mass spectrometry - Analytical Chemistry

---

Richard D. Smith, Jon H. Wahl, David R. Goodlett', and Steven A. Hofstadler Chemical Methods and Separations Group Chemical Sciences Department PacificNorthwest Labratoy Richland, WA 99352 Many of the most difficult chemical, environmental, biochemical, and biomedical analytical problems reqnire a combination of instrumental attributes, including speed, low detection limits, wide linear dynamic range, good sensitivity, and high selectivity. For such demanding applications, the on-line combination of separation methods with MS often provides the most practical or perhaps the only approach. The orthogonal nature of the selectivities provided by a chromatographic or electrophoretic separation in conjunction with MS has long been considered attractive. Indeed, GC/MS is firmly established as a definitive analytical technique for many environmental and clinical analyses. The hallmarks of GClMS are its speed, selectivity, and sensitivity. Unfortunately, however, both GC and t h e conventional ionization methods used in MS (primarily electron impact and chemical ionization) reqnire sample volatilization. Thus, GClMS is not amenable to many analytica! problems without invoking often eomplex and problematic chemical degradation o r derivatization procedures designed to modify sample components to "GC-able" forms. Interest in LCIMS has continued to grow, and the technique has begun to open new avenues for the characterization of biological and biomedical samples (I). The 1980s saw the genesis and rapid development of a high-resolution separation method, capillary electrophoresis (CE), primarily because of the efforts of J. W. Jorgeneon of the University of North Carolina (2).He and his co-workers have

' Current address: Immunobiology Research Institute. Route 22 East, P.O. Bar 999,Annandale, NJ 08801 574 A

demonstrated that CE can generate both rapid and very high resolution separations, based on differences in t h e electrophoretic mobilities of charge-carrying species in a n electric field, in small-diameter fusedsilica capillaries. The advantages of the capillary format for electrophoresis are multifold. First, small-diameter capillaries (generally 50-100-pm id.) generate less Joule heat and dissipate this h e a t more effectively, allowing higher electric fields t h a n can be used with conventional electrophoresis and providing faster and higher resolution separations. Second, the capillary format allows for easy au-

REPORT tomation of sample handling and injection. The CE format allows ready implementation of a range of oncapillary detection methods; most effectively and broadly used are UV absorption and fluorescence emission detectors. Since the first commercial CE instruments appeared in the late 19808, CE technology and its applications have grown explosively. In fact, the rate of growth, use, and commercial implementation has wnsiderably exceeded that seen earlier for LC methods. Improvements in injection methods, detector Sensitivity, capillary surface deactivation, and coating technologies, as well as the introduc-

ANALMICAL CHEMISTRY, VOL. 65, NO. 13, JULY 1.1993

tion of new electmphoretic buffer systems, continue to drive further developments in CE for chemical, biological, and environmental applications. The growth of CE as a viable analytical tool is primarily the result of advances in detection methods and a n increasing recognition of its unique capabilities. CE would not be practical without the sensitivity improvements that have been demonstrated with on-capillary UV and fluorescence detection. Detectable amounts in t h e femtomole (lo-" mol) range can be obtained routinely, although optimized and specialized detection schemes have been reported for which detectable amounts extend to attomole (lo-'' mol) and zeptomole (lo-" mol) levels. Thus, for CE with typical capillary diameters, in which effective injection volumes are generally in the range of 1-10 nL, routinely detectable concentrations are typically on the order of lo-" M for the injected sample. Specialized detection systems allow these detectable concentrations to be extended to < lo-'' M, which is well into the regime of trace analysis. Improved detection limits can be obtained by using electrophoretic methods to concentrate sample wmponents during injection. The ability to manipulate and inject extremely small sample volumes, steps that are generally problematic with LC, provides a basis for using CE to confront extreme analytical challenges (e.g., the analysis of, or sampling from, single biological cells). In addition, CE has the flexibility provided by a range of formats (free-zone electrophoresis, electrokinetic micellar chromatography, isotachophoresis, gel electrophoresis, etc.) and a plethora of methods for manipulating injection conditions and separation specificity. Moreover, methods have been developed or are being investigated for CE application to the analysis of practically any substance that can be dissolved or suspended in a liquid. Finally, from a pragmatic viewpoint, the small sample, buffer, and waste volumes required and generated by CE are much less than those used by 0003-2700193/0365-574Ai$04.00/0

1993 American Chemical Society

LC methods; small volumes are attractive because of the trend toward treating all LC (and CE) effluents as hazardous wastes that require expensive tracking and disposal.

Development of CWMS The early work of Jorgenson and coworkers clearly demonstrated the potential power of CE methods, but combining them with MS required the solution of several conceptual and practical problems. First, all early CE experiments, and nearly all today, involve placement of both ends of the capillary into reservoirs of the conductive buffer, where electrical contact is established to defme the CE field gradient. Second, any CE/MS interface must be compatible with low CE flow rates (5 1 FL/min a t most) and should not induce a pressure-driven (laminar) flow in the capillary t h a t would degrade separation quality. Moreover, detection sensitivities must extend to subpicomole levels for CElMS to be of practical value. The early 1980s saw the introduction of the thermospray interface for LClMS ( I ) , but this approach was impractical for CE because of the necessary flow rates (z 100 pL/min) and inadequate sensitivity. In 1984 Fenn and co-workers presented their initial results on electrospray ionization (ESI) combined with MS (3).In the ESIMS method, liquid solutions are nebulized in a high electric field from the end of a capillary a t flow rates in the 1-10-pLlmin range (4-6). More important, the early ESIMS studies suggested that solute sensitivity was exceptional (4). This work stimulated efforts a t Pacific Northwest Laboratory (PNL) to develop on-line CElMS based on ESI with t h e use of a n electrophoretic capillary, one end of which functioned as the electrospray source rather than being immersed in a buffer reservoir. After constructing ESIMS instrumentation at PNL, results were obtained in 1986 and published in 1987 (7).

a capillary to produce highly charged

ure 2a). With this interfacing method it was necessary to select CE conditions giving rise t o a net electroosmotic flow in the direction of the mass spectrometer, that is, a flow arising at the electric double layer of the capillary surface and creating a flat "plug-like" flow profile away from the point of injection. The metallized capillary terminus also served as the electrospray source by having a 3-6-kV difference in voltage between the terminus and the mass spectrometer sampling aperture, which was 1-2 cm away. In ESI the liquid is nebulized from

droplets, typically only a few micrometers in diameter. The charged droplets drift in the electric field between t h e capillary and t h e mass spectrometer sampling aperture. In transit to the sampling orifice and in transport through the MS interface, the droplets experience conditions that cause evaporation. Because the droplets initially are highly charged and close to the physical limit for their size (the Rayleigh limit), they shrink by evaporation and quickly reach a point at which they shed a portion of their charge. By some still

I Nozzle

High-voltage sample chamber

*

Pump

*

Pum

- ~.

.,.~ ,+! 20-50%) and that very large nonvolatile molecular species are amenable to it (6).In 1987 Fenn and co-workers demonstrated that the ESI of large polymers produced ions with a net charge (2) that increases with molecular size; therefore, conventional mass-to-charge ( m / z ) range instruments could be used t o detect species with molecular weights many times greater than the maximum m/z of the instrument (8). The next year Fenn and co-workers demonstrated that precise molecular weights could also be readily determined for proteins from the distribution of charge states observed across the mass (or more correctly, m/z) spectrum (9). Since then the range of applications of ESIMS has grown very rapidly and is facilitated by the current availability of instrumentation from nearly all MS vendors. The combination of LC with ESIMS is becoming widely accepted because of the broader use of LC and arguably greater ease of interfacing. As the use of CE grows and inevitably displaces LC in some applications, we anticipate that the role of CE/MS will also grow. In some areas, particularly in biological and biomedical research where sample size i s inherently limited or i n creased only a t great expense and effort, the high resolution and sensitivity of CElMS compared with LClMS will ensure a significant analytical role. Realization of the full potential of CE/MS hinges on interface performance and MS detection sensitivity, which we now discuss in greater detail.

Interface methods and performance The initial performance of CE/MS with the metallized capillary terminus design was limited because flow rates were below optimum for the electrospray design used a t the time ( 7 ) and because of frequent spray instabilities. Imposed restrictions on the types of CE buffers effectively precluded the use of aqueous and higher conductivity buffers. In addition, the lifetime of the metallized capillary was limited to only a few days of operation, presumably as a

result of electrochemical processes. Initial studies nevertheless demonstrated CE/MS detection limits for quaternary ammonium salts that extended to subfemtomole levels (10). The ease of interfacing was dramatically improved in 1988 by introduction of the sheath flow design, which removed many of these limitations (11). In this design, a small coaxial flow (1-5 pL/min) of liquid serves to establish the electrical contact with the CE effluent and to facilitate the electrospraying of buffers that could not be directly electrosprayed with earlier interfaces. Figure 2b shows a schematic illustration of a more recent implementation of the sheath flow design that is distinguished by an etched conical tip of the CE capillary. This design serves to enhance the electric field gradient a t the capillary terminus, minimize the effective mixing volume between the sheath liquid and the CE effluent, and increase the stability of the electrospray process. Another variation on the CElMS interface, introduced by Henion and co-workers (IZ),uses a liquid junction to establish electrical contact with the analytical capillary and to provide an additional makeup flow of buffer (Figure 2c). The relative advantages of these designs have been compared ( 1 3 ) , and the coaxial sheath flow interface appears t o

have several advantages. A disadvantage of both approaches, however, is that they depend on an additional flow of liquid that incorporates charge-carrying species (e& . ,!. buffer or solvent impurities) and invariably degrades detection sensitivity. There continues t o be an interest in the development of a more versatile and reliable design that does not depend on an additional liquid flow. We are pursuing several sheathless designs for this purpose, such as that illustrated in Figure 2d, and we show some initial results for one design later. Jorgenson and w-workers are pursuing a somewhat similar microspray approach for LClMS interfacing (14) in which the ion source operates under a vacuum in a manner similar to electrohydrodynamic ion sources for MS. Early generations of commercial CE instrumentation made MS interfacing a difficult task a t best. The design of future commercial instruments will determine the rate a t which CE/MS becomes routinely available. Although CE/MS capabilities exist in a number of laboratories, greater sensitivity (or the generally related issues of better MS scan speed and resolution) is desired.

Sansitivitt The small sample sizes associated with CE.create demands for the best

".I""

Figure 3. Dependence of MS signal intensity on analyte mass flow rate to the electrospray source for relative analyte/background electrolyte flow rates of A. N10, and NIOO. Behavior predicfed using Equation 1; it is expected to be qualitativelycorreCt when the CE current exceeds me ESI current.

576 A * ANALYTICAL CHEMISTRY,VOL. 65, NO. 13, JULY 1,1993

possible detection sensitivity. Generally, CE is not well suited for trace analyses and is probably inappropriate for such applications when large amounta of sample are available and LC separations are adequate. CE is a t its best, however, in samplelimited situations and may constitute the only practical approach in some cases (e.g., sampling from single biological cells). Recent work has demonstrated that CEIMS is capable of providing exceptional sensitivity in certain situations (15).It is essential, however, that consideration of CE/MS application involve a practical understanding of ESIMS performance and sensitivity issues. The response characteristics of a particular detector are referred to as either "concentration sensitive" or "mass flow sensitive." For example, a UV detector functions as a concentration-sensitive detector. Higher flow rates will decrease peak widths but will not increase peak heights; consequently, the peak area is inversely related to the flow rate. Conversely, the conventional electron ionization method used for GCIMS functions as a mass flow-sensitive detector. Higher flow rates will decrease peak widths but also increase peak heights; thus, peak areas are independent of flow rate.

Popular discussions of ESIMS have labeled i t a concentrationsensitive detection method. At the typical sample flow rates used for ESIMS, and thus for LCIMS, this is generally correct. However, because of the small sample size and flow rates associated with CE/MS, it is possible for ESIMS to function in either a concentration-sensitive or a mass flow-sensitive manner. The distinction between these two detection regimes is essential for understanding the CE/MS sensitivity gains that arise when capillary diameters or electromigration rates are reduced. Under most conditions used for ESIMS, the number of charged species in solution delivered to the electrospray source is much greater than the number the electrospray process can transfer to the gas phase as detectable ions. This situation applies for typical CE capillary diameters and buffers in which CE currents range from 5 t o 100 PA. In contrast, ESI currents for aqueous solutions range from 0.1 to 0.5 pA.Thus, even if the entire electrospray current could be converted to desolvated ions in the gas phase, the overall efficiency of the ionization process would be only on the order of 1%. Details of the ion formation pro-

cess and any discrimination in ionization efficiency between buffer components (Le., the background electrolyte that substantially exceeds the analyte in concentration) and analyte species would obviously affect the overall efficiency. The important point, however, and one that is qualitatively consistent with the current phenomenological understanding of ESIMS (5, 6),is that discrimination processes do not substantially alter the expectation that the ionization efficiencv will be low under such conditions. The aualitative features relevant to s e n s i b t y can be described with a simple model. In an adaptation of the approach of Tang and Kebarle (16), the analyte signal intensity, I(A+),can be expressed as a function of the mass flow rate of the analyte, V,(A*), and background electrolyte, VXB+),as

-

where I is the total electrospray murent, P is a variable related to the sampling efficiency of the ESIMS system (assumed constant), and f is a constant representing the fraction of droplet charge converted into gasphase ions. For simplicity, we as-

Flgun 4. Comparison of normal constant field strength (top) and reduced elution s p e d (bottom) CUESIMS analysis of a tryptic digest of bovine serum albumin. The reduced elution speed CElESlMS analysis was mnduued at 300 Vlcm until 1 min before elution of the first analyte. when the electric fieid strength was reduced to M) Vlom. Single-im eiectmphemgramsconespandingto several of the 1rVptic (palypplide) hagmenta are shown on the rQht (designatedLV u68 of the single-lener code for amim ackl residues). (Adapted wilh permission fmm Reference 17.)

ANALYTICAL CHEMISTRY, VOL. 65, NO. 13, JULY I , 1993

-

577 A

REPORT sume no discrimination in the ionization efficiency for A' and . ' B Most relevant t o CE is the limit at which the background electrolyte (buffer) mass flow rate is assumed constant and much greater than the analyte mass flow rate. Here Equation 1 reduces to

The analyte signal intensity in this high-background electrolyte concentration regime is predicted to be directly proportional to the analyte concentration. Moreover, the analyte sensitivity is predicted to be i n versely proportional t o the background electrolyte mass flow rate. Analyte sensitivity can thus be improved by decreasing the background electrolyte mass flow rate. The importance of this observation to detection sensitivity is illustrated in Figure 3, where the relative analyte signal intensity is shown as a function of the analyte mass flow rate. For curve A, the maximum and minimum analyte mass flow rates cover a range from 3 orders of magnitude less that of the background mass flow rate up t o a mass flow rate equivalent to that of the background electrolyte. According to Equations 1or 2, an increase in analyte sensitivity is predicted when the mass flow rate of the background electrolyte is decreased. This i n crease in sensitivity is illustrated in Figure 3 where the middle curve (A/ 10) represents the signal intensity predicted when both the analyte and background mass flow rates are reduced by a factor of 10 relative t o curve A, and the upper curve (A/100) corresponds t o reduction of mass flow rates by a factor of 100 for both flow rates. At low analyte concentrations, the analyte signal intensity is predicted to be 2 orders of magnitude greater when both the analyte and the background ions are reduced by 2 orders of magnitude compared with a decrease only in the analyte mass flow rate (curve A). Thus, a large increase in sensitivity is predicted. This increaee arises directly from the decrease in background electrolyte and is derived from the greater fraction of the analyte converted into gasphase ions. The situation described by Equations l o r 2 clearly fails a t sufficiently low-background electrolyte flows. Equation 2 is not realistic; it predicts that the ion signal intensity will remain constant as the analyte

and background electrolyte mass flow rates are decreased in an indefinite manner. Consequently, a second detection regime must occur when charge-carrying species are no longer supplied to the ESI source at a rate sufficient to sustain the maximum electrospray current (Le., when the CE current becomes comparable to or less than the maximum electrospray current), and the ionization efficiency for this regime is Z(A+)= ZPf V,(A')

(3)

Both Equations 2 and 3 predict that the observed MS signal will be proportional to analyte concentration; however, one important and somewhat subtle distinction exists. For Equation 2, a change in background electrolyte mass flow rate will affect analyte signal intensity. I n t h e course of a given CE separation, V,JB+) is generally eonstant, and the signal intensity will directly reflect electrolyte mass flow rate. In this regime the ESIMS detector appears to function as a mass flow-sensitive detector. The transition between the regimes described by Equations 2 and 3 is evident when further decreases in the background electrolyte do not lead to additional gain in analyte sensitivity and is predicted to occur when the CE current is less than the normal ESI current (17-19). These considerations indicate that analyte sensitivity in CElESIMS may be increased by reducing the mass flow rate of the background components. Separations by CE are generally performed under conditions in which the analyte concentration is much less than the buffer concentration. The lower limits t o CE current with larger capillary diameters are generally defined by trace ionic contaminants; low-conductivity liquids are thus generally impractical for CE. However, the mass flow rate of electrolyte from the analytical capillary can be reduced in several other ways, including reduction of the CE electric field strength or migration rate of electrolyte into the electrospray source, or the use of small i.d. capillaries, which reduces the mass flow rate of both the analyte and the background ions. Experimental studies have recently demonstrated that both approaches are viable and lead t o the anticipated advantages for CE/MS detection (17-19).

Reduced elutlon speed CE/MS The small solute quantities used in CE and the low signal intensities

578 A * ANALYTICAL CHEMISTRY, VOL. 65. NO. 13, JULY 1.1993

generally produced by ESIMS typically result in maximum analyte ion MS detection rates no greater than 106-106 ion counts per second with current quadrupole instrumentation (5).Statistical considerations, therefore, effectively limit the maximum practical scan speeds with quadrupole mass spectrometers. In addition, MS resolution is often sacrificed because it generally comes at the expense of sensitivity. Thus, depending on the m/z range to be examined, solute concentrations, and other factors related to the nature of the solute and buffer species, the maximum practical MS scan speeds are often insufficient to exploit the speed or high-quality separations feasible with CE. The reduced elution speed (RES) method for CE/MS circumvents, to some extent, both the sensitivity and t h e speed limitations of scanning mass spectrometers (1 Involving only step changes in the CE electric field strength, the technique is simple and readily implemented. Before elution of the first analyte of interest into t h e ESI source, the electrophoretic voltage is decreased and electromigration rates are slowed,

n.

L'

00 2.0 4.0 6.0 8.0 10.0

Time (min)

Figure 5. Reconstructed ion current electropherograms obtained for the peptide mixture containing tryptophan (204 Da), leucine-enkephalin (555 Da), and meliftin (2845 Oa), using capillaries of loo-, 50-, 20-,and 10-pm i.d., and displayed on the same absolute intensity scale. The relative amountsof sample injected are proportional to the capillary cross-smional area (i.9.. the amount injected for the IO-wm-i.d. capillary is 2 orders of magnitudesmaller than lor the 1W-pm-i.d. capillary). The gains in ssnsilidily for the 10-pm-i.d. capillary correspand to factors Of 25 to 50.

allowing more or longer MS scans to be acquired. Under conditions i n which the amount of solute entering the ESI source per unit of time exceeds its ionization capacity, the earlier discussion suggests that no substantial decrease in maximum ion intensity will result. The RES method is illustrated in Figure 4, which compares normal and RES total ion electropherograms from CElMS analyses of a tryptic digest of bovine serum albumin (a protein with a mass of 66.430 Da). In each separation a total of 40 fmol of protein was used, producing a separation that yielded more than 100 overlapping peaks for various peptide digestion products. When the RES separation method was used to slow the elution by a fador of 5, only a 20% decrease resulted in maximum ion intensities. Figure 4 shows four of the more than 100 ueaks attributable to tmtic polypeptides observed during ihe RES CElMS experiment. The RES technique allows-more components to be detected and effectively reduces

-

01 0.0 12.0

the complexity of individual mass spectra or the likelihood of missing some components ( a complication that arises because analyte zones for conventional separations are often shorter or comparable to t h e MS scan time). Similar advantages have also been demonstrated for protein mixtures for which high-quality mass spectra can be obtained with injections of only 60 fmol of each protein (17). RES CE/MS provides an increase in the effectiveness of mass suectrometric scanning compared wiih conventional CE/MS methods. The method does not increase solute consumption; does not degrade detection sensitivities for peptide and protein analyses extending into the lowfemtomole regime; and incurs very little loss in ion intensity, which is particularly important for tandem MS methods and their potential application to peptide sequencing. Perhaps more important, RES illustrates a flexibility available with C E separations may be slowed or even stopped instantaneously with little

Ill i

14.0 16.0

18.0

20.0

30 20

10 m/z 589

0

Leu-ENK

10.0

12.0

14.0

16.0

18.0 20.0

(5-10amol?) ? I''

604

2.5

Saa

Figure 6. Comparison of CUMS separations for a mixture of pentapeptides in which the amount of sample injected was varied by more than 2 orders of magnitude. The separations used a bare fused 20-pm-I d capillaw ard a 10-mM acetic acid buffer with a sheath now intedac8

loss of separation efficiency (18)-a feature not readily achieved with LC. It is anticipated that this capability could be particularly useful in conjunction with t h e advanced MS methods that we will discuss later in this article. CUMS wlth small-diameter caplllarles When greater sensitivity is desired or sample volume is extremely limited, smaller capillary diameters are advantageous. For a given buffer and CE electric field strength, the CE current has a quadratic dependence on the capillary diameter. The simple model for ESI sensitivity described earlier leads one to expect that the best sensitivity will be obtained when the CE current is less than the ESI current. Even with low-conductivity 10-mM acetic acid buffer systems often used for CE/ MS, this happens only for capillary diameters of < 30-40 m,somewhat smaller than those typically used for CE. Because of complexities introduced by the sheath flow, it was unczrtain whether the predicted sensitivity gains could be realized. To examine this question, we compared CElMS separations for capillaries ranging from 100- to 5 - p i.d. ~ (19). Figure 5 shows CE/MS seleded ion electropherograms for a simple mixture plotted on the same absolute intensity scale obtained using capillaries of loo-, 50-, 20-, and 10-pm diameters. The electrokinetic injections result in sample sizes proportional to the capillary cross-sectional area, so that 2 orders of magnitude more sample was injected into the 100-fim capillary (- 800 fmol per component) than into the 10-pn capillary (- 8 fmol per component). The absolute signal intensities were found to decrease by factors of only 2-4, rather t h a n t h e 2 orders of magnitude that might be expected, which corresponds to a sensitivity increase by factors of 25-50 for the smallest capillary diameter. Similar experiments with protein mixtures, in which MS signal intensities are generally lower, showed that useful mass spectra could be obtained from subfemtomole injections using 5-pm-i.d. capillaries (15).A detailed study of the role of sample concentration and capillary diameter on CE/MS sensitivity has been published (19).

Appllcatlons Currently a t least a dozen laboratories are actively involved in developing and applying CE/MS techniques

ANALYTICAL CHEMISTRY, VOL. 65, NO. 13, JULY 1,1993

579 A

and methods, based on published work. In addition to those from our laboratory, numerous contributions and applications have been demonstrated by Henion and eo-workers (20-23); Moseley, Tomer, Jorgenson, and their colleagues (24, 25); and Thibault et al. (26-28). Although our focus h e r e is on CEIMS using ESI interfaces, other methods have been reported. In particular, interfaces based on continuous-flow fast atom bombardment have been developed by Tomer, Caprioli, Reinhold, and their respective eo-workers (29-32), as well as others. Wilkins et al. (33)recently dem-

onstrated the off-line combination of CE with high-resolution Fourier transform ion cyclotron resonance (FTICR) MS based on matrix-assisted laser desorption ionization. Currently, the ESIMS interface appears to provide the best approach to on-line CEIMS, although these alternatives may have advantages for specific applications. The range of potential CE/MS applications is far too broad to cover in this article, and it substantially mirrors that for CE. Applications to date include the analysis of oligonucleotides, DNA adducts, peptides, proteins, carbohydrates, pesticides, var-

.. --._...--.-. _-..._

Flgure 7. Comparison of three total ion electropherograi separations of tryptic digests of cytochrome c species.

The separaion us@ a l0ym-i.d. capillav and a new shealhles CEMS inlerfsce. Mass spectra are shown lor represenlalive regions and generally indicate elution of more fhan one wmwnem.

580 A

-

ANALYTICAL CHEMISTRY, VOL. 65, NO. 13, JULY 1,1993

ious metabolites, and small organic and inorganic ions. The emphasis here is on free-solution-zone eledrophoresis, the most widely used CE format. Other CE formats have been interfaced t o MS. C a p i l l a r y i s o t a chophoresis/MS (CITPIMS), developed by U d s e t h e t al. i n o u r laboratory (34), can be a powerful tool in some applications because the background electrolyte is effectively eliminated during analyte detection. As a result, greater signal intensities and improved detection limits can be obtained with use of CITPIMS. In addition, Henion and Garcia (35) demonstrated the combination of capillary gel electrophoresis with pneumatically assisted ESIMS. These developments emphasize the flexibility of ESIMS interfacing methods for CE. Toward ultrahigh-sensitlvlty CEIMS An obvious approach to improving CEIMS detection limits is to modify the concentration during the injection step. Such techniques have been demonstrated for CEIMS; a particularly useful approach involving isotachophoretic sample preconcentration has recently been reported by Tinke e t al. (36) and by Karger and coworkers (37).However, when sample size is limited or such techniques are impractical because of degradation of separation quality, detector sensitivity becomes the crucial issue. Our understanding of ESIMS suggests that sensitivity can be optimized by using small-diameter capESIMS interfaces illaries (< 20 p). allow a wide range of CE buffers to be electrosprayed successfully, providing considerable flexibility despite some constraints (e.g., high salt and surfactant concentrations remain problematic). Aqueous and mixed aqueousIorganic buffers of a t least 0.1 M can be used for CEIMS when dilution occurs by the sheath liquid or liquid-junction buffer. Improved electrospray source designs and the coaxial flow of electron scavengers (e.g., SF,) also serve to extend the range of compatible liquids. Because of practical sensitivity constraints, however, CE buffer concentrations are generally minimized. For the low-conductivity buffers and sheath liquids generally used for CEI MS, 10- to 20-pm capillaries provide good sensitivity while avoiding most of the difficulties associated with smaller capillary diameters, such as low permeability (which makes surface treatment and column flushing

;i!L

difficult) and greater susceptibility to plugging. The major considerations relevant to MS detection are the nature and complexity of the particular sample and pragmatic compromises between detection sensitivity, resolution, and scan speed. For quadrupole mass spectrometers, selected ion monitoring (SIM) yields significantly enhanced detection limits compared with scanning MS operation because of the greater dwell time for signal acquisition a t each selected m / z value. For samples in which analyte molecular weights are known and m h values can be predicted, SIM detection is an obvious choice. If sufficient sample is available, direct infusion can be used to produce a mass spectrum of the unseparated mixture, and the results can be used to guide selection of specific m/z values for SIM detection. The sensitivity gain of SIM detection versus scanning detection with conventional quadrupole MS instruments can be substantial. Figure 6

I =$

at pH 3.4

-peptide’

shows three selected ion electropherograms for injections from serial dilution of a simple pentapeptide mixture. For the 5-IO-fmol injection (top), SIM detection provides excellent signal-to-noise ratios ( S I N ) , and the CE performance and solute zone profile is captured by the MS detector. For injection sizes corresponding to 0.5-1.0 fmol (middle), increased chemical noise attributable to buffer impurities is evident. Injections of 40-75 amol (bottom) still give good response, with estimated detection limits of 10 amol. The chemical noise background for this separation results from “real” signals arising from solution components. Thus, higher purity buffers may provide a basis for an extension to subattomole detection limits. An advantage of MS relative to other detectors is its high specificity. Obtaining the maximum number of theoretical plates possible with CEIMS is rarely required unless closely related mixture components have similar molecular weights. Analyte migration times can shift between runs hecause of (sometimes unavoidable) capillary surface modification and degradation. With less specific detectors, great care is generally necessary to establish reproducible migration times to identify eluents; with CEIMS, approximate migration times are usually sufficient, and one is generally more concerned with MS sensitivity and resolution. A sheathless CWMS Interface Although the coaxial sheath flow interface has facilitated progress in

l

510

Time (win)

CEIMS, i t does have some drawbacks. It contributes electrolyte to the ESI source t h a t can decrease sensitivity (but less than might be expected, presumably because of poor mixing with the CE effluent) and gives rise to chemical noise. Our laboratory has revisited the initial sheathless CEIMS interface a p proach (7) with use of a metallized contoured capillary terminus (38), such as that illustrated in Figure 2d. Figure 7 shows separations of tryptic digests obtained for three cytochrome c proteins (bovine, Candido krusei, and equine, with molecular masses of 12 ma).Approximately 30 fmol of the starting protein (before digestion) was consumed for each separation. The separations were conducted in a 50-em x IO-pmi.d. capillary with its inner surface chemically modified with 3-aminopropyltrimethoxysilane, using a 10mM aqueous ammonium acetate1 acetic acid buffer a t pH 4.4. By scanning rapidly (0.6 s per scan), mass spectra could be acquired for each separation in < 6 min, allowing each of the tryptic fragments, and other mixture components attributable to incomplete digestion, to be identified (38).The routine implementation of such capabilities can have enormous impact for more rapid and sensitive protein mapping and, when used in conjunction with tandem MS, for protein sequencing.

-

The future Recent work with CEIMS using quadrupole mass spectrometers has demonstrated the potential for ex-

-

CUMS at DH8.0

Table 1. Performance characteristics for mass spectrometers with potentfai for CWMS MUS sneclronuhr

IonuUllmlon Sp.nNs *Ikkm*

Quadrupoleb 10-~-10-* Orlhogonal

1300

1500

1700

..jure 8. Applicatior. _.J / M S to the study of ribonuclease S,a noncovalent complex consisting of an A-protein and a smaller A-peptide. The CE sparation in an acetic acid M e r (pH 3.4) gives two peaks with mads spectra cwresponding lo the d m i a t e d A-protein and A-peptide spebes. The sparation in an ammonium bicarbonatebulfer (pH 8.0) gives one broad peak. The mads spectrum W i n e d during eiuiion of this peak is dominated by the intact ribonuciew A-complex. aithough a smaller contributionanributable to the A-protsin toned by ds-iation of the comprex in the ESI intedaca 10 460 O b s e N e d .

time-of-

0.03-0.8

Rnolullon W”

t

103

IO4

200-1000

1-5 0.05-0.5c

104-105

n-2 n.1

Commaslal avallabllW‘

Cod iS KP

1993(?)

zoo-400 150-200

1994(?) 1993 (?)

200-300 400-500

NOW

flightb

ITMSb FTlCR

0.1-0.5 0.03-0.3c

I@@n - l e n>2

*Estimate based on scanning operation or duty cycle. Based on the efiiciency with which a mtinuow ion wnent contributesto useful signd but does not consider trappino efficiency ol ion trap mass rpectmmatry (ITMS) or FTlCR instruments. Estimate range of perionanca arises because of tradeoils berween resolution and 6can speed

eration spectra require restarting from new precursor ion populations fw each ste ‘Based on estimated availability 01 €SI interfaces. but im iementation may SU&! from low ion sampiirqor ion use Nicien and other lim+3tions that hi& CUMS operation (indicatedby “P’). 0 Approximate range, but wqvary mdely ma commercialimprementation and Wions.

ANALYTICAL CHEMISTRY, VOL. 65, NO. 13, JULY 1,1993

581 A

REPORT traordinary sensitivity and the capability to accommodate extremely small sample sizes. For example, sample volumes with 5- to 10-pmi.d. capillaries are generally on the order of 10 pL. Handling of femtomole-sized samples is generally impractical, and integrated or on-line methods of sample handling are essential. Microcolumn methods for protein digestion (using immobilized enzymes), combined with other frontend sample workup strategies, must be optimized for specific applications. It is anticipated that such CE methods and ancillary “picoscale” manipulation techniques will provide an essential toolbox for biochemical analysis on the femtomole and subfemtomole levels. The opportunities for biochemical research are enormous but extremely demanding. For example, the capability for analysis at the single-cell level could provide insights into cellular chemical processes without the necessity of averaging over large cell

populations. Novel studies of DNA damage products, protein production, and important post-translational protein modifications would become possible. Such studies will require MS detection sensitivities extending to the zeptomole range. The demands imposed by materials such as glycoproteins are essentially open-ended because of their possible structural complexity (i.e., heterogeneity arising primarily from the structural complexity of carbohydrates and further convoluted by the number of possible association sites). Comparable challenges exist for the characterization of large biopolymers, in which higher order tandem MS methods (e.g., MS”, where n t 2) may offer the possibility of obtaining complete sequence information as well as the identity and location of structural modifcations. The gentle nature of ESI has also created new opportunities to directly probe noncovalent associations in solution (39-43) and even qualitative aspecta of higher order structure. Al-

7+ Ubiquitin

I

/

1

1205 ’ 12’15 ’

1225’

I

though the gentle interface conditions necessary to preserve such interactions by ESIMS can result in a substantial decrease in sensitivity, CE/MS is still feasible, as demonstrated in Figure 8. Shown are separations of ribonuclease S (RNaseS) with ua? of two Werent pH buffers (44). RNase S is a noncovalent complex consisting of a 13.7-kDa A-protein and a 2166-Da A-peptide. The complex dissociates under acidic buffer conditions, resulting in two peaks in the CE separation. The mass spectrum in Figure 8 was obtained from a CE/MS separation using a n ammonium bicarbonate buffer. The mass spectrum is dominated by peaks arising from intact RNase S (i.e., the A-peptide-A-protein complex). The ability to use CElMS to probe noncovalent DNA-protein, protein-ligand, DNA-drug, a n d other solution interactions is a n exciting prospect. The practicality of such applications will benefit greatly fmm improved MS performance. The prospects for improved performance are excellent. The sensitivity of the ESIMS detector can be attributed to two factors: the overall ionsampling and transmission efficiency of the ESIMS interface (typically only 10-3-10-6 with present instrumentation) and the ion utilization efficiency, which derives from the type of mass spectrometer used and its mode of operation. For example, a quadrupole mass spectrometer with a n ion-sampling efficiency of lo-’ operated in a scanning mode would typically have a n ion utilization efficiency of whereas in SIM mode i t might be IO-’ or better. If 100 ions must be detected to give an S/N of 2 during elution of a n analyte zone, the best achievable sensitivity would correspond to 0.6 and 60 amol during scanning and SIM operation, respectively-even if ESI efficiency were 10096. In fact, our best results are close to these levels of performance. Progress continues in improving ESIMS sampling and transmission efficiency, but efficient utilization of the ions created at atmospheric pressure remains elusive. More promising in the short term are approaches that address mass spectrometric ion utilization efficiency. High-performance CElMS instrumentation, in our opinion, will most likely be based on either orthogonal time-of-flight (TOF) or ion trapping (quadrupole ion trap FT-ICR) mass spectrometers. Table I compares existing quadrupole MS instrumentation with the estimated or antici-

-

~

rigure 9. An initial demonstration of on-line CUFT-ICRMS. Shown on the lee is me total ion eiectrophercgramlor separation of a polypeptide and live proteins, in which the elution rate was reduced by hall prior to elution of the fimt =lute to the elmmopray source. P a m mass M r a obtained during this separation are shown on the right for three of the proteins (ubiquitin.8,565 Da: carbonk anhydrase, 28.802 Da: and myoglobin [minus the heme moiay], 16,951 Oa). Inserts to the mass speclra Show that resolution (30.000-50.000) s~ffidentto resolve the 1-0a w i n g anributableto isotopic contributions is obtained lor each of the proteins.

582 A

-

ANALYTICAL CHEMISTRY, VOL. 65, NO. 13, JULY 1,1993

pated characteristics of three of the most promising contenders among high-performance mass spectrometers for CE/MS. Our laboratory is currently investigating both the orthogonal TOF and FT-ICR a p proaches. The rationale behind this selection is that the TOF offers the greatest scan speeds and possibly the best achievable sensitivity (potentially using nearly every ion t r a n s m i t t e d i n t o vacuum); t h e FT-ICR potentially offers exceptional performance (high MS resolution, precise mass measurements, and high sensitivity simultaneously, as well as MS") but at slower spectral acquisition rates. Thus, the CE/orthogonal TOF ap proach appears to be ideally suited for ultrahigh sensitivity and for situations in which very high separation speeds are desired (45). Such applications may include single-cell analyses, rapid sampling for the analysis of physiological systems, and twodimensional LC/CE/MS. The CE/FT-ICR combination offers exciting opportunities for situations that require high M S resolution, accurate mass measurement, and high sensitivity. Building on the pioneering work of McLafferty et al. (46, 47), we have recently developed a new FT-ICR instrument that has been used to obtain ultrahigh-resolution mass spectra (in excess of IO6) and good sensitivity (below femtomole detection limits for small proteins) for off-line ESIMS (48). Figure 9 shows an initial example from our laboratory t h a t demonstrates the practicality of on-line CE/FT-ICRMS (49). The left side of the figure shows the total ion electropherogram obtained for a separation of a polypeptide and five proteins. The duty cycle (ion injection time divided by the total time between spectra) obtained in these experiments is low (2-10%)) and a mass spectrum could be acquired only every 6 s because of data storage limitations. However, excellent quality spectra with resolutions of 30,000-50,000 were obtained from 500-fmol injections of individual components. Figure 9 also shows partial mass spectra for three of the proteins. As shown by the inserts, the FT-ICR instrument provides resolution of the 1-Da spacing of the isotopic envelope associated with each protein molecular ion charge state. Significantly, such high MS resolution was obtained even for the largest protein examined (carbonic anhydrase, 28,802 Da) for which t h e isotopic p e a k s for the 25+ charge state are

separated by only 0.040 m/z units. The combination of sensitivity and MS resolution demonstrated by these initial results represents a breakthrough i n CE/MS performance. Work in our laboratory is aimed a t obtaining further gains in the duty cycle, sensitivity, and MS resolution of CE/FT-ICRMS. Although only beginning to be realized, this combination of capabilities could ultimately allow the sampling, separation, and structural determination (e.g., sequencing, location of modifications, or noncovalent associations) from attomole and even subattomole quantities of material. A current goal at our laboratory (50) is trapping and sequencing a single (i.e., individual) ionized segment of DNA. Attainment of this ambitious goal will depend on the unique capabilities of nondestructive detection, long trapping times, high sensitivity (one ion if sufficiently charged), and MS" methods (where n can be very large) that are currently unique to FT-ICR (51). Because CE provides the capability to slow or completely stop a separation almost instantaneously, the potential exists to fully exploit the sensitivity and high-resolution capa bility of FT-ICR in conjunction with MS" experiments for structural characterization of biopolymers. The realization of such advanced CE/MS instrumentation should provide the means to address a profoundly expanded range of questions across broad fields of chemical and biological research. This research was supported by internal PNL exploratory research and by the Director, Offce of Health and Environmental Research, US. Department of Energy. Pacific Northwest Laboratory is operated by Battelle Memorial Institute for the U . S . Department of Energy, through contract DE-AC06-76RLO 1830. We thank C. J. Barinaga, J. E. Bruce, G. A. Anderson, R. T.Kouzes, H. R. Udseth, D. C. Gale, and A. Brown for contributions and helpful discussions related to the work described.

References (1) Yergey, A. E.; Edmonds, C. G.; Lewis, I.A.S.; Vestal M. L. Modern Analytical Chemistry; Plenum Press: New York, 1990. (2)-Jorgenson, J. W.; Lukacs, K. D. Science 1983, 222, 266-72. (3) Yamashita, M.; Fenn, J. B. J. Phys. Chem. 1984,88,4671-75. (4) Fenn, J. B.; Mann, M.; Meng, C. W.; Wong, S. F. Mass Spectrom. Rev. 1990, 9,

.".

27-7n

V .

(5) Smith, R. D.; Loo, J. A.; Edmonds, C. G.; Barinaga, C. J.; Udseth, H. R. Anal. Chem. 1990,62, 882-99. (6) Smith, R. D.; Loo, J. A.; Ogorzalek Loo, R. R.; Busman, M.; Udseth, H. R. Mass Spectrom. Rev. 1991, 10, 359-451. (7) Olivares, J. A.; Nguyen, N. T.; Yon-

ker, C. R.; Smith, R. D. Anal. Chem. 1987,59, 1230-32. (8) Won$, S. F.; Meng, C. K.; Fenn, J. B. Proceedings of the 35th ASMS Conference on Mass Spectromety and Allied Topics, Denver, CO, May 24-29, 1987; pp. 33-34. (9) Meng, C. K.; Mann, M.; Fenn, J.B. 2.Php. D 1988, 10, 361-68. ( 1 0 ) S m i t h , R. D.; Olivares, J. A.; Nguyen, N. T.; Udseth, H. R. Anal. Chem. 1988, 60,436-41. (11) Smith, R. D.; Barinaga, C. J.; UdSeth, H. R. Anal. Chem. 1988, 60, 194852. (12) Lee, E. D.; Muck, W.; Henion, J. D.; Covey, T. R. J. Chromatogr. 1988, 458, 313-21. . ~ . (13) Pleasance, S.; Thibault, P., Kelly, J. J. Chromatogr. 1992,591,325-39. (14) Jorgenson, J. W.; Dohmeier, D. M.; Austelc T. L. Presented a t Pittcon '92, Pittsburgh Conference, New Orleans, LA, March 9-12, 1992. (15) Wahl, J. H.; Goodlett, D. R.; Udseth, H. R.; Smith, R. D. Anal. Chem. 1992,64, 21 - --94-95 - - -. (16) Tang, L.; Kebarle, P. Anal. Chem. 1991, 63, 2709. (17) Goodlett, D. R.; Wahl, J. H.; Udseth, H. R.; Smith, R. D. J. Microcol. Sep. 1993, 5. 57-62.

press. (20) Lee, E. D.; Muck, W.; Henion, J. D.; Covey, T. R. Biomed. Environ. Mass Spectrom. 1989, 18, 844. (21) Muck, W.; Henion, J. D. J. Chromatogr. 1989, 495, 41-59. (22) Johansson, I . M.; Huang, E. C.; Henion, J. D.; Zweigenbaum, J. J. Chromatogy. 1991, 554, 311-27. ( 2 3 ) J o h a n s s o n , I . M.; P a v e l k a , R.; Henion, J. D. J. Chromatogr. 1991, 559, 515-28. (24) Moseley, M. A.; Shabanowitz, J.; H u n t , D.; Tomer, K. B.; Jorgenson, J. W. J. Am. SOC.Mass Spectrom. 1992, 3 , 289-300. (25) Deterding, L. J.; Parker, C. E.; Perkins, J. R. J. Chromatogr. 1991,554, 32938. (26) Thibault, P.; Paris, C.; Pleasance, S. Rapid Commun. Mass Spectrom. 1991, 5, 484-90. (27) Thibault, P.; Pleasance, S.; Laycock, M. V. J. Chromatogr. 1991,542,483-501. (28) Thibault, P.; Pleasance, S.; Laycock, M. V. Proceedings of the 39th ASMS Confer-

ence on Mass Spectromety and Allied Topics, Nashville, TN, 1991; pp. 593-94. (29) Moore, W. T.; Caprioli, R. M. Techniques in Protein Chemisty II; Academic Press: New York, 1991; pp. 511-28. (30) Moseley, M. A.; Deterding, L. J.; Tomer, K. B.; Jorgenson, J. W. Rapid Commun. Mass Spectrom. 1989, 3, 87-93. (31) Moseley, M. A.; Deterding, L. J.; Tomer, K. B.; Jorgenson, J. W. J. Chromatogr. 1990,516, 167-73. (32) Reinhoud, N. J.; Niessen, W.M.A.; Tjaden, U. R. Rapid Commun. Mass Spectrom. 1989, 3, 348-57. (33) Castoro, J. A.; Chiu, R. W.; Monnig, C. A.; Wilkins, C. L. J. Am. Chem. SOC. 1992, 114, 7571-72. (34) Udseth, H. R.; Loo, J. A.; Smith, R. D. Anal. Chem. 1989, 61,228-32. (35) Garcia, F.; Henion, J. D. Anal. Chem. 1992,64,985-90. (36) Tinke, A. P.; Reinhoud, N. J.; Niessen, W.M.A.; T'aden, U. R.; van der Greef, J. Rapid dommun. Mass Spectrom. 1992, 6, 560.

ANALYTICAL CHEMISTRY, VOL. 65, NO. 13, JULY 1 , 1993

583 A

REPORT include combining LC and CE with MS for biochemical applications.

(37) Thompson, T. J.; Foret, F.; Vouros, P.; Karger, B. L. Anal. Chem., in press. (38) Wahl. J.,H.;,Smith, R. D., submitted

for publication in J. Chramatafl.

(39) Ganem, B.; Li, Y-T.; Henion, J. D. J. Am. Chem. Sac. 1991, 223.6294-95. (40) Katta, V.; Chait, B. T.J. Am. Chem. Sac. 1991, 223, 8534-35. (41)Baca. M.; Kent, S.B.H. J. Am. Chem. Sot. 1992, 214,3992-93. (42) Light-Wahl, K. J.; Sprin er D L

Winger, B. E.; Edmonds, C. 6,;’Cim;: D. G., 11; Thrall, B. D.; Smith, R. D. J. Am. Chem. Sac. 1993. 225, 803-04. (43) Goodlett, D. R.; Camp, D. G., 11; Hardin, C. C.; Corregan, M.; Smith, R. D. E o / . M n n Spectrom. 1993, 22, 181-

Richard D. Smith (leP., is a senior staff scientist and group leaderfor the Chemical Methods and Separations Croup in the Chemical Sciences Department at Pacific Northwest Laboratory. He obtained his Ph.D. from the University of Utah and spent a year as an NRC postdoctoral ossaciate at the Naval Research Laboratory before joining PNL in 1976. His current research interests include the development of separation techniques, their combination with MS, and new methodsfor the ultrasensitive structural characterization of large molecules based on MS. Jon H. Wahl received his B.S. degree in chemistryfrom Michigan Technological University and his Ph.D. in analytical chemistryfiom Michigan State University. He is currently a postdoctoral researcher at PNL involved in the develop ment of CE/ESIMS. His research interests

Steven A. Hofstadler (leP., received his B.S. degree in chemistryfiom the University of New Mexico and his Ph.D. in chemistryfrom the University ofTexas. He is currently a postdoctoral research associate focusing on the development and bioanalytical applications of ESI-IT-ICRMS. He willjoin the PNLstaffthisyearasasenior research scientist. David R. Goodleffreceived his M.S. degreefiom Auburn UniveniQ and his B . D . in biochemistryfrom North Carolina State University. He recently completed a postdoctoral appointment at PNL and has joined the Immunobiology Research Institute. His research interests include the relationship between strndure andfundion in biological mammolenrles.

AN AMERICAN CHEMICAL SOCIETY JOURNAL

WhereyW Can Learn More About What’s New in the Chemistry of NorrNucIear Energy Sources E x l i limimthly issue brings y o u reports on the l,itest research < mn i x nuclear energy snurces: petrnleum, coal, shale nil, tar sands, hiomass, syntiicls, C chemistry, organic geochemistry. applied catalysis, and cnmbustion. Under the leadership of Editor John W. L a w n o f l c h i g h Uniacrsin and Associate Editor Michael T. Klein of the University of lklawarc, i keeps ?no current (in the latest news i i i the t r a n s ~ i r i i i ~ i t iutilization. ~~n, fi)rmatinn, and p r n d u c ~ tinn of kissil hicls. Plus, in each issue ynu’ll find th(~ught-pr(i\r,kiiig communications, bixik reviews, and guest fcaturc essays by major figures in the field on non-nuclear tiicls. To subscribe, write: American Chemical Swiety, Member and Subscrihcr Services, P.O.R n x 3337, Columbus, OH 43210. Or, call toll-free in the U.S. 1-800-333-9511and charge ynur nrder. Outside the U.S. call 614-447-3776; or fax your order (in 614-447-3671.

584 A

ANALYTICAL CHEMISTRY, VOL. 65, NO. 13, JULY 1.1993

2YRs

I 1

$102

* For rates outside the U.S.. Canada. and Mexim. call or write the ACS.

For information o n composition and submission of manuscripts, write: John Larsen, Editor, Energy & Fuels, Depamnent of Chemistry, 6 East Packer Avenue, Lehigh University, Bethlehem, PA, U.S.A. or Telephone 215-7583489. .This pcihbrmm I‘ aalrhlc m rnicn4ilm. rnimfi