A Gated Electrostatic Ion Trap To Repetitiously Measure the Charge

Anal. Chem. 1997, 69, 4162-4168

A Gated Electrostatic Ion Trap To Repetitiously Measure the Charge and m/z of Large Electrospray Ions W. Henry Benner

Human Genome Center Instrumentation Group, Engineering Science Department, Lawrence Berkeley National Laboratory, 1 Cyclotron Road MS-70A-3363, Berkeley, California 94720

The design and operation of a new type of ion trap provides a way to measure mass, charge, and velocity of large individual electrospray ions (>1 MDa, >250 charges) repeatedly during the time the ion is trapped. The trap consists of an image charge detection tube mounted between two ion mirrors. The mirrors are sets of parallel electrodes drilled with holes aligned with the bore of the detector tube. The device relies on voltages applied to the electrodes to establish symmetrically opposing ion focusing mirrors for the purpose of cycling ions through the charge detection tube many times. This design does not use magnetic or radio frequency fields to trap ions. Gating one of the mirrors to 0 V while maintaining voltages appropriate for reflecting ions on the opposite mirror allows an ion to pass through the holes in the mirror and enter the detector tube. A low-noise charge-sensitive amplifier, connected to the tube, reproduces the image charge of individual ions as they pass through the detector tube. When a highly charged electrospray ion enters the detector tube, its image charge triggers a circuit that enables the entrance electrodes, thus closing the electrostatic gate to the trap. Ion mass is calculated from simultaneous measurements of ion charge and velocity every time an ion passes through the detector. Individual ions have been trapped for as long as about 10 ms, during which time they cycled 450 times through the detector tube. At this level of trapping time, a theoretical precision for charge measurement as small as about two electrons rms can be achieved for 200 eV/charge ions carrying more than 250 charges. The operation of the system is demonstrated by trapping 2.88 MDa ions of DNA. Presently, the determination of the mass of electrospray ions larger than about 1 000 000 Da is possible with two mass spectrometry techniques. The first relies on Fourier transform ion cyclotron resonance (FTICR),1 and the second utilizes the simultaneous measurement of charge and time-of-flight.2 In the FTICR method, ions are injected into a trapping cell where the resonance condition defined by the magnetic and radio frequency fields definitively resolve the mass/charge (m/z) ratio of the trapped ions. FTICR has been available for many years and operates at very high m/z resolution. Bruce et al. com(1) Smith, R. D.; Cheng, X.; Bruce, J. E.; Hofstadler, S. A.; Anderson, G. A. Nature 1994, 369, 137-139. (2) Fuerstenau, S. D.; Benner, W. H. Rapid Commun. Mass Spectrom. 1995, 9, 1528-1538.

4162 Analytical Chemistry, Vol. 69, No. 20, October 15, 1997

mented3 that the high resolution achieved with FTICR suggests that the numerous m/z states for electrospray ions exceeding 1 MDa should be resolved. In practice, this goal is confounded by heterogeneity of the population of trapped ions. A number of additional factors addressed in a recent review article by Holliman et al.4 also degrade resolution in FTICR analysis. The mass analysis of large electrospray ions is, nevertheless, possible with FTICR. In a series of papers reported by the Smith group,5-8 a technique was developed for analyzing individual electrospray ions, thus avoiding heterogeneity with respect to the population of trapped ions. With this individual ion analysis technique, they studied the dimer of bovine albumin (133 kDa), poly(ethylene glycol) (5 MDa), and Coliphage T4 DNA (110 MDa). The serial trapping of individual plasmid DNA ions more recently has been used to acquire statistical information about the uniformity of ions from a sample.8 This approach is, evidently, difficult, as judged by the observation that only 62 ions were analyzed in order to acquire limited statistical information about the mass of the ions comprising the sample. Currently, the FTICR technique is not well suited for rapidly analyzing a large number of individual ions sequentially, as is required for determining the average mass of a population of megadalton ions in a sample. In addition to mass determinations, FTICR can be adapted to study the charge state of large electrospray ions and the influence of electrospray conditions on the amount of charge carried by electrospray ions. The determination of charge on individual electrospray ions is an important measurement that helped to elucidate some aspects of electrospray ion formation9 and the role solution parameters, most notably surface tension, dielectric constant, and conductivity, play in the electrospray charging process.10 Chen et al.5 used FTICR to extend the study of ion charge to the regime of large molecular ions and, by means of an experimentally determined ion ejection curve, measured the charge on megadalton ions with an accuracy of approximately 10%. (3) Bruce, J. E.; Cheng, X.; Bakhtiar, R.; Wu, Q.; Hofstadler, S. A.; Anderson, G. A.; Smith, R. D. J. Am. Chem. Soc. 1994, 116, 7839-7847. (4) Holliman, C. L.; Remple, D. L.; Gross, M. L. Mass Spectrom. Rev. 1994, 13, 105-132. (5) Chen, R.; Cheng, X.; Mitchell, D. W.; Hofstadler, S. A.; Wu, Q.; Rockwood, A. L.; Sherman, M. G.; Smith, R. D. Anal. Chem. 1995, 67, 1159-1163. (6) Chen, R.; Wu, Q.; Mitchell, D. W.; Hofstadler, S. A.; Rockwood, A. L.; Smith, R.D. Anal. Chem. 1994, 66, 3964-3969. (7) Cheng, X.; Bakhtair, R.; Orden, S. van; Smith, R. D. Anal. Chem. 1994, 66, 2084-2087. (8) Cheng, X.; Camp, D. C., II; Wu, Q.; Bakhtiar, R.; Springer, D. L.; Morris, B. J.; Bruce, J. E.; Anderson, G. A.; Edmonds, C. G.; Smith, R. D. Nucleic Acids Res. 1996, 24, 2183-2189. (9) Fenn, J. B. J. Am. Soc. Mass. Spectrom. 1993, 4, 524-535. (10) Cole, R. B.; Harrata, A. K. J. Am. Soc. Mass Spectrom. 1993, 4, 546-556. S0003-2700(97)00163-7 CCC: $14.00

© 1997 American Chemical Society

With much simpler instrumentation, we have determined the mass and charge of individual megadalton electrospray ions.2 This recent development uses a sensitive low-noise charge-sensitive amplifier to capture the image charge of individual ions as they pass through a metal detector tube. The transient image charge signal consists of a pulse with an approximate square-wave shape whose rise and fall correspond to ion entry and exit times in the tube. The amplitude of the resulting differentiated charge pulse is proportional to ion charge, and, by timing the flight of ions with known energy, ion m/z is determined simultaneously. Ion mass is calculated simply by multiplying m/z and z. Currently, we have a detector system that has, at best, a rootmean-square (rms) noise of 50 electrons. In our system, this is equivalent to a peak-to-peak noise signal of (130 electrons. An amplifier operating at this noise level can readily distinguish ions carrying at least 250 charges from baseline noise. Signals from ions with less charge can also be detected, but transients in the background signal interfere with timing and charge measurements. With this detector, we routinely detect and mass analyze positively charged DNA ions between 1.5 and 5 MDa.11 These ions typically have a m/z between 2500 and 4000. We have not determined an upper mass limit for our direct charge detection technique. The weight of polystyrene latex particles as large as about 100 nm in diameter, corresponding to a molecular mass of 3.3 × 108 Da, has been measured. The upper mass limit lies above this value. A primary advantage of our measurement approach is the rate highly charged individual ions are analyzed. In our previously reported instrumentation format,2 ions make a single pass through the tube detector. In that format, several thousand ions can be analyzed in a few minutes, thus supplying enough data for calculating statistically significant measurements of the mass of molecules in a sample population. The cost advantage of this technology, when compared to ICR, is also obvious because large magnets and ultrahigh vacuum are not needed. These two advantages of our approach are balanced, however, by the low precision of the single-pass charge detection approach. In practice, fairly accurate but imprecise mass measurements are obtained. In our one-pass format, the dominant cause of poor mass resolution observed for megadalton DNA is due to imprecision of the charge measurement. Ion velocity is measured more precisely than charge. An estimate of the relative errors associated with charge and velocity measurements were determined using an electronic pulser to generate charge signals that simulate DNA ions flying through the detector tube. The use of a pulser eliminates measurement variations caused by fluctuations of ion charge and velocity characteristic of large molecules. By introducing 10-µs-wide 0.5-mV pulses into the charge-sensitive preamplifier, as typically produced by transiting 3-MDa ions formed by positive mode electrospray, the relative standard deviation (n ) 100) of the charge measurement is 0.054, compared to a relative standard deviation of 0.013 for the velocity measurement. These values illustrate the relative importance of the charge determination in limiting the precision of the overall mass measurement. There are a limited number of options for improving charge measurement precision for the purpose of obtaining better mass (11) Schultz, J. C.; Hack, C.; Benner, W. H. Manuscript in preparation.

Figure 1. Waveform generated with a pulser simulating an ion passing through the detector tube. The vertical scale is 0.5 V/div, and the horizontal time scale is 5 µs/div. The upper trace corresponds to a single ion passing once through the detector tube and displays amplifier noise of 50 electrons rms. The lower trace results when 100 of these waveforms are summed and averaged. Averaging decreases noise to 5 electrons rms, thus improving signal-to-noise by 10-fold, and demonstrates the improvement that is gained when the charge on an ion is measured repeatedly.

measurements. The reduction of noise in the charge measurement circuit will not be very easy. With the current detector operating with a noise level of 50 electrons rms, further reduction in the noise level is constrained by fundamental limitations in the charge-sensitive circuitry. An approach that bypasses this limitation and provides a more substantial improvement in the precision and accuracy of the charge measurement is to remeasure the charge on individual ions. Assuming the source of electronic noise in the detector circuit is not correlated with the image charge signal, each additional measurement of ion charge reduces the noise associated with the measurement by a multiplication factor of (1/n)1/2, where n is the number of measurements that are averaged. The efficacy of signal averaging is shown in Figure 1. The upper trace is a wavelet, simulating the passage of an ion, generated by introducing approximately 10-µs-wide 0.5-mV test pulses onto a test capacitor connected to the input of the charge sensitive amplifier. A differentiating and shaping amplifier transforms the preamplifier signal into the bipolar pulse in Figure 1. The lower trace in Figure 1 results after 100 of these wavelets are averaged. The noise associated with the upper trace is 50 electrons rms, but, after 100 of these pulses are averaged, the noise fluctuations associated with the base line is reduced to 5 electrons rms. Several approaches might be used to remeasure the charge on an ion repetitively to benefit from signal averaging. A linear series of detectors would accomplish this goal, but, for this approach, each detector requires its own amplifier, and a series of 100 detector tubes is impractical if a 10-fold reduction in noise is targeted. This paper describes a gated electrostatic ion trap that recirculates ions through an image charge detector tube so that ion charge and velocity can be measured repetitiously, thus providing the opportunity for signal averaging. Analytical Chemistry, Vol. 69, No. 20, October 15, 1997

4163

Figure 2. Three-dimensional view, produced with Simion 6.0, of the potential valley created to trap ions. The thick black lines show the location of the electrodes and detector tube. The vertical axis of this plot is voltage. The detector tube and L1 (lens 1 in the trapping electrodes) are at ground potential; L2 ) -100 V, L3 ) 100 V, L4 ) 200 V, L5 ) L6 ) 300 V. The line extending along the centerline axis of the detector tube and part of the way up the potential valley shows the path followed by a trapped ion during one cycle through the detector tube.

EXPERIMENTAL SECTION We started with the premise that, if ions could be trapped efficiently between pairs of ion mirrors, then it might be possible to place our charge detector tube between the mirrors and introduce ions into the detector tube by momentarily gating one of the mirrors. Calculations performed with Simion 6.0 began by arbitrarily aiming an ion of known energy and m/z into an ion mirror defined by holes aligned in several parallel electrodes. The size of the holes, the spacing between the electrodes, and the voltages applied to them were adjusted until reflected trajectories were nearly parallel to incoming trajectories. Then, potential gradients associated with many different pairs of ion mirrors were examined to find the best conditions for trapping ions between the mirrors. The best-performing potential gradients looked like symmetrically opposed, rising potential valleys located at the ends of the detector tube. Figure 2 shows a typical 3D potential gradient that efficiently traps ions. After trapping conditions were established, additional Simion calculations showed that it is possible to gate one of the ion mirrors for the purpose of introducing ions into the trap. The Kingdon trap is another example of a pulsed electrostatic trap.12,13 The potential gradient in Figure 2 is produced with two sets of five electrodes attached to end caps mounted close to the ends of the detector tube. Each electrode in this model contains a centering hole through which ions travel. The lens numbering system progresses from 1, the end cap, to 6, the electrode farthest from the end cap. The detector tube is separated from the end caps by a small space. The detector tube is an essential part of the trap because, without it, it is impossible to know when to enable the entrance electrodes. The potential valley depicted in Figure 2 results when the following voltages are applied to the electrodes: L1 ) 0, L2 ) -100, L3 ) 100, L4 ) 200, L5 ) 300, L6 (12) Yang, L.; Church, D. A.; Weinberg, G.; Wang, Q. Nucl. Instrum. Methods B 1993, 73, 37-39. (13) Sekioka, T.; Terasawa, M.; Awaya,Y. Rad. Effects Defects Solids 1991, 117, 253-259.

4164 Analytical Chemistry, Vol. 69, No. 20, October 15, 1997

) 300. Setting L6 ) L5 and applying a negative potential on L2 creates the rising potential valley, and the negative potential on L2 additionally prevents the potential gradient from extending into the detector tube. Only ions in a specific range of energy are trapped by the potential valley in Figure 2. Higher energy ions or charged residue particles fly out of the valley and are not captured. Less energetic ions are not trapped because they roll off the potential saddle between lens L1 and lens L2. With the indicated voltages applied to the electrodes, Simion modeling indicated that ions with 215-230 eV/charge can be confined for several hundred passes through the detector, providing that they enter the trap within 1 mm of its centerline. Trapping times were found to be sensitive to ion m/z only in the sense that high-m/z ions travel relatively slowly and, therefore, have longer trapping times than lighter ions. Simulated trapping times for 1000 m/z ions slightly exceeded 3 ms, a limit imposed by computer RAM and not a consequence of instabilities associated with the trap design. The slope of the potential valley can be better comprehended by examining a plot of the potential along the centerline of the bore of the trapping electrodes as presented in Figure 3. The centerline potential controls ion velocity. As an ion exits the detector tube, it accelerates slightly until it passes through L2 and then rapidly decelerates as it climbs in the potential valley between L2 and L5. The height it attains in the potential valley depends on ion energy/charge. When the magnitude of ion energy/charge equals the magnitude of the local potential (compare, for example, 100 eV/charge with 100 V), the ion stops, turns around, and accelerates back down the potential valley. An identical potential valley awaits the ion at the opposite end of the detector tube, where the ion is forced to turn around again. The Simion modeling described above guided the construction of a gated electrostatic ion trap. It is shown in Figure 4. The detector tube (37.5 mm × 6.5 mm i.d.) is held axially in the bore of a metal block (3 cm diameter, 5 cm long) with two polyethylene

Figure 3. Grids representing the electrodes in the trap, drawn by Simion software. Juxtaposed is a plot of the potential along the center bore of the trap. As an ion travels from right to left, it begins at ground potential in the detector tube and accelerates until it passes L2, and then it decelerates in the increasing potential field. These conditions trap ions possessing about 200 eV/charge.

Figure 5. Photograph of the ion trap attached to a vacuum flange.

Figure 4. Diagram of the ion trap. Trapping plates on the left and right sides of the detector module define the potential field that forces ions to cycle back and forth through the detector tube. A support arm, attached to the bottom of the detector block, holds the detector assembly rigidly to minimize vibrations and shields an internal FET from rf noise.

disks that also provide electrical isolation. The metal block provides stability and electrical shielding. Pump-through ports in the polyethylene disks allow the entire assembly to be evacuated efficiently. Metal end caps on the block, designed with internal tubes that line up and face each end of the detector tube, provide additional shielding at the ends of the detector tube without adding significant stray capacitance. The oval pattern, drawn inside the detector tube, roughly reproduces a Simion-calculated ion trajectory. Its flight path extends beyond the ends of the detector tube and into the ion mirror, where its image charge is shielded from the detector tube. As an ion returns toward the detector tube, its image charge is, again, impressed on the detector tube. This design of the end caps and the way the electrodes are attached to the caps forces the image charge signal to appear abruptly, thus optimizing timing measurements. The pattern also shows that a trapped ion generally follows a path parallel to the axis of

the detector tube and only occasionally crosses the centerline. As a result, nearly identical image charge signals are generated for each transit through the detector tube. The ion mirrors were constructed from five square (5 cm × 5 cm, 0.05 cm thick) stainless steel plates separated with insulating spacers (0.2 cm long). Centering holes (0.5 cm diameter) were drilled in all of the electrodes, and small tabs on the edge of each plate provide locations for attaching power supply wires. The holes in the electrodes were aligned with the bore of the detector tube. A larger tube (4 cm diameter, 15 cm long) was attached perpendicularly to one of the longer sides of the metal block and serves as a pedestal for attaching the detector assembly (detector tube, trapping electrodes, and the shielding block) to a 6-in.diameter vacuum flange. Wires leading from electrical feedthroughs in the vacuum flange to the electrodes wrap around the outside of this support tube. A field-effect transistor (FET), along with its feedback resistor and capacitor, is located inside this supporting tube near the metal block. Wires leading from the FET to the rest of the preamplifier circuit, located outside the vacuum chamber, travel inside the support tube. The mounting structure was optimized to minimize stray capacitance associated with the detector tube and the wire connecting the detector tube to the FET. The structure’s rigidity minimized microphonic contributions to the background signal. A picture of the entire assembly is shown in Figure 5. Operation of the gated trap proceeds as follows: (1) Initially, all potentials applied to the electrodes on the entrance side of the Analytical Chemistry, Vol. 69, No. 20, October 15, 1997

4165

detector tube are maintained at ground, while the potentials on the opposing electrodes are set to predetermined values designed to reflect and focus ions of a selected energy toward the detector tube. (2) An electrospray interface directs ions through the entrance electrodes and into the detector tube. A detectable charge pulse from a single ion entering the detector tube triggers a circuit that enables the potentials on the entrance electrodes. These potentials are established in a time interval less than is required for the ion to return through the detector tube. (3) The potentials on each mirror are held constant over the lifetime of the trapped ion. (4) As an ion passes back and forth through the detector tube, the amplified and differentiated image charge pulses are digitally recorded for the duration of the trapping time. The resulting waveform consists of wavelets corresponding to single passes of an ion through the detector tube. (5) After a trapped ion has been lost, most likely through collision with the wall of the detector tube or one of the lens electrodes, the entrance plates are returned to 0 V, and the trap is rearmed to receive the next ion. The trap is also rearmed when an ion is detected but not trapped. By repeating these five steps, a large number of trapped ion waveforms can be accumulated with the current data system at a rate of several waveforms per second. Ion charge and velocity are determined for each wavelet in a waveform. Ion charge is deduced from the peak-to-peak voltage of a wavelet and a detector response factor of 425 electrons/V. The time between a positive pulse and an ensuing negative pulse corresponds to time-of-flight through the detector. We prefer this data analysis approach over Fourier transform analysis of the waveform because it avoids confusing transit time with the nearly equal time for an ion to turn around in the mirror. Mass is calculated from 2qV/(vm2 vg2), where q is ion charge, V is ion acceleration voltage (equal to the dc voltage applied to the ion guide), vm2 is measured ion velocity, and vg2 is the velocity component due to gas flowing through the electrospray source. Operation of the trap was tested with plasmid DNA ions. A 4.3-kilobase-long circular DNA molecule of a bacterial plasmid, described as pBR322, was supplied by Boehringer-Mannheim at a concentration of 250 µg/mL in 10 mM Tris-HCl, 1 mM EDTA, pH ) 8.0. The sample was not desalted. It was diluted 1/25 with 1:1 acetonitrile-water before it was electrosprayed at approximately 1 µL/min using an Analytica API ion source. pBR322 has a molecular mass of 2.88 MDa when all of its phosphate groups are catonized with sodium or a molecular mass of 2.69 MDa when it is in the H+ form. Additional electrospray source conditions include setting the cylinder voltage to -1300 V, end plate to -4000 V, and capillary entrance to -4400 V. The electrospray source was equipped with a factory-installed hexapole ion guide modified to operate at a rf peak-to-peak voltage of 700 V. It was operated in rf-only mode and electrically floated at 230 V to produce ions with 230 eV/charge. The drying gas was a countercurrent flow of approximately 2L/min of nitrogen, heated to 300 °C. An approximately 5-cm-long section near the middle of the glass transfer capillary in the electrospray source was wrapped with resistance wire and electrically heated with 5 W to desolvate the ions. RESULTS Figure 6 shows the waveform created by a single highly charged electrospray ion of pBR322 as it recirculated through the trap. The entire waveform composes wavelets corresponding to 4166 Analytical Chemistry, Vol. 69, No. 20, October 15, 1997

Figure 6. The lower oscillatory waveform describes the cycling of a 2.88 MDa DNA ion in the trap. For this waveform, the vertical scale is in volts, and the displayed trapping time is 1 ms. Pulse height provides a measure of ion charge, and the time between a positive peak and the ensuing negative peak is the time needed for an ion to fly through the detector tube. Ion mass is calculated each time an ion passes through detector tube and is plotted with open circles. The vertical scale for the mass data is megadaltons.

single passes of the ion through the detector tube. The time between a positive pulse and the ensuing negative pulse represents the time the ion spent in the detector tube, and the time between a negative pulse and the next positive pulse corresponds to the time it takes an ion to turn around in one of the ion mirrors. The shape of each wavelet is roughly the same because its shape does not depend on the direction it travels through the detector tube. This particular ion carried an average of 1040 charges, and the 1-ms record shows the ion recycled more than 51 times through the trap. The actual trapping time was longer than 1 ms, but only 1 ms of data is presented so that individual wavelets are easy to recognize in this figure. The open circles above the waveform indicate the mass values calculated for each of the 51 cycles through the detector tube. These values fall between 2.59 and 3.02 MDa, and, when averaged, the mean ( SD is 2.79 ( 0.09 MDa for this ion. The 95% confidence interval of the measurements is 0.01 MDa. The value of 2.79 MDa compares favorably to the expected mass of 2.88 MDa for pBR322 DNA in the sodium form. The difference between 2.79 MDa and the expected value of 2.88 MDa is likely due to the formation of an electrospray ion that is not completely sodiated. When ions from this same sample were analyzed with the one-pass method (not shown here), the resulting histogram of the mass of 1000 ions displayed a peak centered at 2.90 MDa. Its width spanned a mass range extending from the mass of pBR322 in H+ form (2.69 MDa) to ions completely sodiated. This width indicates that, even though pBR322 ions were electrosprayed from a sodium-containing buffer, they do not always form with a maximum number of sodium ions, as estimated from the total number of base pairs plus enough sodium ions to account for the charge the ion carries. The value of 2.79 MDa for the ion in Figure 6 clearly falls in the expected mass range.

The length of time an ion can be confined in this gated electrostatic trap determines the precision with which ion mass can be calculated. The time an ion is trapped depends on factors related to the trajectory an ion follows as it passes through the detector tube and turns around in the mirrors. The most stable trajectory results when an ion follows a radially centered path through the tube and turns around in the external trapping field without deviating from its centerline position. An ion following a centerline trajectory will remain confined in the trap until it is slowed by gas collisions or spontaneously fragments. Simion showed that ions entering more than 1 mm off the centerline or not traveling parallel to the axis acquire a slightly different trajectory each time they turn around in one of the mirrors and strike the electrodes or the tube after a relatively small number of passes. When a 2-mm-diameter aperture was placed in the hole of L6 to confine entering ions within 1 mm of the longitudinal axis of the detector tube, trapping times increased and the fraction of ions trapped improved also. The longest time an ion has been trapped so far is about 10 ms, during which time it oscillated 450 times through the detector tube. Trapping times this long suggest that charge measurements for individual ions could be as precise as the rms noise level of the detector (50 electrons rms) divided by 4501/2 or (2.3 electrons rms. Typical performance of the gated electrostatic ion trap, operated with the aperture, traps about one out of 10 entering ions. The fraction of ions trapped can be as large as 1/5 but can also be very small when the voltages applied to the ion mirrors are not adjusted properly. Several other factors influence efficiency and trapping time. Collisions of ions with background gas molecules reduce the energy of the ions and contribute to unstable ion trajectories. The presence of a gas jet flowing through the trap, created by the electrospray source, might be significant. The background gas pressure surrounding the trap was in the low 10-8 Torr range for this experiment. Ions were observed to lose velocity only when the pressure in the trap was above about 10-6 Torr. A few ions were observed to lose charge during the time they recirculated in the trap. A rough estimate is that one ion out of 104 loses charge while trapped. The appearance of waveforms produced by trapped pBR322 ions varies widely depending on ion m/z. Four different trapped ion waveforms are presented in Figure 7 to show typical variability. Figure 7a is a waveform of an ion trapped for nearly 10 ms. This ion carried 690 charges, and its mass was calculated to be 2.68 MDa, which corresponds to a pBR322 ion in H+ form. The statistical significance of the mass values we report here is insufficient to determine exactly the number of sodium ions attached to megadalton ions. It is possible that the value of 2.68 MDa indicates molecular heterogeneity associated with ions from this sample of pBR322, but it seems more reasonable to conclude that this ion carried very few sodium ions. A mass value for pBR322 smaller than 2.68 MDa could indicate the presence of DNA impurities in the sample or ion fragmentation. We did not see evidence for ion fragmentation occurring in the electrospray source. Figure 7b shows the waveform of a pBR322 ion carrying 2580 charges. Its mass is 6.6 MDa, a value much greater than expected for pBR322; 6.6 MDa is not close to an integer multiple of the molecular weight of pBR322, and, therefore, molecular agglomeration is an unlikely explanation for the weight of this ion. We see very large ions, such as this one, in all plasmid DNA samples we have analyzed. Commonly used plasmid purification procedures evidently do not completely remove bacterial genomic

Figure 7. Four trapped ion waveforms, showing charge and m/z variability of pBR322 electrospray ions. The upper waveform in each panel presents the entire waveform, and the lower panel shows an expanded view. Closing the electrostatic gate shortly after an ion enters the trap induces a large negative pulse at the beginning of each waveform. The x-axis is time, and the y-axis is volts. Charge is approximately 425 electrons/V. (a) An ion carrying 690 charges and trapped for 9.5 ms (upper trace, 1 V/div and 1 ms/div; lower trace, 1 V/div and 20 µs/div). (b) An ion carrying 2580 charges and trapped for about 5 ms (upper trace, 2 V/div and 0.5 ms/div; lower trace, 2 V/div and 20 µs/div). (c) An ion carrying 590 charges and trapped for 1.8 ms (upper trace, 1 V/div and 0.2 ms/div; lower trace, 1 V/div and 20 µs/div). (d) An ion carrying 270 charges and trapped for 3 ms (upper trace, 1 V/div and 1 ms/div; lower trace, 1 V/div and 20 µs/ div).

DNA from the plasmid preparations, and it is likely that this large ion is a piece of the genome of the bacterial host. Figure 7c shows the waveform of an ion carrying 590 charges. This ion is likely a portion of a pBR322 molecule because its mass is only 2.2 MDa. DNA samples always contain small amounts of sheared molecules. It is, therefore, difficult to determine the origin of this small molecule because it could be a contaminant or could have fragmented during transfer through the electrospray source. In Figure 7d, the waveform of an ion carrying 270 charges is characterized by a low S/N ratio. This charge level is close to the lower limit of charge quantification. Its mass is 1.6 MDa. Mass values, calculated from 549 trapped pBR322 ion waveforms, provide a way to calculate the average mass of pBR322 ions obtained from this sample. The 549 trapped ion waveforms were recorded in about 4 min, and an additional 2 min of computer time was needed to extract average mass values from the waveforms. Figure 8 shows a histogram of frequency vs ion mass for pBR322. A predominating peak at 2.88 MDa corresponds to sodium-adducted pBR322 ions, and a much smaller peak at 5.85 MDa corresponds to pBR322 dimer ions. The resolution (M/ ∆m) of the peak at 2.88 MDa is approximately 25 and is a factor of 5 improvement over the mass resolution obtained with the onepass method. An experimental resolution of 25 is less than instrumental resolution and results because the sample was not desalted. Instrumental resolution has not yet been satisfactorily determined because it depends on whether the digitized data are smoothed before the amplitude is measured. Smoothing the data improves resolution, but this might bias the data. Resolution is significantly larger than 25. At a resolution of 25 for plasmid DNA not desalted, the trapping technique rivals gel separations in terms Analytical Chemistry, Vol. 69, No. 20, October 15, 1997

4167

Figure 8. Histogram of the average mass obtained from each of 549 pBR322 trapped ion waveforms.

of resolution and greatly surpasses gel performance when analysis time is considered. For example, when we examined the purity of the plasmid DNA preparation by separating it on a 0.7% agarose gel, mass resolution varied among gels but generally fell between 20 and 30. Gel analysis time was 4 h. We chose to use DNA that was not desalted to demonstrate a capability for electrospraying large DNA ions without introducing tedious cleanup procedures as required in other mass spectrometry procedures. Using a salty buffer allows us to directly compare gel separations with mass spectrometric analyses. These results demonstrate that it is now possible to measure megadalton DNA rapidly with mass spectrometry. It should be noted that the mass measurement technique described here is amenable to direct calibration since it depends only upon the detector tube length, pulse height of the image signal, and ion transit time. The relationship between signal amplitude and induced charge is determined by depositing a known voltage on a 0.215-pF test capacitor connected to the input of the FET. Measurement of physical length of the detector tube is accurate to about one part in 500, but we know the effective tube length is different from the actual length. The effective tube length, more accurately termed the electric tube length, is the length value used to calculate ion velocity and is different from the physical length because of the way the image charge is captured by the detector tube. It is nearly 2% longer than the physical tube length.11 Pulse amplitude and ion transit time measurements are determined with a self-calibrating digitizer and are accurate to within a fraction of a percent. As noted earlier, the accuracy of the mass measurement is dominated by the charge-measurement accuracy. Now that ion charge can be measured with improved accuracy with the trapping technique, the relative inaccuracy of velocity and energy measurements will need to be reevaluated.

4168

Analytical Chemistry, Vol. 69, No. 20, October 15, 1997

CONCLUSIONS A gated electrostatic ion trap mass spectrometer provides a new measurement capability for electrospray ions exceeding about 1 MDa. Although the technique was demonstrated as a tool for determining the mass of plasmid DNA, several other applications could benefit from this technology. In any genomic sequencing project, it is necessary to clone chromosomal fragments and then order the longest possible clones into a minimally overlapping tiling path along the sequence of the chromosome. Assembly of a minimally overlapping tiling path is complicated because the cloning process introduces random lengths of genomic DNA into cloning vectors. Thus, searches are made to identify long clones with minimal overlap. It is relatively easy to determine if clones overlap, but it is much more time consuming to measure the length of overlap because the length determinations are made using gel electrophoresis. Therefore, the sizing step is often avoided. It is possible the mass spectrometry procedure we have developed could be used to identify overlap in clones rapidly. Another application relates to chromosome sorting. It is currently performed by spraying a solution of fluorescently tagged chromosomes. Specific chromosomes are identified by their fluorescence patterns, but the process is not perfect. Chromosomes might be identified more accurately by their masses. After the mass of a chromosome is determined, the chromosome could be released from the trap and deflected into a collection vial. Subcellular structures occur as noncovalent associations of component molecules and might be investigated with the gated electrostatic ion trap. Electrospray mass spectrometry has already been used to study noncovalent complexes such as DNA-protein interactions. It is possible that much larger complexes than these, which fall outside the range of many mass spectrometers, could be examined with the gated electrostatic ion trap. Finally, this technique operates over a mass range occupied by small microbes. It might possibly be used to identify virus particles, or other microbes, including bacterial spores. ACKNOWLEDGMENT This work was supported by the Director, Office of Energy Research, Office of Health and Environmental Research, Human Genome Program, of the U.S. Department of Energy under Contract No. DE-AC03-76SF00098. Dr. Stephen Fuerstenau is recognized for his work related to the development of the image charge detection scheme. William Searles contributed to the fabrication of the trapping cell, and Norman Madden developed the low-noise detector electronics. Dr. Joseph Jaklevic provided helpful editorial suggestions. Received for review February 10, 1997. Accepted August 4, 1997.X AC970163E X

Abstract published in Advance ACS Abstracts, September 15, 1997.