a82

Anal. Chem. 1990, 62,882-899

PERSPECTIVE: ANALYTICAL BIOTECHNOLOGY

New Developments in Biochemical Mass Spectrometry: Electrospray Ionization Richard D. Smith,* Joseph A. Loo, Charles G . Edmonds, Charles J. Barinaga, and Harold R. Udseth

Chemical Methods and Separations Group, Chemical Sciences Department, Pacific Northwest Laboratory, Richland, Washington 99352

The principles, development, and recent application of electrospray loniration-mass spectrometry (ESI-MS) to Mdogical compounds are reviewed. ESI-MS methods now allow determination of accurate molecular weights for proteins extending to over 50 000, and in m e cases well over 100 000. Similar capabilltles are being developed for oligonucleotides. The instrumentation used for ESI-MS is briefly described and it is shown that, although Ionization efficiency appears to be uniformly high, detector sensltivlty may be directly correlated wlth molecular weight. The use of tandem mass spectrometry (e.g., MS/MS) for extending collision-induced dissociation (CID)methods to the structural studies of large mdecules Is described. For example, effective CID of various albumin species (molecular weight -66 000) can be obtained, far larger than obtainable for singly charged molecular Ions. The combination of capillary electrophoresis, In both free solution zone electrophoresis and isotachophoresls formats, as well as microcolumn liquid chromatography with ESI-MS, provides the capability for on-line separation and analysis of subplcomole quantities of proteins. These and other new developments related to ESI-MS are illustrated by a range of examples. Fundamental conslderations suggest even more Impressive developments may be antidpated related to detection sensitivity and methods for obtaining structural Information.

I. INTRODUCTION Over the past two decades mass spectrometry has advanced to the point where it has become one of the most broadly applicable tools in analytical chemistry. The rapid pace of this evolution has been spurred by periodic advances in mass spectrometer hardware, computer interfacing, new extended mass spectrometric methods (e.g., tandem mass spectrometry), interfaces with separation methods, and an increasing performance/cost ratio. The ion formation process is the starting point for mass spectrometric analyses and dictates the scope and utility of the method. Ionization methods based upon ion “desorption”, the direct formation or emission of ions from solid or liquid surfaces, have allowed increasing application to nonvolatile and thermally labile compounds. These eliminate the need for neutral molecule volatilization prior to ionization and generally minimize thermal degradation of the molecular species. These methods include field desorption (FD) ( I ) , plasma desorption (PD) (2, 3 ) , laser desorption (LD) ( 4 , 5 ) ,

fast particle bombardment (e.g., fast atom bombardment, FAB, and secondary ion mass spectrometry, SIMS) ( 6 ) ,and thermospray (TS) ionization (7).FAB is easily implemented on magnetic sector and quadrupole instruments and is most widely used. TS is the most broadly applied for the on-line combination with liquid chromatography, although the continuous flow FAB (8, 9 ) methods, given an eluent at compatible flow rates (i.e., 55 hL/min compared to 0.1-2 mL/min for TS), have also shown significant potential. The ionization methods amenable to nonvolatile biological compounds have overlapping ranges of application. Molecular weight (or relative molecular mass, M,) limitations of each method are also somewhat varied and can be only qualitatively defined. Ionization efficiency for these methods is generally highly compound and matrix dependent. Other factors such as mass spectrometer transmission efficiency, masking effects due to “matrix” (or “background”) contributions, and detector efficiency may conspire in limiting the M, range. Currently available results suggest,the upper molecular weights are -8000 for TS (IO),-25000 for FAB ( I I ) , and -45000 for PD (12). As M, increases, molar sensitivity decreases for all these methods. Since TS is practiced mainly with quadrupole mass spectrometers, sensitivity typically suffers disproportionately at higher mass-to-charge ratios (m/z). Commercial time-of-flight (TOF) instrumentation has been available for several years for P D (2, 3). An advantage of TOF instrumentation is that the m / z range is limited only by detector efficiency. Some of the most important successes for biomolecules have been obtained by using the FAB desorption technique. For example, in combination with high-performance (double focusing, electric and magnetic sector) mass spectrometers, powerful methods for protein sequencing have been developed based upon tandem MS (i.e., MS/MS) of enzymatic digests (13,14). Depending upon instrumentation, polypeptides with M, up to 2500 (and in some cases 4000 depending upon molecular structure and sequence) can now be addressed. In general, however, the principal limitations of these mass spectrometric methods arise from the decreasing efficiency with which molecular ions will be formed a t increasingly large M,. In the last two years, two “new” ionization methods have begun a revolution that is profoundly expanding the role of mass spectrometry in biological research. Although these methods, based upon matrix-assisted laser desorption (LD) ( 4 , 5 , 1 5 )and electrospray ionization (ESI),have origins going back more than two decades, only very recently has their potential been demonstrated for biomolecules exceeding a molecular weight of 100000. These methods show extremely

0003-2700/90/0362-0882$02.50/0C 1990 American Chemical Soclety

ANALYTICAL CHEMISTRY, VOL. 62, NO. 9, MAY 1, 1990

high ionization efficiencies (i.e., very high [molecular ions produced]/ [molecules consumed]) and, for ESI in particular, the capability for very precise M , measurements. The purpose of this article is to review the historical development, physical basis, and current state of ESI-MS application to biomolecules. The precision, accuracy, and limitations of MI measurements by ESI-MS are described. The important subject of sensitivity, which defines the ultimate potential of the technique, is considered from the viewpoints of sample size, concentration and flow rate, overall detection efficiency (i.e., [ions detected]/ [molecules introduced]),and actual ionization efficiency. The range of applications is illustrated by examples including polypeptides, intact proteins, and oligonucleotides. The high charge state molecular ions produced by ESI allow tandem mass spectrometry to be extended to much higher MI than previously possible. The potentially important role of ESI for on-line interfacing of various capillary electrophoresis (CE) methods also enhances the promise of ESI-MS. Such combined CE-MS and microcolumn LC-MS methods will be crucial for future extension to the attomole ( mol) sample range. It must be emphasized that the areas discussed are undergoing rapid advances. Here, we attempt to convey an accurate summary of the current state-of-the-art. However, past experience tells us that the present capabilities, based upon relatively crude “first generation” instrumentation (and often evaluated by performance with “known” samples of unknown quality) will almost certainly be inferior to that of a year or two hence. 11. HISTORICAL PERSPECTIVE Although the study of electrospray phenomena extends back many years, perhaps over two and one-half centuries to the work of Bose (16)and certainly to that of Zeleny early in this century (13,the seminal research into the use of electrospray as an ionization method was due to Malcom Dole and coworkers (18, 19). More than 20 years ago these workers performed extensive studies into the electrospray process and defined many of the important experimental parameters. The purpose of Dole’s studies was to use ESI to produce gas-phase macro-ions. Experimental evidence was presented for ionization of zein ( M , 50000) (20)and lysozyme (21). Interpretation of their results was problematic, however, because a mass spectrometer was not available and only ion retardation (18-20) or ion mobility (21)measurements were feasible. The concluding paragraph of their 1971 paper (20) is particularly prophetic: “This research began with the ultimate hope of being able to measure molecular weights and molecular weight distributions of samples of high molecular weight compounds by means of a mass spectrometric type technique. The results of this paper as well as of the previous work demonstrate that the electrospray technique can produce intact gas phase macromolecules of variable charge and aggregation. The technique can be applied only to electrosprayable solvents and is limited in this respect. Furthermore, only m / z ratios can be measured. A separate technique for determining the charge independently of the mass is greatly to be desired. Or perhaps a technique can be developed in which the charge per species can be controlled. It is hoped to initiate work along these lines. At any rate the results of this paper demonstrate for the first time that a molecular beam containing protein ions can be produced.” In 1984 combined electrospray ionization-mass spectrometry (ESI-MS) was reported, essentially simultaneously, by both Yamashita and Fenn (22)and Aleksandrov et al. (23,24). Fenn and co-workers also demonstrated ESI-MS in the negative ion mode (25),originally explored by Dole and coworkers (18-20). However, the Soviet researchers independ-

-

883

ently demonstrated such accomplishments as the on-line combination with liquid chromatography in 1984 (24),and over the next few years the application to oligosaccharides(26), intact polypeptides with M, to 1500 (27),and methods for their sequence determination based upon chemical digestion (28). An interesting contrast between the work of Aleksandrov et al. and that of nearly all other researchers was the use of a magnetic sector instrument, as opposed to the quadrupole instruments. The initial research of the Fenn group also stimulated our research beginning in 1984 aimed at combined capillary electrophoresis-mass spectrometry (CE-MS). Of particular importance for the ESI of macromolecules is the phenomenon of multiple charging. Since the ESI process involves the formation of highly charged liquid droplets, as discussed in more detail below, the possibility of multiple charging is implicit. The multiple charging of lysozyme by ESI was reported by Dole on the basis of ion mobility measurements (21),but difficulties in interpretation suggested a charge state of only 3+, too low by a factor of almost 5 based upon subsequent MS studies. In 1985 both Fenn and coworkers (29)and Aleksandrov et al. (26,27)observed dominant contributions for doubly charged polypeptides such as bradykinin ( M , 1060) and gramicidin S (M, 1141). Aleksandrov et al. later reported the analysis of a polypeptide of MI 3492 which produced both doubly and triply protonated molecular ions (30). Most significant, however, was the report by Fenn and co-workers of poly(ethy1ene glycols) of average molecular weights up to 17500 with as many as 23 charges (due to sodium ions) (31). Since the higher molecular weight poly(ethylene glycol) samples had relatively broad molecular weight distributions, only unresolved “humps” of ions from m/z -550 to >1400 were observed for higher molecular weight samples. Further studies required a polymer of well-defined molecular weight. The next step involved the study of such naturally occurring polymers: proteins. In 1988, Fenn and co-workers (32) first reported ESI-MS spectra of intact multiply protonated molecular ions of proteins up to MI 40 OOO, having as many as 45 positive charges. With this development ESI-MS reached critical mass. 111. THE ELECTROSPRAY IONIZATION PROCESS Although the use of electrospray ionization for mass spectrometry is relatively recent, it and related phenomena have been extensively investigated (I7,33-41). Electrospray ion production requires two steps: dispersal of highly charged droplets at near atomspheric pressure, followed by conditions resulting in droplet evaporation. An electrospray is generally produced by application of a high electric field to a small flow of liquid (generally 1-10 ML/min) from a capillary tube. A potential difference of 3-6 kV is typically applied between the capillary and counter electrode located 0.3-2 cm away (where ions, charged clusters, and even charged droplets, depending on the extent of desolvation, may be sampled by the mass spectrometer through a small orifice). The electric field results in charge accumulation on the liquid surface a t the capillary terminus; thus the liquid flow rate, resistivity, and surface tension are important factors in droplet production. The high electric field results in disruption of the liquid surface and formation of highly charged liquid droplets. Positively or negatively charged droplets can be produced depending upon the capillary bias. The negative ion mode requires the presence of an electron scavenger such as oxygen to inhibit electrical discharge (25). While a wide range of liquids can be sprayed electrostatically into vacuum (33),or with the aid of nebulizing gas (19,42),the use of only electric fields for nebulization leads to some practical restrictions on the range of liquid conductivity and dielectric constant (35, 36). Solution conductivity of S10-5 9-’ is required at room

884

ANALYTICAL CHEMISTRY, VOL. 62, NO. 9,MAY 1, 1990

temperature for a stable electrospray at useful liquid flow rates (35),corresponding to an aqueous electrolyte solution of 5104 N. Most stable electrospray performance is generally obtained when the current due to ion migration to the liquid electrospray tip is well matched to electrolyte concentration. Fluids with higher surface tensions require a higher threshold voltage for stable electrospray production (-8 kV for water vs -4 kV for methanol given a 1 cm capillary-counter electrode distance) (36). Higher dielectric liquids produce higher currents (37). While ESI currents for a typical water/ methanol/5% acetic acid solution would be in the range of 0.1-0.5 FA, conducting liquids can produce much higher ESI currents (up to 500 FA for liquid metals) (39,40). For conducting liquids, stable operation is obtained under very low flow rate conditions, where droplet production is minimized and direct field induced ion evaporation apparently dominates (39, 41). This variation upon ESI is referred to as electrohydrodynamic ionization when practiced under vacuum (43). In the mode found most useful for ESI-MS, an appropriate liquid flow rate results in dispersion of the liquid as a fine mist. A short distance from the capillary the droplet diameter is often quite uniform and on the order of 1 pm (38). Of particular importance is that the total electrospray ion current increases only slightly for higher liquid flow rates. Increasing flow rates result in formation of larger droplets and ultimately electrical breakdown, although the use of a nebulizing gas can produce stable results at flow rates as high as 100 pL/min (42). The use of higher voltages does not substantially increase the electrospray ion current until the onset of a corona discharge (generally at >6 kV). There is evidence that heating is useful for manipulating the electrospray process. For example, recent results (44) indicate that slight heating allows aqueous solutions to be readily electrosprayed, presumably due to the decreased viscosity and surface tension. Both thermally assisted (45) or gas-nebulization-assisted (42) electrosprays allow higher liquid flow rates to be used, but decrease the extent of droplet charging. The reduced droplet charging can result in a lower average charge state for multiply charge ions or lower ionization efficiency for singly charged ions. Initial work with “electrically assisted” thermospray indicates that the extent of multiple charging increases compared with conventional TS (44). For example, it has been shown that myoglobin with at least 13 positive charges can be formed by inducing greater charging on thermosprayed droplets, less than half the average charge state observed by ESI. These results clearly suggest that average charge state may be continuously adjustable between the limiting cases of electrospray ionization (electric field, no heating) and thermospray ionization (heating, no electric field). These results also suggest that the charge-state distribution somehow reflects the extent of charging in the initial droplet population. The maximum extent of charging results for the electrospray process, and any methods aimed a t “assisting” ESI are expected to decrease charging (46). The formation of molecular ions requires conditions effecting evaporation of the initial droplet population. This can be accomplished at higher pressures by a flow of dry gas at moderate temperatures (C60 “C), by heating during transport through the interface, and (particularly in the case of ion trapping methods) by energetic collisions a t relatively low pressure. Droplets shrink to the point where repulsive Coulombic forces approach the level of droplet cohesive forces (e.g., surface tension). Two mechanisms are commonly cited for ion formation from charged droplets; droplet fission at the Rayleigh limit (47) and direct field evaporation of ions (34). A field of -lo8 V/m is required for evaporation of “preformed” ions, as opposed to the larger field strengths required for desorption ionization of neutral species from

solution (34). The maximum droplet charge is proportional to d3I2 (d = droplet diameter) for Rayleigh fission and d2 for direct ion evaporation. For liquid metal electrosprays Kelly (48) showed that direct field evaporation might be expected to dominate at some small droplet size, determined by the surface tension and the electric field strength a t the droplet surface. Iribarne and Thomson (49, 50) used a similar approach to rationalize ion formation from charged aqueous droplets. For large (>1pm) droplets there is no doubt that the Rayleigh limit applies, where droplets explode, typically shedding more than 10% of their charge and about an order of magnitude less mass (51). Thus, it has been postulated that droplets undergo a cascade of fission processes yielding smaller and smaller droplets until the electric field at the droplet surface is sufficient for ion evaporation (48-50). The ion evaporation mechanism has been criticized by Rollgen et al. (52-54), who argue that the actual field strength required for ion evaporation from even the smallest aqueous droplets almost always exceeds that for Rayleigh fission (54). They claim the ion evaporation process yielding a single ion is improbable since such an event would almost certainly induce a Rayleigh explosion (causing a jet of very small charged droplets to be emitted). Additionally, these workers note that droplet disintegration will be augmented by a nonuniform charge distribution (as likely for large multiply charged ions in a low conductivity medium) (54). Zolotoi et al. (55) also point out that the evaporation of “bare” (unsolvated) small ions is reasonable only at very high field strengths. They show that the relative rates of ion evaporation predicted for small “bare” anions and cations differ by over 20 orders of magnitude, even though these species appear to be produced in ESI mass spectra with comparable efficiencies. Shiryaeva and Grigor’ev have also shown that instabilities in small multiply charged clusters can explain the formation of singly charged ions (56). It may be that a process yielding jets of very small droplets by Rayleigh fission (51) would be mechanistically indistinguishable from a process involving the field evaporation of highly solvated ions. Although the detailed processes underlying ESI remain uncertain, the macroscopic features are becoming clear. The very small droplets produced by ESI appear to allow almost any species carrying a net charge in solution to be transferred to the gas phase after evaporation of residual solvent. Mass spectrometric detection then requires that ions have a tractable m / z range (14000for quadrupole instruments) after desolvation, as well as to be produced and transmitted with sufficient efficiency. The wide range of solutes already found to be amenable to ESI-MS, and the lack of substantial dependence of ionization efficiency upon molecular weight, suggest a highly nondiscriminating and broadly applicable ionization process. IV. ESI-MS INSTRUMENTATION The electrospray ion source functions at near atmospheric pressure. Atmospheric pressure ionization (AP1)-MS has been practiced for many years, and commercial instrumentation has been available (e.g., Sciex, Thornhill, Ontario) but has not been widely utilized. The recent developments in ESI-MS are rapidly changing this situation. The electrospray “source” is typically a metal or glass capillary incorporating a method for electrically biasing the liquid solution relative to a counter electrode. Solutions, typically water-methanol mixtures, containing the andyke and often other additives, such as acetic acid, flow to the capillary terminus. The liquid sheath ESI source, developed at our laboratory, differs substantially from that of other ESI devices, which generally utilize direct infusion through a metal capillary. Our design consists of a 100-Mm-i.d. fused silica capillary which

ANALYTICAL CHEMISTRY, VOL. 62, NO. 9, MAY 1, 1990

885

CAD GAS

INLET Liquid Sheath Electrode

f

CRYO-SURFACE CHANNEL

Fused Silica

Stainless Steel Capillary

Figure 1. Schematic diagram of the instrumentation used for ESI-tandem TAGA 6000E modified to provide an additional stage of pumping.

protrudes -0.2 mm from a stainless steel tube (57). A liquid sheath, typically methanol or acetonitrile, establishes electrical contact to the analyte solution eluting from the fused silica capillary. Other sheath components are often added to avoid solute precipitation or enhance stability for CE-MS interfacing. One feature of this ESI source is that essentially any solvent system can be used. The analyte solution is mixed with the second, largely organic, sheath solvent (which may contain water and/or specific electrolytes) at the point of electrospray formation. A voltage of between +3.0 and +6.0 kV (for positive ions) is applied to the stainless steel tube in the case of the sheath ESI source, and directly to the capillary for conventional arrangements, establishing the electrospray. Typical ESI flow rates are -1-10 pL/min. For the sheath interface design, syringe pumps control the flow of analyte and sheath liquids at typically 0.25-1.0 and 3-5 pL/min, respectively. We do not distinguish here between ESI and nebulization-assisted ESI, called "ion spray" ( 4 2 ) ,since the instrumentation and resulting spectra are nearly identical. Our experience is that the nebulization-assisted mode of operation provides a slight improvement in electrospray stability at the cost of reduced sensitivity. However, this approach has some of the same advantages of the sheath flow method, allowing operation with aqueous samples. Typically a countercurrent flow of warm gas (up to -80 "C)between the nozzle and the ESI source is used to aid desolvation, although sufficient heating during transport into the mass spectrometer can also accomplish effective desolvation. Chait and co-workers have shown that direct heating of a stainless steel capillary can provide significant droplet desolvation during transport into vacuum without any gas flow a t atmospheric pressure (58). A similar approach can allow ESI operation on instruments equipped for thermospray ionization (44). Studies with injection into a quadrupole ion trap have shown that desolvation can be completed after trapping by collisions with the helium bath gas (59). We have also found that the ability to "collisionally heat" ions by the

mass spectrometry at the authors' laboratory, based upon a Sciex Table I. Typical Electrospray Ion Source Characteristics

Ionization total current (1-5) X lo-' A unit charges/s (0.6-3) X lo'* droplet diametera -1-2 wm Ion Sampling (Nozzle-Skimmer or Capillary Inlet-Skimmer)* % of total

ionization

through nozzled focused into quadrupole detectede

-10-2

-lod

-10-5

current, A -2 x 10-9 -2 X lo-" -2 x 10-12

total ions/sc -1010

-lo8 -107

"Estimated at -0.3 cm from capillary. Droplets generally become too small to be visible (