32 1085.753 35 1085.753 80 0.41 815.999 45 815.999 45 0.00 1143.795 22 1143.798 50 2.87 817.002 85 817.002 56 0.36 1201.837 08 1201.838 90 1.51* 839.999 45 839.999 27 0.22 1259.878 95 1259.882 00 2.42 841.002 85 841.002 62 0.28 1317.920 81 1317.914 80 -4.56 842.006 25 842.005 92 0.39 1318.924 21 1318.920 00 -3.19 843.009 65 843.009 22 0.51 1319.927 61 1319.921 60 -4.55 912.334 59 912.334 41 0.20* 1347.735 40 1347.743 50 6.01 1348.738 80 1348.739 90 0.82 1349.742 20 1349.740 40 -1.33 1375.962 68 1375.961 20 -1.07* 1434.004 54 1434.003 50 -0.73 average 0.17 -0.15 Figure 3. InCAS with cesium iodide and stachyose: (A) one shot st dev 0.24 2.70 of CsI followed by two shots of stachyose: (B) two shots of stachyose followed by one shot of CsI. a Asterisks indicate calibrant peaks. tended to distort the peak shapes to produce poor centroids that yield high errors (for example, the peak at m/z 528.1648 in Table 1 as a minor isotope had signal/noise ∼4 and -5.1 ppm mass The resolving power (M/∆M at fwhm) in all of these spectra error). As a particularly accurate example, the C70 radical cations ranged from ∼45 000 for the low-mass ions at m/z 132 to ∼4500 •+ are observed at m/z 839.999 27, whereas C70 should be at at m/z 1400. That this resolution is lower than might be anticipated 839.999 45. This is an error of 0.22 ppm. Also interesting is the for an FTMS spectrum is a result of the wide mass range selected. 2•2+ diradical cation (C70 ) at 419.999 55. The diradical cation There is a fixed depth to the transient digitizer and the capacity overlaps with the second harmonic of the cyclotron frequency of to Fourier transform a maximum of one megapoint in the array the singly charged ion (2ω+) except for the additional loss of an processor. All transient signals were well above noise level electron. The two peaks are clearly separated, as shown in the throughout the detection range. This range of resolving powers inset of Figure 2A. As a side issue, although separating these two increases the error in the mass assignment at the high-m/z region would in theory require significantly higher resolving power than of the spectrum, as is particularly evident with Figure 2B (Table indicated, the frequency shifts caused by the cell’s trapping field 2, column 3) where the substance P molecular ion at 1347.74 has and space charge (typically corrected for in the B term of the a mass error of 6 ppm despite being bracketed by calibrant poly- calibration equation) make the resonant frequencies of the ions (proplyene glycol) ions. This limitation can be overcome simply slightly nonlinear so that the 2ω+ harmonic is actually separated by increasing the memory depth of the transient digitizer and/or more than expected (∼6.5 Hz peak separation or ∼0.010 Da) from performing the Fourier transform in the pentium processor rather the 2+ diradical cation peak. Bruce et al. in the DeCAL calibration14 than on the array processor. Figure 2A demonstrates this quite have developed a calibration method that depends on this effect. nicely, as this spectrum was acquired using the full one megapoint

5884 Analytical Chemistry, Vol. 72, No. 24, December 15, 2000 acquisition memory available in the current IonSpec software; to the ions of one sample for calibration generates extremely accurate get adequately centroided peaks for the calibration, the data had masses on the spectrum. In this case, the mass accuracy on six to be processed in custom-built software to allow zerofilling prior different spectra internally calibrated in this way ranges from 0.2 to peak picking. The result, in this case, is dramatically better (σ ) 0.2 ppm) to -0.15 ppm (σ ) 2.7 ppm). This accuracy is mass accuracy (∼0.2 ppm σ compared to 2.7 ppm σ; see Table limited by the resolution, implying that much better mass accuracy 2). The memory limitation has already been partially addressed can be obtained by increasing the ability to acquire and transform by IonSpec, by increasing the memory depth to two megapoints longer transients. and implementing the Intel Signal Processing Library FFT functions in version 7.0 of their software, but the new software ACKNOWLEDGMENT was not used in this study due to other limitations that have not This research was supported by NIH Grant P41 RR10888 for yet been resolved. The two-megapoint depth to the transient the Boston University Mass Spectrometry Resource. The authors digitizer, though a significant improvement, will still be the limiting express thanks to E. Mirgorodskaya, D. Young, and M. McComb factor to the improvement of mass accuracy beyond this point. for assistance in sample preparation and to A. LaFleur and M. Strem for providing the fullerene samples. CONCLUSIONS A new and simple method for calibrating MALDI-FTMS spectra is demonstrated here. The method involves accumulation of ions Received for review July 5, 2000. Accepted September 19, from different samples together in the FTMS Penning trap prior 2000. to detection. Internal calibration of the spectrum based on using AC000770T

Analytical Chemistry, Vol. 72, No. 24, December 15, 2000 5885