Investigation of intact protein complexes by mass spectrometry


  • Albert J. R. Heck,

    Corresponding author
    1. Department of Biomolecular Mass Spectrometry, Bijvoet Center for Biomolecular Research and Utrecht Institute for Pharmaceutical Sciences, Utrecht University, The Netherlands
    • Department of Biomolecular Mass Spectrometry, Bijvoet Center for Biomolecular Research and Utrecht Institute for Pharmaceutical Sciences, Utrecht University, Sorbonnelaan 16, 3584 CA Utrecht, The Netherlands.
    Search for more papers by this author
  • Robert H. H. van den Heuvel

    1. Department of Biomolecular Mass Spectrometry, Bijvoet Center for Biomolecular Research and Utrecht Institute for Pharmaceutical Sciences, Utrecht University, The Netherlands
    Search for more papers by this author


  II.Electrospray Ionization of Biomacromolecules00
  III.Mass Analyzers for Biomacromolecular Mass Spectrometry00
 A.  Collisional Cooling and/or Focusing00
 B.  Alternative Mass Analyzers00
 IV.Solution-Phase Properties of Proteins Complexes00
 A.  Protein–Protein Interactions00
 B.  Cooperativity and Cofactors00
 C.  Dynamics of Assembly00
  V.Gas-Phase Properties of Protein Complexes00
 A.  Tandem Mass Spectrometry00
 B.  Charge Separation in Protein Dissociation00
 VI.Future Outlook00

Mass spectrometry has grown in recent years to a well-accepted and increasingly important complementary technique in structural biology. Especially electrospray ionization mass spectrometry is well suited for the detection of non-covalent protein complexes and their interactions with DNA, RNA, ligands, and cofactors. Over the last decade, significant advances have been made in the ionization and mass analysis techniques, which makes the investigation of even larger and more heterogeneous intact assemblies feasible. These technological developments have paved the way to study intact non-covalent protein–protein interactions, assembly and disassembly in real time, subunit exchange, cooperativity effects, and effects of cofactors, allowing us a better understanding of proteins in cellular processes. In this review, we describe some of the latest developments and several highlights. © 2004 Wiley Periodicals, Inc., Mass Spec Rev


BIRD, blackbody infrared radiative dissociation; CID, collision-induced-dissociation; ESI, electrospray ionization; FAD, flavin adenine dinucleotide; FMN, flavin mononucleotide; FT-ICR, Fourier transform ion cyclotron resonance; MALDI, matrix assisted laser desorption ionization; m/z, mass-to-charge; TOF, time-of-flight.


In this review, we address recent advances in protein mass spectrometry, with special attention given to the analysis of larger intact proteins and protein complexes. The biomacromolecular mass spectrometry research has been given an enormous impetus with the introduction of the new soft ionization methods matrix assisted laser desorption ionization (MALDI) (Tanaka et al., 1987, 1988; Karas & Hillenkamp, 1988; Hillenkamp & Karas, 1990; Hillenkamp et al., 1991) and electrospray ionization (ESI) (Wong, Meng, & Fenn, 1988; Fenn et al., 1989; Chowdhury, Katta, & Chait, 1990a,b; Meng & Fenn, 1990; Light-Wahl et al., 1993; Mann & Wilm, 1995) in the late 1980s. The enormous impact of both these new ionization methods is of course best illustrated by the Nobel prize in Chemistry of 2002 (Cho & Normile, 2002), which was partially awarded to John Fenn and Koichi Tanaka for their roles in the development of these soft ionization methods.

Through these and other recent developments mass spectrometry has become a well-accepted and increasingly important technology in proteomics (protein identification and quantification, protein profiling, and protein interactions) (Hamdan & Righetti, 2002; Mo & Karger, 2002; Aebersold & Mann, 2003). The proteomics-based mass spectrometry techniques provide information on the molecular masses of proteins and their primary sequences, but do not directly provide any information on higher order protein structures and interactions.

In parallel to the use of mass spectrometry in proteomics for primary structure elucidation, the technique has become a complementary tool in structural biology for the investigation of secondary, tertiary, and quaternary structures of protein complexes, and their interactions with DNA, RNA, ligands, and cofactors. Indeed, soon after the introduction of MALDI and ESI, these ionization methods were already used for the study of intact proteins and non-covalent protein complexes (Ganem, Li, & Henion, 1991; Miranker et al., 1993; Mirza, Cohen, & Chait, 1993; Smith & Light-Wahl, 1993). In particular, ESI is well suited to detect and investigate non-covalent complexes by transferring whole intact assemblies into the vacuum inside the mass spectrometer. The observation of these non-covalent complexes by ESI mass spectrometry strongly suggests that at least part of these native conformations persist through the ionization process. The measurement of non-covalent protein complexes is challenging caused by the fact that they tend to dissociate easily, next to additional experimental problems such as the high mass-to-charge (m/z) ratios they typically acquire, and the potential of non-specific adduct formation, which may contribute to and/or broaden ion signals.

The investigation of non-covalent protein complexes by mass spectrometry has already been reviewed extensively (Fitzgerald et al., 1996; Przybylski & Glocker, 1996; Loo, 1997, 2000; Winston & Fitzgerald, 1997; Veenstra, 1999a,b; Bakhtiar & Nelson, 2000; Krutchinsky et al., 2000; Hernandez & Robinson, 2001; Daniel et al., 2002). Why review this field of research again? The review of Loo (1997) in this journal in 1997 provides an excellent introduction with an extensive overview of nearly all studies reported in the field up to 1997. One would not be able to provide a similar overview at present as the number of publications has grown tremendously. However, much progress has been made in recent years, and with the Nobel Prize awarded to John Fenn in 2002 this seems to be an appropriate timing to review some of the recently reported advances. Taking the review by Loo as a starting point we have chosen here to describe a certainly not comprehensive perspective on some of the exciting achievements made in the last 5 years, with special attention to the biophysical backgrounds in this field of research.


Mass spectrometric measurements are carried out in the gas-phase on ionized analytes. By definition, a mass spectrometer consists of an ion source that generates gas-phase ions, a mass analyzer that measures the m/z ratio of the ionized analytes, and a detector that registers the number of ions at each m/z value. In this review, we focus on the mild ionization method ESI as source to generate gas-phase ions. This method is most appropriate to detect and analyze larger intact proteins and protein complexes. However, we should not forget that MALDI is also applied to study these assemblies. Indeed, several recent reports show that MALDI is capable of keeping non-covalent protein interactions intact in the gas-phase vacuum of the mass spectrometer (Kiselar & Downard, 2000; Strupat et al., 2000; Wattenberg et al., 2000; Friess et al., 2002). However, the dried, often acidic, organic matrix is not an ideal environment to keep the proteins under physiologically relevant conditions. Moreover, MALDI is not the method of choice for very large proteins as the resulting singly charged ion signals often result in broad mass peaks (partially induced by adduct formation), typically in the order of several hundreds of Daltons for a protein with a molecular weight of 100 kDa. The rapidly expanding literature of the ESI method indicates that this ionization method has more benefits for most biomacromolecular assemblies.

The ESI method has emerged as a powerful technique for producing intact ions in vacuo from large and complex species in solution (Wong, Meng, & Fenn, 1988; Fenn et al., 1989). As described in detail by Kebarle (2000) in electrospray, initially small charged droplets are formed at the tip of the capillary, because of the action of the applied electric field on the solution at the capillary tip (1–4 kV), making the liquid at the capillary tip enriched for positive ions (in positive ion mode). This charging leads to droplets with an elongated meniscus (Taylor cone). Evaporation of solvent from the released droplets decreases the radius of the droplets and since the charge is conserved, at some critical radius Coulombic forces overcome the surface tension of the liquid and lead to fission of the droplets into even smaller droplets. Repeated evaporation and fission lead to very small charged droplets, which are the precursors of the gas-phase ions. This ionization process takes place at atmospheric pressure and is, therefore, very gentle (without significant fragmentation of analyte ions in the gas-phase). The introduction of nanoflow ESI has played an important role in the analysis of protein complexes by ESI as this made minute amounts of sample feasible without compromising signal intensity (Wilm & Mann, 1996). Moreover, nanoflow ESI generates smaller droplets, which makes the desolvation process more efficient.

Several factors may influence the electrospray process such as pH of the solvent, non-volatile salts, contaminants, volatile buffer, and percentage of organic modifier. To illustrate the effects of some of these factors we measured mass spectra of the ferredoxin-dependent glutamate synthase from Synechocystis sp. under different solution conditions (Fig. 1). This enzyme with a monomeric mass of 165 kDa is of extreme importance for ammonia assimilation in plants and bacteria, where it catalyzes the formation of two molecules of L-glutamate from L-glutamine and 2-oxoglutarate (Vanoni & Curti, 1999). The enzyme contains a flavin mononucleotide (FMN) and a 3Fe-4S cluster as non-covalently bound cofactors. It has been proposed before, that this enzyme functions via the transfer of two electrons from two reduced ferredoxin molecules via the 3Fe-4S cluster to the FMN cofactor (van den Heuvel et al., 2002). Figure 1A represents the ESI mass spectrum in the positive ion mode of glutamate synthase sprayed from an aqueous solution containing 50% acetonitrile and 0.1% formic acid. Under these conditions, glutamate synthase denatures and, therefore, accommodates many positive charges as indicated by the broad distribution centered around [M + 66H]66+ at m/z 2,500. Using cesium iodide as the external calibrant we were able to determine the average mass of the glutamate synthase monomer to be 165,526 ± 11 Da, this being the mass of the protein that has lost its FMN cofactor and 3Fe-4S cluster. The calculated molecular mass of glutamate synthase excluding the cofactors and the N-terminal presequence of 36 amino acids is 165,477 Da. Thus, the mass spectrometric analysis of intact proteins under denaturing conditions can provide a molecular weight of a protein yielding information about possible post-translational modifications and iso-proteins, however, the experiments will not give any information on non-covalent interactions. These measurements under denaturing conditions may be difficult to perform for very large proteins or protein complexes. First, some of these proteins are instable and prone to aggregation in acidified or organic solutions. Second, the mass spectra may become very complex due to the high charge states of the different proteins.

Figure 1.

Nanoflow ESI spectra of 1 μM glutamate synthase from Synchocystis sp. A: Glutamate synthase sprayed from an aqueous solution containing 50% (v/v) acetonitrile and 0.1% (v/v) formic acid in positive ion mode shows a broad charge distribution centered around [M + 66H]+ at m/z 2,500. B: Glutamate synthase sprayed from neutral 50 mM ammonium acetate buffer in positive ion mode shows monomeric (m/z 7,000) and dimeric (m/z 9,000) species. The FMN and 3Fe-4S cofactors are non-covalently bound to the protein. C: Glutamate synthase sprayed from neutral 50 mM ammonium acetate buffer in negative ion mode shows monomeric (m/z 7,500) and dimeric (m/z 10,500) species. D: Glutamate synthase sprayed from neutral 50 mM triethyl ammonium bicarbonate in positive ion mode shows monomeric protein ions around m/z 9,000, but no dimeric ions. At m/z values higher than 12,000 we did not detect any ions belonging to the dimeric protein form. The bound FMN and 3Fe-4S cofactors are represented by (⧫) and (▴), respectively.

When glutamate synthase was measured from aqueous ammonium acetate at neutral pH the ESI mass spectrum in the positive ion mode of the monomer showed a narrow distribution of relatively low charge states centered around [M + 24H]24+ (Fig. 1B). The molecular weight determined from this spectrum is 166,222 ± 20 Da, which indicates that the species detected under these conditions includes the non-covalently bound FMN (457 Da) and 3Fe-4S (296 Da) cofactors. The measured molecular weight is in close agreement with the calculated mass based on the primary sequence including the cofactors (166,230 Da). It is often reported that the measured mass of large folded protein assemblies is somewhat higher then expected from the primary sequence. It has been speculated that the higher observed mass is because of trapping of small molecules or counterions in proteins (Nettleton et al., 1998; Rostom et al., 2000; Pinkse et al., 2003). This may especially be pronounced in multi-protein assemblies, which generally have cavities or holes. Indeed, the measured mass of the dimeric form of glutamate synthase, centered around m/z = 9,200, [2M + 36H]36+, is significantly higher then expected on the basis of the primary sequence including the cofactors (322,638 ± 23 vs. 322,460 Da) (Fig. 1).

To obtain a thorough understanding of the conditions that lead to the multiple protonation and multiple charging in general, one needs to consider the mechanism that leads to formation of multiply charged globular proteins in the gas-phase. The actual mechanism by which the solvent free gaseous ions ultimately are formed from the very small and highly charged droplets is still under some dispute, but is thought to be different for small ions compared to larger native-like proteins and protein complexes. The ion evaporation mechanism is thought to be the dominant mechanism in the production of small ions such as the conventional organic and inorganic ions (Iribarne & Thomson, 1976). However, with ESI, as described in this review, proteins and protein complexes with masses over 100 kDa may be ionized, and for such species it is quite unlikely that they evaporate from a droplet. An alternative mechanism has therefore been developed, which is expected to be dominant in the production of macro-ions of globular proteins. The charged residue mechanism was originally proposed by Dole and coworkers (Dole et al., 1968) and later validated by others (de la Mora, 2000; Felitsyn, Kitova, & Klassen, 2002). In this model, the water molecules in the smallest droplets containing a native protein evaporate completely and the protein is charged by the residual charges, which have been accumulated on the droplet during the ESI process, often in the form of ammonium ions. These excess charges are transferred to basic sites on the surface of the globular proteins. In this model, the number of charges that a globular protein or macromolecular complex obtains depends on the number of basic charges at the surface of the protein when the protein is transferred from solution-phase into gas-phase. For the analysis of the stoichiometry of protein assemblies, which are present in more than one oligomeric state, this effect is actually advantageous. For instance, the ions of dimeric protein complexes have usually and fortunately less than two times the amount of charges than the monomeric protein. If this would not be the case, these two different assemblies would be measured in the mass spectra at overlapping m/z ratios, complicating their analysis.

In a somewhat simplified view, amino acids that are at the surface of the globular protein or non-covalent complex, i.e., attainable by the ammonium ions, which contain a site that is more basic than ammonia (in the case ammonium acetate is used as volatile buffer salt) will become charged during the ionization process (Peschke, Blades, & Kebarle, 2002). On the basis of the charged residue mechanism, it has been argued that the number of charges a non-covalent complex will obtain upon ionization by electrospray can simply be predicted from the mass of the complex, when the complex is electrosprayed from aqueous ammonium acetate at a constant neutral pH (Peschke, Blades, & Kebarle, 2002). This theoretical model assumes that the proteins in the spray solution retain their solution-phase structures, at least until shortly before they are transferred to the gas-phase. In the calculated curves it is assumed that the proteins and their complexes have spherical structures. The calculated charge can be predicted by the stability limit for Coulombic fission of water droplets, the Rayleigh limit (Rayleigh, 1882), with a size equivalent to the protein or protein complex. The predicted charge is then given by:

equation image(1)

in which γ is the surface tension of the water droplet, ε0 the electrical permittivity of vacuum, e the elementary charge, and R the radius of the droplet. By assuming that the radius of a protein is directly correlated to the molecular weight of a protein and that the density of a protein is similar to that of water, an even simpler equation could be derived:

equation image(2)

in which M is the mass of the complex in Dalton.

Several research groups have validated this equation by accumulating a large number of experimental data on the ESI of globular proteins and protein complexes. Table 1 shows for a number of globular proteins and protein complexes the molecular weight and the charge of the most intense ion signal in both positive and negative ion mode (data from our group). An accumulated set of maximum intensity ion charges, from our own and other laboratories, are plotted versus the molecular mass in Figure 2. It is evident that in positive ion mode the predicted charges as calculated by Eq. 2 are in very close agreement with the measured masses over a range from 30 kDa to 1 million Da, further supporting the charged residue model. In negative ion mode, the globular proteins and non-covalent complexes obtain significantly lesser number of charges. For instance, the monomeric glutamate synthase from Synechocystis sp. with a molecular weight of 165 kDa obtains in positive ion mode approximately 24 ion charges (Fig. 1B), whereas in negative ion mode this is reduced to 22 charges (Fig. 1C). Taking all data together, we can conclude that the measured positive charge is approximately 90% of the Rayleigh model predicted charge and that the negative charge is approximately 70% of the predicted charge (Table 1). The latter percentage corresponds well with earlier data (Tolic et al., 1997; de la Mora, 2000). As the initial charged residue model leads to a predicted charge that is independent on the ionization mode (negative or positive mode) the observed discrepancy cannot be explained by the charged residue mechanism.

Table 1. Molecular weight and charge of the most abundant ion signal in both positive and negative ion mode for a number of globular proteins and protein complexesThumbnail image of
  • Measured in our laboratory by nanoflow ESI from 50 mM aqueous ammonium acetate at neutral pH.

  • aPheA2, flavin reductase component from two-component phenol hydroxylase; PHBH, p-hydroxybenzoate hydroxylase; VAO, vanillyl-alcohol oxidase; GltS, glutamate synthase; SR1, single ring mutant of GroEL; gp23, bacteriophage capsin protein.

  • bRaleigh model predicted charge, ZR = 0.078 × M½.

  • c(Measured mean ion charges/Raleigh model predicted charge) × 100.

  • Figure 2.

    Number of observed mean charges of a number of globular proteins and protein complexes compared with the Raleigh limit model predicted charge. The number of observed charges is very close to the Raleigh limit on water droplets of the same size as the protein. All proteins were sprayed from 50 mM ammonium acetate at neutral pH. (○) Represents positive ion mode and (●) negative ion mode.

    Both de la Mora and Felitsyn have argued that the experimental charge may also be influenced by an evaporation effect of the smaller buffer ions (de la Mora, 2000; Felitsyn, Kitova, & Klassen, 2002). The fact that the larger deprotonated acetate anion is more volatile than the protonated ammonium ion, would lead to more charges in the positively charged ion droplets compared to the negatively charged ion droplets. The model proposed by de la Mora and Felitsyn also predicts that when other than ammonium acetate buffer salts are used this may influence the charging of globular proteins and protein complexes. It has been shown that triethylammonium bicarbonate can also be used in ESI mass spectrometry for studying intact proteins and non-covalent complexes (Lemaire et al., 2001). In comparison with ammonium bicarbonate or acetate solutions, the use of triethylammonium bicarbonate resulted in significantly less ion charges, which may be explained by the higher gas-phase proton affinity of protonated triethylammonium bicarbonate (Lemaire et al., 2001). Moreover, it has been proposed that the multiply charged ions generated in a triethylammonium bicarbonate solution are more stable compared to the traditionally used solutions, making them more suited for the analysis of macromolecular complexes (Lemaire et al., 2001). When ferredoxin-dependent glutamate synthase was sprayed from triethylammonium bicarbonate buffer at neutral pH the average number of ion charges was 18 (Fig. 1D) compared to 24 charges when sprayed from an ammonium acetate buffer in positive ion mode. These results indicate that the observed experimental charges are influenced by an evaporation effect whereby the small ‘buffer’ ions interact with the globular proteins just before they become free gaseous ions.


    Unlike the analysis of peptides and smaller individual (denatured) proteins, which can be carried out effectively in mass spectrometers with low m/z range capabilities, the study of intact macromolecular complexes mainly depends on the m/z range of the instrument. Although other instruments may also have extensive m/z ranges, at present the mass spectrometric analysis of biomacromolecules is largely the domain of time-of-flight (TOF) mass spectrometry (Tang et al., 1994). This is not only because of the virtually unlimited m/z range, but also because of the achievable high sensitivity and speed of TOF analysis. TOF measurement depends on a pulsed ion beam and, therefore, the development of orthogonal hybrid instruments that allow a proper combination of a continuous electrospray source with pulsed ion detection has been essential for protein complex mass spectrometry (Morris et al., 1996; Loboda et al., 2000). In one of its simplest forms ESI is combined with TOF analysis using a combination of quadrupole or hexapole filters orthogonally coupled to a TOF tube. These quadrupole and/or hexapole filters do not allow mass selection of the ions, but just transmit and focus the whole ion beam, and act as an interface between the high pressure at the ESI source front end and the high vacuum of the TOF region. These instruments have become, with some small modifications in the interface (see below), extremely valuable tools in the analysis of very large protein complexes.

    Quadrupole TOF mass spectrometers are also well suited for the study of large proteins and protein complexes and have in addition the potential for tandem mass spectrometry. Quadrupole TOF machines combine a quadrupole mass filter with an orthogonal TOF analyzer (Morris et al., 1996; Loboda et al., 2000). These spectrometers provide supplementary structural information for ions isolated in the quadrupole, dissociated via collisional activation, and analyzed in the TOF analyzer. In most commercial instruments the quadrupole mass range is limited to m/z 3,000–4,000, which allows the isolation of ions of folded proteins and/or protein complexes up to a molecular mass of 50–60 kDa. Larger non-covalent complexes often exhibit ion charge states appearing well above the m/z 4,000 range and are, therefore, usually not amendable for tandem mass spectrometry. Recently, Robinson and coworkers have introduced a quadrupole TOF instrument that has been constructed to overcome this mass range limitation with a custom-built quadrupole operating up to 22,000 m/z (Sobott et al., 2002). This has allowed them to dissociate protein complexes well in excess of 60 kDa and subsequently to determine their subunit composition. This ability to perform tandem mass spectrometry on large complexes becomes especially important when the composition of for instance heterogeneous complexes is unknown. The tandem mass spectrometry capabilities of this quadrupole TOF were elegantly demonstrated in the multi-component human plasma proteins transthyretin and retinol-binding protein with the ligands thyroxine and retinol (McCammon et al., 2002). Transthyretin is responsible for transfer of the hormone thyroxine, and through binding with retinol-binding protein, retinol (vitamin A). The precursor ions of this multi-component system of approximately 80 kDa were not well resolved in the mass spectrum. Ion selection and subsequent collision-induced-dissociation (CID) of the ions clearly revealed fragments in the mass spectrum. The isolated peaks were consistent with the presence of a complex consisting of tetrameric transthyretin with two molecules of thyroxine and retinol-binding protein with one molecule of retinol. This instrument developed specifically for biomacromolecular mass spectrometry still provides a mass resolution of 3,000 at m/z 45,000.

    A. Collisional Cooling and/or Focusing

    Several recent reports have shown that transmission of high mass ions requires pressures in the first vacuum stages of the mass spectrometer to be increased by reducing the pumping speed or adding a collision gas (Krutchinsky et al., 1998; Rostom et al., 2000; van Berkel et al., 2000; Schmidt, Bahr, & Karas, 2001; Tahallah et al., 2001; Sobott et al., 2002). Thus, high m/z ions do not benefit from high vacuum conditions within the mass spectrometer as might be expected from multi-protein complexes that are only held together by non-covalent interactions. That this, at first glance, may be an unexpected phenomenon was nicely illustrated by the ESI mass spectra of the folded flavoprotein vanillyl-alcohol oxidase (Fig. 3) (Tahallah et al., 2001). Vanillyl-alcohol oxidase is a homooctameric enzyme with each monomeric subunit of approximately 65 kDa containing a covalently bound flavin adenine dinucleotide (FAD) cofactor. At low concentrations, the octameric assembly is in equilibrium with the dimeric species (Fraaije, Mattevi, & van Berkel, 1997; van Berkel et al., 2000). In these experiments, the intensity of His61Thr vanillyl-alcohol oxidase (with added FAD) ions was measured as a function of pressure (2–7 mbar) in the first hexapole of the instrument (G1). Under standard pressure conditions (i.e., 2 mbar), no ions could be detected, however, upon increasing the source pressure the total ion current increased continuously. At intermediate pressures (i.e., 3–5 mbar) the ions originating from the dimeric species were dominant, however, upon increasing the pressures further the ions originating from the octameric species became most intense (Fig. 3). Further studies have revealed that optimal pressure conditions are different for different ionic species, and are largely dependent on the molecular weight (Tahallah et al., 2001). Krutchinsky et al. (1998) have suggested that larger ions may acquire substantial kinetic energies (of more than 1,000 eV) when they are sprayed out of the supersonic jet. This may have a negative effect on the transfer and orthogonal extraction into the TOF region. The increased pressure in the preceding quadrupoles/hexapoles may act as a collisional dampening interface. It is now widely accepted that a combination of collisional dampening, increased cooling of the ions, and more efficient desolvation is effective in the enhanced detection of ions with high m/z values. The fact that the detection of ions is m/z dependent makes it evident that this pressure effect phenomenon should be carefully addressed when relating ion abundance in mass spectra to solution-phase abundance of non-covalent protein assemblies, especially in equilibrium measurements between different components (Ayed et al., 1998).

    Figure 3.

    Nanoflow ESI spectra of 5 μM His61Thr vanillyl-alcohol oxidase with 5 μM added FAD in 50 mM ammonium acetate, pH 6.8 as measured at different source pressures. The most abundant ions originate from the dimeric and octameric protein species. The pressure read-outs for the first Pirani gauge (G1) are given for each spectrum. Reproduced with modifications from Tahallah et al. (2001).

    B. Alternative Mass Analyzers

    Although, as clear from the previous section, TOF analyzers seem to be most appropriate for studying macromolecular complexes, impressive results have also been reported by Loo and coworkers using sector instruments (Sannes-Lowery et al., 1997; Kuchumov, Loo, & Vinogradov, 2000), which have the advantage that tandem mass spectrometry may be performed on the non-covalent complexes at rather high collision energies. Also fourier transform ion cyclotron resonance (FT-ICR) instruments can be used to investigate large macromolecular non-covalent complexes, e.g., the 38.5 kDa pentameric Shiga-like toxin (Kitova et al., 2001). However, at the relatively high m/z values these complexes attain in electrospray, the mass resolution in FT-ICR is no longer better than in TOF analyzers. In FT-ICR, the ions are trapped within an ion trap and they thus can be specifically manipulated, for instance by collisional activation, infrared or black-body radiation, and/or gas-phase reactions with other gaseous molecules (Felitsyn, Kitova, & Klassen, 2001, 2002; Breuker et al., 2002; Oh et al., 2002).

    Quadrupole ion trap analyzers are not routinely used for the analysis of protein complexes as most standard instruments have a limited m/z range of 3,000–4,000, however, in principal the m/z range of ion traps can be extended (Wang et al., 2000). The application of ion traps may be advantageous as they also allow the storage of ions for extended periods of time so that ion/molecule reactions may be performed as well as multiple stages of mass spectrometry for structural elucidation. So far, ion trap mass spectrometry has been little exploited to study large non-covalent complexes, although a few papers have described ion traps with extended m/z ranges to study such complexes and/or protein ions of high m/z values (Stephenson, Cargile, & McLuckey, 1999; Wang et al., 2000; Cargile, McLuckey, & Stephenson, 2001; Reid, Stephenson, & McLuckey, 2002; Stephenson et al., 2002). In our group, we have evaluated the use of the LCQ-Deca (ThermoFinnigan) ion trap with an extended m/z range of up to 20,000. Using this setup, we have recorded the nanoflow ESI spectrum of vanillyl-alcohol oxidase sprayed from ammonium acetate at neutral pH (Fig. 4). With the ion trap, we observed ions originating from the dimer at m/z values around 6,000 and from the octamer at m/z values around 11,000. This study indicates that the detection of these large protein assemblies by quadrupole ion trap is feasible and should be exploited in more detail. However, by comparing the performance of the standard ion trap and the ESI-TOF (Fig. 3) we had to conclude that the performance of the ESI-TOF is better. As for ESI-TOF we observed a positive effect on the transmission and detection of large non-covalent complexes upon increasing the pressure in the ion trap.

    Figure 4.

    Mass spectrum of 10 μM vanillyl-alcohol oxidase in 50 mM ammonium acetate, pH 6.8 as measured with the LCQ-Deca ion trap. Ions originating from the vanillyl-alcohol oxidase dimer (m/z 6,000) and octamer (m/z 11,000) were detected.

    In addition to the more traditional mass analyzers described above, recent reports have also introduced ion mobility analyzers into the field of biomacromolecular mass spectrometry. In this approach, the charged ions/particles are separated in a high-pressure drift tube, which allows the determination of their collisional cross-section, providing information on their size and to some extent globular structure. Such drift tube differential mobility measurements have been applied to intact proteins, glycoproteins, non-covalent protein complexes (streptavidin, avidin, alcohol dehydrogenase, and catalase), and human rhinoviruses (Bacher et al., 2001). In the latter case, infectious, intact human rhinovirus could be distinguished from a heat-degraded viral particle having lost its infectiousness. The heat-treated particle actually exhibited a somewhat longer mobility time (i.e., apparent higher mass), which has been explained by an increase of capsid protein shell diameter by penetration of water into the virus capsid causing its swelling. Thus, the mass of the degraded particle is not higher, albeit its size is bigger than that of the infectiousness virus particle.

    Finally, large megaDalton particles have also been investigated by charge-detection TOF mass spectrometry (Fuerstenau & Benner, 1995; Fuerstenau et al., 2001). In this technique, the detection is based on the simultaneous measurement of the charge of the particles using the image current they induce on the flight tube, and their m/z value using TOF methodology. Although the obtained molecular weights are not extremely accurate, it has been shown that this technique can be used to study very large complexes as tobacco mosaic virus particles with molecular weight of approximately 40 million Da (Fuerstenau et al., 2001). This indicates that the mass-limit achievable by ESI has probably not been reached yet.


    A. Protein–Protein Interactions

    Many proteins interact in vivo strongly or weakly with alike proteins forming oligomers or with other proteins such as in functional protein complexes and/or signaling pathways. In addition, proteins interact with DNA, RNA, and a variety of other biomolecules (e.g., peptides, sugars etc.). Investigation of oligomeric proteins is of utmost interest for instance to improve our understanding of enzymatic regulation and protein stability. It is believed that protein oligomerization improves the stability of the involved proteins against proteolysis and thermal denaturation. Knowing that such multimeric interactions can be retained during the transfer of proteins from solution-phase to gas-phase, mass spectrometry provides an ideal approach to investigate such protein–protein interactions as the measured molecular mass of the oligomer directly reveals the number of subunits in the quarternary structure. This can be illustrated by using the nanoflow ESI spectra of ferredoxin-dependent glutamate synthase from Synechocystis sp. in the positive ion mode (Fig. 1B). Small-angle X-ray scattering data have indicated that the enzyme is monomeric in solution, however, in the crystalline state glutamate synthase is shown to be dimeric (van den Heuvel et al., 2003). The ESI mass spectra of glutamate synthase revealed that in the gas-phase the enzyme is in equilibrium between the monomeric and dimeric form.

    The ESI mass spectrometry studies on the quaternary structure of 4-oxalocrotonate tautomerase is one of the classical studies in which the oligomeric state of a protein was probed (Fitzgerald et al., 1996). 4-oxalocrotonate tautomerase was estimated to be a pentamer on the basis of size-exclusion chromatography and ultracentrifugation (Chen et al., 1992), however, X-ray crystallography and mass spectrometry clearly showed that the protein forms a hexamer of 41 kDa (Roper et al., 1994; Fitzgerald et al., 1996). Four single point-mutated 4-oxalocrotonate tautomerases, on the other hand, only form monomeric protein ions. These data thus provide information about the oligomerization behavior of wild type protein and residues important in maintaining the hexameric structure.

    Following the pioneering work of the research group of standing, several other groups have used electrospray mass spectrometry for the analysis of the stoichiometry of oligomeric protein complexes of known and unknown composition. Interesting studies are reported on complexes of proteasome activator proteins (Yao et al., 1999; Zhang et al., 1999), the chaperone complex GroEL (Rostom & Robinson, 1999; Ashcroft et al., 2002), and many different variants of hemoglobin (Zal et al., 1997a,b, 2000; Apostol, 1999; Lippincott et al., 2000; Versluis & Heck, 2001). ESI mass spectrometry experiments have confined that GroEL is composed of 14 non-covalently bound subunits arranged in two heptameric rings with a total molecular mass of 800 kDa (Rostom & Robinson, 1999). For its functional activity, GroEL requires the presence of the co-chaperone GroES. The association of the mobile loop of GroES and the model substrate helix D from rhodanese to GroEL was probed by mass spectrometry (Ashcroft et al., 2002). Together with the results from fluorescence binding studies it was concluded that chaperone GroES and substrate proteins have, at least partially, distinct binding sites even in the intact GroEL tetradecamer.

    Hemoglobin is in most mammalian systems present as a tetrameric complex of approximately 60 kDa, being possibly in equilibrium with dimers in lower abundance, as also indicated by mass spectrometry. However, in terrestrial aquatic, marine, and deep sea annelids hemoglobin forms larger multimeric complexes. Mass spectrometry has been used extensively to study these larger complexes such as that of the multi-million Dalton hemoglobin complexes of Lumbricus terrestris and Alvinella pompejana (Zal et al., 1997a,b; Kuchumov, Loo, & Vinogradov, 2000). The study on the relative distribution of hemoglobin S and fetal hemoglobin in blood from patient material of patients with homozygous sickle cell disease (Ofori-Acquah et al., 2001) is a related example of the use of mass spectrometry focused on non-covalent oligomeric structures of hemoglobin. In this study, the researchers concentrated on the quantitative distribution of asymmetric hemoglobin hybrid and other tetrameric species in blood of patients with sickle cell disease. Therefore, the non-covalent association of hemoglobin subunits in hemolysates was studied by ESI mass spectrometry. The spectra of both patient and fetal blood revealed intact hetero-hemoglobin tetramers of both α2β2 and hybrid α2γβ. A unique tetrameric marker protein of an average mass of 64.6 kDa was identified in hemolysates from patients with sickle cell disease in accordance with the calculated mass of the asymmetric hemoglobin hybrid.

    B. Cooperativity and Cofactors

    Oligomerization of proteins, in particular enzymes and proteins involved in signal transduction may be advantageous to provide a means of allosteric interactions between the subunits. In some multimeric protein assemblies, these interactions may give rise to cooperative binding of substrates or ligands, which is believed to play a role in the regulation of their activity (Koshland & Hamadani, 2002).

    Van Dorsselaer and coworkers performed one of the first studies in which cooperativity effects were probed by mass spectrometry (Rogniaux et al., 2001). A semi-quantitative interpretation of the ESI mass spectra provided the relative abundance of all the distinct enzymatic species with and without multiple copies of the oxidized cofactor nicotinamide adenine dinucleotide. This allowed them to deduce cooperativity mechanisms that accompany the binding of oxidized cofactor to several different oligomeric glyceraldehyde-3-phosphate dehydrogenases and one tetrameric alcohol dehydrogenase. Overall, the obtained results were in excellent agreement with the cooperative behavior as measured by fluorescence quenching, calorimetry, and ultra violet spectroscopy. The method is highly attractive in terms of the required amount of biological material as only a single ESI mass spectrum is necessary.

    Another elegant study revealing the ability of mass spectrometry to investigate cooperativity is the work on the transthyretin system (McCammon et al., 2002). The tetrameric human protein transthyretin transfers the hormone thyroxine and, through its non-covalent interactions with the retinol binding protein, retinol. Transthyretin is thought to be one of the human proteins involved in the formation of amyloid fibrils, which is associated with diseases such as Alzheimer's, type II diabetes, and the transmissible spongiform encephalopathies. Under normal conditions, transthyretin transports thyroxine and retinol, but misfolding of wild-type and single point mutations lead to senile systemic amyloidosis and familial amyloid polyneuropathy, respectively. Dissociation of the tetrameric transthyretin complex is thought to be the first step in the formation of transthyretin amyloid fibrils (Miroy et al., 1996). McCammon et al. (2002) examined a series of 18 possible inhibitors of this process by mass spectrometry. The ligands were evaluated for their ability to bind to and stabilize the tetrameric structure, their cooperativity in binding to the transthyretin complex and their ability to compete with the natural ligand thyroxine. The observation of the multifaceted ten-component complex containing six protein subunits (four transthyretin and two retinol binding proteins), two retinol molecules, and two synthetic ligands allowed them to conclude that ligand binding does not inhibit association of transthyretin with retinol bound retinol binding protein. Thyroxine and the probed synthetic ligands showed different cooperativity characteristics. Negative, positive, and non-cooperative mechanisms could be observed, amongst them the well-established negative cooperativity for thyroxine binding.

    The influence of cofactors on the stability of oligomeric protein assemblies can be quite dramatic. It has been reported that the binding of non-covalent cofactors to proteins induces subunit association and improves resistance of the protein against thermal and chemical denaturation. The oligomerization behavior of the flavoenzyme vanillyl-alcohol oxidase and its site-directed mutant His61Thr provides a nice illustration (Tahallah et al., 2002). Whereas wild type vanillyl-alcohol oxidase contains a covalently bound FAD cofactor, the His61Thr variant is a flavin-free apo-protein. It is well known that wild type vanillyl-alcohol oxidase forms primarily octameric assemblies of 507 kDa (Mattevi et al., 1997; van Berkel et al., 2000). In contrast, the apo-protein His61Thr predominantly forms dimeric assemblies of 126 kDa (Fig. 5) (Tahallah et al., 2002). ESI mass spectrometry was used to monitor non-covalent binding of FAD to His61Thr vanillyl-alcohol oxidase (Fig. 5). Binding of the cofactor could be monitored both to dimeric and octameric species as association resulted in an increased mass. The data clearly show that binding of the flavin cofactor coincides with octamerization of His61Thr, suggesting that cofactor binding is involved in association of dimeric vanillyl-alcohol oxidase subunits to octameric assemblies. This major effect on the oligomeric state of His61Thr vanillyl-alcohol oxidase is somewhat surprising as the flavin-binding pocket is well embedded within the monomeric subunits. The crystallographic model of wild type vanillyl-alcohol oxidase (Mattevi et al., 1997) and the His61Thr variant (Fraaije et al., 2000) has even revealed that the absence of the covalent flavin bond does not induce any major structural changes within the enzyme. Still from the mass spectrometric and complementary size-exclusion chromatography data it is evident that the absence of the cofactor induces dimerization. Therefore, it has been speculated that upon flavin binding small conformational changes in the cofactor binding pocket of the dimeric species are transmitted to the protein surface, promoting octamerization.

    Figure 5.

    Nanoflow ESI mass spectra of the His61Thr vanillyl-alcohol oxidase protein in 50 mM ammonium acetate, pH 6.8. A: Mass spectrum of 3 μM FAD-deficient His61Thr shows a largely dimeric species around m/z 5,500. B and C: Mass spectra of His61Thr with added FAD at a ratio of protein to cofactor of 1:1 and 1:10 show intense octameric peaks centered around m/z 9,500. Reproduced with modifications from Tahallah et al. (2002).

    An interesting example of the use of mass spectrometry in studying local cooperativity is the study towards the unfolding of lysozyme (Canet et al., 2002). Lysozyme is an extremely well studied protein, which is identified to be associated with a recently discovered human amyloidotic disease. The known cases of this non-neuropathic systemic amyloidosis result from single point mutations in the lysozyme gene, which result in one or the other of two mutant proteins (Asp67His and Ile56Thr). In their native states, the two lysozyme mutants are structurally similar to wild type protein. However, hydrogen exchange experiments monitored by mass spectrometry and nuclear magnetic resonance revealed that the amyloidogenic mutation Asp67His reduces the stability of the β-domain and the adjoining C-helix. Moreover, mass spectrometry revealed that the transient unfolding of these regions occurs with high degree of cooperativity. This behavior results in the simultaneous unfolding of the β-domain and the C-helix, whereas the remaining part of the protein is still folded. The authors have postulated that this species is likely to initiate aggregation events that ultimately lead to well-defined β-sheet rich fibrillar structures that are found in patients carrying the mutated gene encoding lysozyme.

    C. Dynamics of Assembly

    Conformational properties of proteins are influenced by many factors including environmental parameters such as pH and temperature. Whereas heat-induced changes of individual proteins can be monitored by an array of different methods, such as circular dichroism and calorimetry, influences of the temperature on higher order quaternary protein assemblies are more difficult to study by conventional methods. The appearance of electrospray mass spectra of native proteins is known to depend on the temperature of the solution. ESI sources often incorporate heating elements, including some that can be used to heat the protein containing electrospray capillary. Therefore, in studying protein conformational properties care has to be taken not to heat the drying gas and the ion source too much. On the other hand, the ability to control the temperature of the solution can, and has been, used to study for instance the thermal unfolding of proteins online (Mirza, Cohen, & Chait, 1993). Several studies have shown that results obtained on the thermal stability of proteins by such an ESI-based method correlate well with data obtained by conventional methods.

    Recently, a thermo-controller system was introduced that can be directly coupled to a nanoflow electrospray setup (Benesch, Sobott, & Robinson, 2003). This device was used to probe the thermal stability of the protein lysozyme, but also the temperature induced disassembly of a small heat shock protein complex. Small heat shock proteins are present in virtually all organisms and act to increase cellular tolerance against stress conditions. Moreover, it is postulated that they can prevent improper protein associations. A mini-Peltier cooled device was constructed in which the used nanospray needles were embedded. The major advantage of this device is that still small volumes of solution can be used, whereby the temperature of the solution can be varied between approximately 10 and 100°C. The wheat small heat shock protein TaHSP16.9 studied is a 17 kDa protein which forms at room temperature and under physiological conditions a dodecamer with a total molecular weight of just over 200 kDa. It has been shown that the quarternary structure of the protein is very dynamic and that major conformational changes occur upon heat shock temperature. Figure 6 shows the ESI mass spectra of this protein complex studied by temperature-controlled nanoflow electrospray from a 200 mM ammonium acetate solution. The spectra clearly reveal the heat induced dissociation of the dodecamer into primarily monomers. By plotting the decreased signal intensity of the dodecamers as a function of the total ion intensity, a transition curve for the dissociation could be extracted, which in turn provided a melting temperature of the complex of 60°C.

    Figure 6.

    Thermo-controlled dissociation of the dodecameric heat shock protein TaHSP16.9 in 200 mM ammonium acetate, pH 6.8 as measured by nanoflow ESI mass spectrometry. Spectra at different temperatures show the dissociation of the dodecamer (m/z 6,000) into sub-oligomeric species (low m/z). Selected charge states for monomers (solid type) and dimers (outlined type) are labeled. The inset shows the percentage of dodecamer versus temperature. A melting point of 60 ± 2°C was determined from the transition. Reprinted with permission from Benesch, Sobott, & Robinson (2003). Copyright 2003 by American Chemical Society.

    Pinkse et al. (2003) have reported another example in which the dynamics of the very large multi-protein complex urease from Helicobacter pylori has been studied. H. pylori is a gram-negative spiral bacterium infecting approximately half of the world's population and an important etiologic agent in a variety of diseases (Covacci et al., 1999). The abundant protein complex urease is essential for survival in the acidic environment of the stomach and its pathogenic capacity. The enzyme catalyzes the hydrolysis of urea, present in gastric juice, to ammonia and carbamate. It has been postulated that the generated ammonia neutralizes the gastric acidity and forms a neutral microenvironment that surrounds the bacterium enabling it to colonize the gastric lumen (Eaton et al., 1991). Urease consists of a 26.5 kDa α-subunit and a 61.7 kDa β-subunit (Hu & Mobley, 1990; Evans et al., 1991). Nanoflow ESI mass spectrometry has established that the urease complex has a molecular weight of 1,063,900 ± 600 Da, corresponding to a dodecameric (αβ)12 assembly (Fig. 7A). The observation that the dodecamer readily disassembles into (αβ)3 subunits is in strong support for a ((αβ)3)4 architecture. Such an architecture is consistent with the recently published X-ray structure, revealing a spherical assembly of twelve catalytic αβ subunits with an outer diameter of ∼160 Å (Ha et al., 2001). When urease was sprayed at a higher monomer concentration, 24-, 36-, and even 48-mers with molecular masses over 4 million Dalton were observed from which the individual charge states could still be identified (Fig. 7B). These aggregates are likely to be formed aspecifically and not of any biological relevance. Mass spectrometry also showed that in vitro fully denatured α- and β-subunits reassemble when reconstituted in ammonium acetate buffer, hereby forming the (αβ)3 substructure, but not the dodecameric complex (Pinkse et al., 2003). The fact that the (αβ)3 urease subunits did not reassemble to the dodecameric ((αβ)3)4 assembly was expected as for proper in vivo assembly of urease several auxiliary proteins are needed (Park & Hausinger, 1995).

    Figure 7.

    Nanoflow ESI mass spectra of urease in 200 mM ammonium acetate, pH 8.0. A: Mass spectrum of 20 μM urease reveals primarily (αβ)12 subunits. B: At an increased monomer concentration of 40 μM urease forms aspecific 24-, 36-, and 48-mers with m/z values of 18,000, 21,000, and 27,000, respectively.

    Electrospray mass spectrometry has also proven to be successful in determining subunit exchange in oligomeric proteins. Fluorescence resonance energy transfer is a widely accepted approach to study subunit exchange among oligomers, however, this technique is unable to detect individual species and their abundances. Mass spectrometry-based monitoring subunit exchange has first been described for type II DNA-binding proteins (Vis, Dobson, & Robinson, 1999). The HU proteins from Bacillus stearothermophilus and B. subtilis are homodimers of interlaced polypeptides containing 90 and 92 residues, respectively. The primary sequences of the two proteins are highly identical and the X-ray models show a high degree of structural conservation (White et al., 1989). The structural similarity permits the formation of heterodimers, which allows monitoring the slow interconversion between the two homodimeric species (Fig. 8). The obtained results correlate well with the rate of dissociation of homodimers into monomers and the difference in free energy between homo- and heterodimeric species.

    Figure 8.

    Interconversion between the HU proteins from B. stearothermophilus (HBst) and B. subtilis (Hbsu) in 500 mM ammonium acetate as measured by ESI mass spectrometry. Solutions of the two proteins were mixed, heated to 80°C, and cooled down to 20°C. Mass spectra were obtained at 0, 2, 20, and 500 min. The fractions of the three dimers are initially randomly distributed, but when equilibration takes place those of both homodimers increase, whereas that of the heterodimer decreases. Reprinted with permission from Vis, Dobson, & Robinson (1999). Copyright 1999 by The Protein Society.

    Real-time monitoring of subunit exchange has been further explored by using small heat shock proteins from pea (PsHSP18.1) and wheat (TaHSP16.9). Under the mass spectrometry conditions used, the two heat shock protein mainly form homodecamers with molecular masses just over 200,000 Da, which corresponds well with the X-ray model of TaHSP16.9 (van Montfort et al., 2001). Upon mixing the two protein assemblies at different molar ratios, heterododecameric species were found from which the composition was dependent on the initial subunit ratio (Sobott et al., 2002). Following the kinetics of subunit exchange suggested that dimeric assemblies are the predominant species of exchange between the two heat shock proteins. Thus, these heat shock proteins are very dynamic but stable with dimeric species moving out and in of the oligomer. This may imply that in cases in which different heat shock proteins of the same class are expressed in the same cellular compartment we have to expect that subunits interchange.


    A. Tandem Mass Spectrometry

    As described in the previous sections, mass spectrometry can provide information on intact protein complexes in solution-phase, such as molecular mass, stoichiometry, dynamics of assembly, cooperativity, and ligand binding. Mass spectrometry can, however, also be used to dissect protein assemblies in the gas-phase in a sequential fashion to identify for instance the building blocks and possibly even obtain information regarding topology, quaternary structure, and stability.

    Typically, in tandem mass spectrometry, ions of a protein complex are mass selected and via collisions with an inert gas activated and reacted. Structural information can then be inferred from the resulting mass-analyzed product ions, which may be used to obtain further structural information. In the initial CID studies on protein complexes, the ESI generated ions were accelerated in the interface region between the source and the first quadrupole of the mass spectrometer (Tang et al., 1994), often termed nozzle-skimmer induced dissociation. Such experiments, however, lack proper precursor ion selection hampering the determination of the origin of the resulting fragment ions. With modern instruments, tandem mass spectrometry experiments on protein complexes are relatively straightforward to perform, however, the obtained data from measurements in vacuo have to be treated with care when they are used to deduce solution-phase structural properties of the studied non-covalent protein complexes. To date, the mass spectrometric studies on the decomposition pathways of protein complexes do not give a conclusive answer whether or not the solution-phase properties and structures of these non-covalent complexes are (sufficiently) retained during the electrospray process and in the vacuum. Some of the tandem mass spectrometry studies show a correlation between the gas-phase and the solution-phase structural and thermodynamical properties, but others reveal serious conflicts. For instance, quarternary structures may change by the transfer to the gas-phase, since water molecules cannot longer assist in stabilizing the molecular structures.

    Some of the earlier studies by Smith and coworkers showed that dissociation patterns observed in tandem mass spectrometry do not coincide with expected thermodynamical and structural properties of the non-covalent protein complexes in solution. They studied the decomposition patterns of the tetrameric assemblies concanavalin A, human hemoglobin and avidin and found that the predominant dissociation channel led to the formation of mostly monomeric and trimeric protein (Light-Wahl et al., 1993; Smith & Light-Wahl, 1993; Light-Wahl, Schwartz, & Smith, 1994; Schwartz, Light-Wahl, & Smith, 1994). Notably, both hemoglobin and concanavalin A are known to be composed in solution of two dimers, heterogeneous for hemoglobin and homogeneous for concanavalin A. A related study on the gas-phase dissociation of hemoglobin from bovine, porcine, and human origin also revealed the monomer–trimer dissociation as the predominant reaction channel. Moreover, dissociation pathways such as the expulsion of heme dimers by the holo-tetrameric hemoglobin complexes were observed, being in conflict with known solution-phase properties of hemoglobin (Versluis & Heck, 2001). From several other studies it has become evident that in general larger multi-protein complexes show a tendency to decompose by CID via a similar pathway, i.e., expulsion of a small monomeric subunit. This phenomenon has also been described for the hexameric α2β4 complex of archaeal GimC/prefolding homologue (MtGimC), which primarily eliminates β monomers following collisional activation (Fandrich et al., 2000), the above described multi-protein transthyretin complex, which decomposes in the gas-phase primarily via loss of a single transthyretin or retinol binding protein subunit (Sobott et al., 2002) and the dodecamer of the small heat shock protein TaHSP16.9 (Benesch, Sobott, & Robinson, 2003). Thus, the expulsion of a single small moiety seems to be a general pathway in the gas-phase dissociation of protein complexes, even when this is not expected from the solution-phase structural and thermodynamical properties of the complex.

    Other tandem mass spectrometric studies show better correlations between gas-phase and solution-phase structural and thermodynamical properties. For instance, Williams and coworkers have reported that the gas-phase ionic stability of complementary DNA duplexes is more stable when compared to non-complementary DNA duplexes, establishing that the Watson–Crick base pairing might be preserved in the gas-phase and can thus exist in the absence of any solvent molecules (Schnier et al., 1998). Moreover, Douglas and coworkers have determined the relative activation energies for the loss of heme in non-covalent myoglobin ions, whereby they systematically mutated amino acids involved in the heme binding. A correlation was found between the solution-phase Arrhenius activation energy for heme loss of the studied mutants and their gas-phase kinetic stability of the corresponding ions (Hunter, Mauk, & Douglas, 1997).

    B. Charge Separation in Protein Dissociation

    As found for other tetrameric complexes, the gas-phase dissociation of the streptavidin tetramer follows the typical monomer–trimer decomposition pattern. Surprisingly, the resulting monomer fragment moiety carries a disproportionately large fraction of the charge (Schwartz, Light-Wahl, & Smith, 1994). A priori one might expect upon gas-phase dissociation of the tetramer ions the trimer would carry ¾ of the charge and the monomer ¼ of the charge. However, the predominant dissociation pathway of the tetramer of streptavidin S4 followed the pathway; S414+ dissociates into S7+ and S37+. This disparate charge distribution is unexpected as the trimer has not only a mass three times the monomer, but is also significantly larger. Since this early report, such skewed charge distributions have been reported many times in the dissociation of larger protein complexes (Fitzgerald et al., 1996; Rostom et al., 1998; Rostom & Robinson, 1999; Zhang et al., 1999; Felitsyn, Kitova, & Klassen, 2001, 2002; Nesatyy, 2001; Versluis & Heck, 2001; Versluis et al., 2001; Mauk et al., 2002; Benesch, Sobott, & Robinson, 2003; Jurchen & Williams, 2003), and have led to much research and speculation about the origin of this phenomenon.

    Schwartz, Light-Wahl, & Smith (1994) were the first to come up with a model explaining this fascinating, anomalous charge distribution. They have suggested that the asymmetric dissociation process is analogous to the fission of charged droplets in the electrospray process in which the smaller offspring droplets are believed to form with much higher m/z than the parent droplet. By assuming that the volume of the concomitant fragment ions is directly related to the mass, their model explains to some extent the observed inequality of charge partition over the two fragments. In a related study, however, different specific and aspecific protein homodimers also show striking disparate charge distributions whereas the simple charged droplet fission model predicts for dissociations of protein homodimers an equal charge for both the concomitant equal-mass fragments (Versluis et al., 2001). The dimeric glyoxalase I from Escherichia coli was observed to dissociate under CID conditions primarily via the pathway; Gly211+ dissociates into Gly8+ and Gly3+. Versluis et al. (2001) suggested that the model of Smith and coworkers could still be valid, but only if it was assumed that the two monomers formed in the gas-phase collisional activation process had very different shapes, whereby the highly charged entity would become more unfolded and less compact than the less charged concomitant entity. Others have argued that the disparate charge distribution may also be related to the electrostatic repulsion of the fragment ions (Mauk et al., 2002). If the reverse reaction is considered, i.e., association of the charged monomers, there will be a Coulomb barrier with a height proportional to the product of the charges on the fragment ions. For protein dimers, this barrier is minimized for the most asymmetric distribution. When this Coulomb barrier contributes to the activation energy for dissociation, the most asymmetric charge distributions will have the lowest energies for dissociation. Competing with this effect, however, is the additional energy required to move charges to new sites in the complex prior to dissociation.

    Klassen and coworkers have revealed that the disparate charge distribution observed in the CID spectra of ionic protein complexes is reproduced when these ions are activated through infrared absorption using the blackbody infrared radiative dissociation (BIRD), which is performed inside a Fourier transform ion trap (Tholmann, Tonner, & McMahon, 1994; Dunbar & McMahon, 1998; Felitsyn, Kitova, & Klassen, 2001, 2002; Kitova, Bundle, & Klassen, 2002). The BIRD induced thermal dissociation of the homodimer ecotin showed that the asymmetric partitioning of charge was somewhat dependent on the charge state of the precursor dimer and the temperature (Felitsyn, Kitova, & Klassen, 2002). In this paper, it is argued that isomeric forms of the dimer may pre-exist in the gas-phase, whereby the charge is already asymmetrically divided over the two monomer entities. An additional feature of such BIRD experiments is that the dissociation kinetics may be followed as a function of temperature. From the resulting Arrhenius plots, activation energies for dissociation and Arrhenius factors may be determined. For instance in BIRD experiments on the dissociation of the homo-pentameric Shiga-like toxin it was found that the magnitude of the Arrhenius parameters is highly dependent on the charge state of the pentamer. Also the activation energies are charge dependent between 35 and 80 kcal/mol, increasing with decreasing charge state of the precursor. For some of the lowest observed charge states of the pentamer (11+), Arrhenius parameters as high as 1039 sec−1 have been reported, indicating an unusual high transition state entropy. This suggests major structural changes taking place during the dissociation process (Felitsyn, Kitova, & Klassen, 2001). It has been suggested that the disassembly of the complex is taking place via unfolding of a leaving subunit, being energetically unfavorable, albeit compensated by transfer of the charge to this subunit, which would be entropically favorable. Additionally, it has been argued that the apparent gas-phase basicity of unfolded proteins may be substantially higher than that of more compact native proteins, making it also energetically somewhat more favorable to transfer charges to the unfolded subunit. The increased apparent basicity may originate from exposure of extra lysine and arginine residues in the unfolded protein and from the fact that in an elongated structure the charges may be located further away from each other reducing the Coulombic repulsion.

    Williams and coworkers have systematically studied several factors that may influence the charge separation in the dissociation of ionic protein dimers (Jurchen & Williams, 2003). In addition to confirming that the internal energy and the charge state of the precursor is important, they have also investigated the influence of the conformational flexibility of the monomer entities. The native and reduced α-lactalbumin dimers were subjected to sustained off-resonance irradiation CID, a method in which an ion is activated incrementally by many low-energy collisions, inside a fourier transform mass spectrometer. They found that in the oxidized state, without disulfide bridges, the charge partitioning is highly asymmetric, whereas in the reduced state the charge portioning is quite symmetric. Similar data on other native, reduced, and chemically cross-linked proteins led them to conclude that asymmetric charge partitioning is greatly reduced when the conformational flexibility of the leaving fragments is diminished.

    These studies show that disparate charge partitioning depends on a number of different factors, such as internal energy, dimer gas-phase conformation, charge state, and conformational flexibility of the subunits. The general picture seems now to be that the asymmetric charge partitioning proceeds via protein unfolding in the dissociative transition state, which requires substantial energy, but is entropically favorable. The extensive amount of work on the asymmetric charge distributions, and the above given explanation that leads to these effects, provides another caution for the interpretation of structural information on protein complexes in solution from such gas-phase data.


    The development of electrospray by the group of John Fenn and others has led to the fact that very large biomacromolecular complexes can be investigated by mass spectrometry. In his words “it has made it possible to led these elephants fly.” It now seems that the transition from solution-phase into gas-phase of large non-covalent protein complexes or even virus particles is not any more a bottleneck. Additionally, several mass analyzers are available for analyzing the by electrospray generated ions, even when they exhibit m/z values far above 100,000 Th. Therefore, the future for analyzing protein complexes by mass spectrometry seems to be bright.

    Most examples described in this review focus on relatively homogeneous protein complexes, often oligomers of one or two different protein subunits. We have to keep in mind, however, that protein complexes can be far more complex containing various different proteins, RNA, DNA, and additional smaller molecules. In fact, recent proteomic studies have suggested that hundreds of heterogeneous protein networks exist in living cells. Biomacromolecular mass spectrometry as described here may play a prominent role in validation and structure determination of these protein complexes. The nanoflow ESI mass spectrometric structural analysis of intact 70S ribosome complexes, consisting of proteins and RNA, by the group of Robinson is one of the first examples in which such a heterogeneous assembly has been studied (Rostom et al., 2000). The greatest challenge now is not only to validate well-characterized complexes, but also to define unknown complexes consisting of proteins, RNA, DNA, cofactors, and other biomolecules. These protein assemblies are not static but often very dynamic entities with transient binding partners. Therefore, a second major challenge is the evaluation of the dynamics of assembly and disassembly of protein complexes by mass spectrometry. This will yield unique information in this intriguing field of structural biology.


    We thank all current and alumni members of the Biomolecular Mass Spectrometry group in Utrecht. We particularly thank Cees Versluis, Claudia Maier, Martijn Pinkse, Nora Tahallah, Isabel Catalina, and Esther van Duijn who have contributed to some of the work reported in this review. Additionally, we express our gratitude to many esteemed colleagues for useful discussions over the years on this exciting subject, in particular Ken Standing (University of Manitoba), Carol Robinson (Cambridge University), Peter Roepstorff and Thomas Jorgensen (University of Southern Denmark), Joseph Loo (UCLA), Fred McLafferty (Cornell University), Willem van Berkel (Wageningen University), Byung-Ha Oh (Pohang University), and John Klassen and Paul Kebarle (University of Alberta). We dedicate this review to John Fenn, who made all this possible.

    Biographical Information

    Prof. Dr. Albert J.R. Heck received his PhD in mass spectrometry from the University of Amsterdam in 1993, working under supervision of Nico Nibbering. After a post-doctoral period at the laboratories of Prof. Dr. Richard Zare at Stanford University and Dr. David Chandler at Sandia National Laboratories, he took up a lectureship in Chemistry at the University of Warwick, UK (1996–1998). In 1998 he became full professor of Biomolecular Mass Spectrometry, an inter-facultary position shared by the departments of Chemistry and Pharmacy of Utrecht University, The Netherlands. He and his group (approx. 25 people) are embedded in the Bijvoet Research School for Biomolecular Research and the Utrecht Institute for Pharmaceutical Sciences. His main interest is to develop and implement innovative mass spectrometric methods for the more efficient and detailed characterization of biomolecules in relation to their biological function. The emphasis is on the identification and structural characterization of proteins and their post-translational modifications as well as the investigation of protein complexes and protein interactions important in e.g. protein folding, protein ligand binding, and the formation of tertiary and quaternary structures. Heck is recipient of the Golden Medal of the Dutch Royal Society for Chemistry (KNCV) and has been appointed as ABC University professor in 2003. Since 2004 Heck is scientific director of the Netherlands Proteomics Centre.

    Dr. Robert van den Heuvel received his PhD degree in biochemistry at Wageningen University (The Netherlands) under the supervision of Dr. Willem van Berkel and Prof. Dr. Colja Laane in 2001. In 2001 Robert also received an individual Marie Curie fellowship and a long-term EMBO fellowship allowing him to start a post-doctural studies in the protein crystallography laboratory of Prof. Dr. Andrea Mattevi at the University of Pavia (Italy). In this period Robert van den Heuvel became interested in the biological function of supramolecular protein complexes and protein–protein interactions. In 2003, he joined the Biomolecular Mass Spectrometry group of Prof. Dr. Albert Heck at Utrecht University (The Netherlands) as an assistant professor. His main research interests are in the structural characterization of heterogeneous protein assemblies, protein–protein interactions, and protein–ligand interactions in relation to their biological function using mass spectrometry as the primary tool.