Digital ion trap mass analysis of high mass protein complexes using IR activation coupled with ion/ion reactions

Abstract

Native mass spectrometry (MS) focuses on measuring the masses of large biomolecular complexes and probing their structures. Large biomolecular complexes are readily introduced into mass spectrometers as gas-phase ions using electrospray ionization (ESI); however, the ions tend to be heavily adducted with solvent and salts, which leads to mass measurement errors. Various solution clean-up approaches can reduce the degree of adduction prior to introduction to the mass spectrometer. Gas-phase activation of trapped ions can provide additional adduct reduction, and charge reduction ion/ion reactions increase charge state separation. Together, gas-phase activation and charge reduction can combine to yield spectra of well separated charge states for improved mass measurements. A simple gas-phase collisional activation technique is to apply a dipolar DC (DDC) field to opposing electrodes in an ion trap. DDC activation loses its efficacy when ions are trapped at low q values, which is true of the high m/z ions generated by charge reduction ion/ion reactions. Digital ion trapping (DIT) readily traps high m/z ions at higher q values by varying trapping frequency rather than amplitude, but the low frequencies used to trap high m/z ions also decreases the efficacy of DDC activation. We demonstrate here using ions derived from GroEL that IR activation of ions shows no discrimination against high m/z ions trapped with DIT, because they can be focused equally well to the trap center to interact with the IR laser beam. Following pump out of excess background gas, IR activation can also induce efficient dissociation of the GroEL complex. This work demonstrates that IR activation is an effective approach for ion heating in native MS over the unusually wide range of charge states accessible via gas-phase ion/ion reactions.

1. Introduction

As mass spectrometry expands to include the measurement of large biological complexes, challenges associated with measuring analytes with high mass-to-charge ratios (m/z) and wide mass distributions have motivated instrumentation developments [1]. Electrospray ionization (ESI) typically generates gas-phase ions of large biomolecules over a range of charge states [2]. The range of charge states permits mass determination using mass analyzers with modest m/z ranges [3]. However, ESI also can lead to high levels of solvation and adduction, especially on gas-phase ions generated under ‘native’ conditions via the charged residue mechanism of ESI [4,5]. High mass gas-phase ions are often generated with high degrees of salt adduction, which leads to poorer peak signal levels and, often, to poor resolution of charge states. Even if individual charge states are visible and confidently assigned, solvation and adduction can significantly affect the accuracy of the mass measurement.

Solution-phase and gas-phase approaches have been employed to mitigate errors associated by solvation and adduction. Chromatography, buffer exchange procedures, and other solution cleaning procedures, for example, can greatly enhance the quality of ESI mass spectra [6,7]. Additionally, instrument adaptions, such as smaller nESI tip diameters, can generate smaller initial droplets leading to gas-phase ions with less initial adduction [8]. The use of relatively heavy background gases and relatively high voltage gradients along the ion path in regions of relatively high pressure has shown to drive off volatile adducts [9,10]. Instruments equipped with ion traps can accumulate and store gas-phase ions for long periods of time and facilitate processes that can further decrease adduction, such as dipolar DC (DDC) activation [11–13], IR activation [14], and ion/ion reactions aimed at removing salts [15,16]. Ion traps also facilitate charge reduction ion/ion reactions that generate lower charge states that are more readily resolved due to the increased m/z spacings between adjacent charge states [17–19]. The combination of charge reduction ion/ion reactions with gas-phase activation thus has the potential to improve both the accuracy of charge state assignments and the accuracy of peak positions on the m/z axis leading to more accurate mass measurements of high mass analytes. In this work, we compare DDC and IR activation, two forms of slow heating used in tandem mass spectrometry [20], applied to ions derived from nano-ESI (nESI) of GroEL under ‘native’ conditions both before and after proton transfer ion/ion reactions. We demonstrate that IR activation provides a greater degree of flexibility than collisional heating (e.g. DDC) in driving off salts and other small adducts, particularly at low charge states.

2. Experimental

2.1. Materials

Chaperonin 60 from Escherichia coli (GroEL, 801 kDa) and perfluoro-1-octanol (PFO, 400 Da) were purchased from MilliporeSigma (St. Lous, MO). GroEL was prepared based on published procedures [21]. 1 mg of GroEL powder was dissolved in 160 μL buffer A (20 mM tris-HCL, 50 mM potassium acetate, 0.5 mM ethylenediaminetetraacetic acid, 5 mM magnesium chloride) with 2 mM adenosine-5′-triphosphate (ATP) adjusted to pH 7. The solution was shaken slowly for 30 min. 40 μL ethanol was added to the solution, and it was shaken again for another 30 min. 400 μL acetone was added and the solution was allowed to sit for 5–15 min to allow precipitation of the protein. The mixture was centrifuged at 5000 g for 5 min and liquid was decanted. The pellet was dissolved in 200 μL buffer A with ATP. Buffer exchange into 150 mM ammonium acetate was performed with 100 kDa cut-off Amicon centrifugal filters (MilliporeSigma). The solution was centrifuged at 14,000 g for 10 min and washed with 500 μL ammonium acetate at 14,000 g for 10 min. The concentrated sample was recovered at 2000 g for 2 min and diluted with ammonium acetate to 500 μL. Positive GroEL ions were generated via nano-ESI (nESI) with 700–900 V. PFO was dissolved in 99:1 (v/v) methanol/ammonium hydroxide to a concentration of ~300 μM. Negative singly-deprotonated PFO dimers were generated via nESI with −700 to −800 V.

2.2. Instrumentation

An in-house built and modified 3D ion trap mass spectrometer (shown in Fig. 1) was used for the experiments. The ion trap was operated as a digital ion trap [22,23] using an in-house designed instrument controller and custom electronics [24]. A typical experiment first accumulated positive GroEL ions into the ion trap where they were trapped with a 120 kHz ± 400 V square wave applied to the ring electrode. After ramping the frequency to 300 kHz, negative PFO dimer ions (m/z = 799) were admitted into the ion trap. Both polarities were stored together for the ion/ion reaction. The frequency was then ramped back to a lower frequency (<120 kHz), depending on the desired m/z scan range, to eject unreacted PFO ions and to stabilize high-m/z product ions at higher q values in preparation for mass analysis. For mass analysis, the frequency was scanned in proportion to the square of its inverse (or linear with m/z) to an even lower frequency again depending on the desired scanning m/z range. Resonance ejection at q = 0.5 was effected by applying a dipolar square wave (2.5–10 V) to the end cap electrodes that was phase-locked at one-fourth the frequency of the trapping square wave.

Fig. 1.

3D digital ion trap mass spectrometer with IR laser and pulsed gas valve.

IR activation was performed using a Synrad (Mukilteo, WA) fancooled v40 CO₂ laser (40 W @ 10.6 μm with a beam diameter of 2.5 ± 0.5 mm) that was focused through 1 mm diameter holes in the ring electrode. The supplied laser controller had a TTL input with which it could be gated on and off for a specified amount of time. When on, a built-in command signal (0–5 V @ 5 kHz) triggered firing of the laser. The percentage of laser output was controlled with the supplied controller that set the duty cycle of the command signal. In this publication, the duty cycle percentage is reported when comparing different laser output percentages. The beam was directed through the holes in the ring electrode using a series of silver coated mirrors (>96% reflectance for 2–20 μm) purchased from Thorlabs (Newton, NJ). The beam was focused to the trap center using a zinc selenide plano-convex lens (>97% average transmission for 7–12 μm) and passed through a BaF₂ wedged window (~70% transmission @ 10.6 μm) into the vacuum chamber. The lens and window were also purchased from Thorlabs.

A TTL controlled pulsed valve was installed in the background gas (N₂) line so that higher pressures (~2 mTorr) could be used to trap and cool incoming ions, and lower pressures (<1 mTorr) could be used during activation periods to limit collisional cooling [25]. Following the methodology of reference [25], the valve was opened for 5 ms prior to injection of GroEL which filled the trap with nitrogen gas that had built up pressure behind the closed valve during the rest of the previous scan. The initial pulse of gas was enough to efficiently trap both GroEL and PFO for the ion/ion reactions. If the valve was left open during the entire injection period of GroEL (500 ms), longer pump out times were required to reduce the trap pressure for ion activation without providing an increase in ion signal. Pulsing gas prior to mass analysis did not increase signal intensity, possibly because enough residual gas from the initial pulse was present to cool product ions before the mass scan.

3. Results and discussion

3.1. DDC vs. IR activation

Fig. 2 provides a series of product ion spectra involving cations of GroEL in reaction with anions of PFO and illustrates the effect of using either DDC or IR activation coupled with charge reduction ion/ion reactions (The pre-ion/ion mass spectrum is shown in Fig. 3a). The precursor ion population can be activated prior to charge reduction, or the product population can be activated following charge reduction. Prior to activation, the gas valve was closed and ions were trapped for 500 ms so that excess gas could be pumped from the trap to a pressure of approximately 0.8 mTorr to reduce collisional cooling during activation. Without any activation (Fig. 2a), the relative mass error was+1.9% and the 4+ charge state had an ‘apparent resolution’ (m/Δm FWHM of the envelope of ions at a particular charge state, which includes adducts, isotope distributions, etc.) of 180. The m/z scale for these post ion/ion reaction spectra was calibrated using the charge states of Fig.2e assuming a GroEL mass of 801 kDa. Quoted mass errors are therefore relative to any mass errors associated with the scale of Fig. 2e. For Fig. 2b, the trapping square wave frequency was decreased to 90 kHz to maximize the DDC voltage that could be applied without losing precursor ions due to the low mass cut-off (LMCO). The maximum DDC voltage without losing ions from the DDC high mass cut-off (HMCO) was 35 V. With 35 V DDC for 100 ms applied before charge reduction, the mass error decreased to +1.1% and the effective resolution of the 4+ charge state increased to 200. Using 40% IR for 100 ms (total fluence of 326mJ/mm²) before charge reduction (Fig.2c), the mass error further decreased to +0.5% with an effective resolution of 308.

Fig. 2.

Low charge states of GroEL generated via ion/ion reactions with PFO subjected to 100 ms of (a) no activation, (b) 35 V DDC before the ion/ion reaction, (c) 40% IR before the ion/ion reaction, (d) 35 V DDC after the ion/ion reaction, and (e) 30% IR after the ion/ion reaction. A zoomed-in portion highlighting the 4+ charge state is shown to the right of each spectrum. Spectra were collected using a frequency scan from 40 to 15 kHz over 500 ms (scan rate of 762,838 m/z s⁻¹) with ions ejected at q = 0.5 and calibrated using the charge states in (e) with a mass of 801 kDa. Dashed lines indicate the expected m/z for the given charge states.

Fig. 3.

Spectra of initial charge states of GroEL with 100 ms of (a) no IR activation, (b) 20% IR activation, and (c) 40% IR activation. Insets of (a) and (b) show zoomed portions of the spectra. Spectra were collected by scanning the trapping frequency from 300 to 45 kHz over 2 s (scan rate of 24,034 m/z s⁻¹) with ions ejected at q = 0.5 and calibrated using the spectrum in (c).

Whereas activation prior to charge reduction (Fig. 2b–c) simply highlights the ability of IR activation to more effectively overcome collisional cooling and drive off more solvated molecules, given our current DIT electronics, activation following charge reduction (Fig. 2d–e) illustrates an inherent weakness in DDC activation. To apply 35 V DDC to the product ions following charge reduction without losing all of the low charge states, the trapping square wave frequency was reduced to 23 kHz during DDC activation. The LMCO from the trapping frequency combined with the HMCO from the DDC effectively isolated the 4+ to 6+ charge states (Fig. 2d); however, the mass error did not decrease (+2.1%) and the effective resolution of the 4+ charge state actually decreased to 122 – likely because of the decrease in signal. A model describing the temperature change due to DDC can explain the apparent inability of DDC to drive off weakly bound adducts in this case. Using a diatomic background gas in a pure quadrupolar trapping field, the dipolar field will increase the ions’ temperature by: [11]

$$\Delta T_K=\frac{2m_g}{5k_{b}m}{K}_i$$ (1)

where m_g is the mass of the background gas, k_b is the Boltzmann constant, m is the mass of the ion, and K_i is the average ion kinetic energy. Following the derivation in Ref. [11], but using the well depth potential of a square wave driven 3D ion trap, D = 0.205qV_RF, the effective potential and effective field across the end cap electrodes are:

$$V(Z)=0.205q_Z{V}_{RF}{\left(\frac{Z}{Z_0}\right)}^2$$ (2)

$$E(Z)=-\frac{d}{dZ}V(Z)=-0.410q_Z{V}_{RF}\frac{Z}{Z_0^2}$$ (3)

where Z is the axial displacement of the ion of interest, q_Z is the Mathieu q parameter of the ion of interest, V_RF is the trapping square wave voltage, and Z₀ is the maximum axial displacement. The equilibrium axial displacement of the ion with an applied dipolar field will satisfy E_DDC = −E(Z_e) where Z_e is the equilibrium axial displacement and E_DDC is the field due to the applied dipolar DC voltage. The applied dipolar voltage is then V_DDC = 2E_DDCZ₀, and the effective potential at the equilibrium axial displacement can then be written as:

$$V\left(Z_e\right)=\frac{V_{DDC}^2}{3.28q_Z{V}_{RF}}$$ (4)

where V_DDC is the applied DDC voltage. At equilibrium displacement, the average kinetic energy is K_i = zeV(Z_e) where z is the charge of the ion of interest and e is the fundamental charge. Making this substitution into Equation (1) gives:

$$\Delta T_K=\frac{2m_g}{5m{k}_b}\frac{ze}{3.28q_Z}\frac{V_{DDC}^2}{V_{RF}}=\frac{m_g{\Omega }^2{r}_0^2}{32.8k_b}{\left(\frac{V_{DDC}}{V_{RF}}\right)}^2$$ (5)

where Ω is the trapping frequency, and r₀ is the inner trap radius. Note that for a sine wave driven 3D ion trap, the constant in the denominator of Equation (5) would be 20 due to its well depth potential being D = 0.125qV_RF. In digital trap operation, trapping frequency rather than voltage is variable and determines the temperature increase. Thus, although 35 V DDC was used for both experiments in Fig. 2b and d, the trapping frequency was ~3.9 times greater when activating the precursor ion (Fig. 2b) than when activating the product ions (Fig. 2d), leading to a ~15 fold greater temperature increase in the former case relative to the latter heating due to DDC. Using our experimental parameters (m_g = 28 Da, $r_0^2=0.707\text{cm}$), pre-ion/ion DDC activation provided ΔT = 12.6 K (Ω = 90 kHz) and post-ion/ion DDC activation provided ΔT = 0.8 K (Ω = 23 kHz).

Unlike DDC activation, IR activation of the products (Fig. 2e) appears to have a similar effect as activation of the precursor ions. Applying 30% IR activation (total fluence of 244 mJ/mm²) for 100 ms after charge reduction results in a mass error of 0.1% and 4+ peak resolution of 417. In this case the temperature increase is related to how well the ion cloud interacts with the laser beam. For a well-focused laser beam that is narrower than the ion cloud, a tighter ion cloud leads to greater spatial overlap. Digital operation provides better trapping of high m/z ions than traditional sine wave trapping because the trapping frequency can be easily reduced to move high m/z ions to high q values [24]. The size of an ion cloud can be predicted by its “spring constant” given by: [26,27].

$${\text{Z}}_{\max }=\sqrt{\frac{2E}{{\kappa }_Z}}=\sqrt{\frac{2E}{m{\omega }_z^2}}$$ (6)

where Z_max is the maximum Z-dimension ion displacement, E is the ion kinetic energy, and ω is the ion secular frequency. Assuming that all ions are thermalized to room temperature (~0.04 eV), the average precursor charge state of 64+ trapped with a 90 kHz square wave would have a maximum axial displacement of 24 μm (q_z = 0.45, ω_z = 20.4 kHz). The 4+ charge state trapped with 40 kHz would have a maximum axial displacement of 170 μm (q_z = 0.14, ω_z = 2.9 kHz). Both displacements are expected to be less than the beam radius based on the hole (0.5 mm radius) that the beam passes through. Fig. 2c and e shows that IR activation can be equally effective at lower trapping frequencies, unlike DDC activation, because ions can still be focused to the trap center even with very high m/z ratios and very low charges. The lower power required to desolvate the low charge states vs. the initial charge states may be related to the fact that the low charge states were activated later in the scan function (following the ion/ion reaction) than the initial charge states, and therefore more background was pumped out prior to activation.

3.2. IR fragmentation

Coupling charge reduction ion/ion reactions with an activation approach allows the study of fragmentation patterns over a wide range of different charge states. To observe efficient fragmentation beyond the loss of weakly bound adducts, as noted above, it was necessary to increase the time for excess background gas pump out from 500 ms to 2 s, which reduced the pressure to approximately 0.4 mTorr. Increasing the pump out time beyond 2 s showed no observable difference in complex desolvation and fragmentation. Fig. 3 illustrates desolvation and fragmentation of native GroEL charge states centered at ~64+ (Fig. 3a) using 20% IR (Figs. 3b) and 40% IR (Fig. 3c), respectively (Note that Fig. 2 suggests that with 500 ms pump out prior to IR activation 40% IR power only desolvated the complex without fragmentation of the complex, whereas Fig. 3 shows that 40% IR power led to extensive dissociation of the complex following 2 s of pump out time.). The m/z scale for the spectra in Figs. 3–5 was calibrated using the charge states of Fig. 3c assuming an intact GroEL mass of 801 kDa, a monomer mass of 57,214 Da, and a tridecamer mass of 743,786 Da. IR desolvation decreased the relative mass error from +1.3% to +0.1% shown in Figs. S1a–b. Note that a 1% mass error for the closely spaced native charge states is a much more significant problem than for the widely spaced ion/ion product charge states. The 64+ charge state with no activation measured at m/z 12,678 which is closest to the theoretical m/z of the 63+ charge state (m/z 12,714). After activation, the 64+ charge state measured at m/z 12,528 which is closest to the theoretical m/z of the 64+ charge state (m/z 12,516). This highlights a major benefit for reducing charge via ion/ion reactions to very low charge states. Even with well resolved native charge states, like in Fig.3a, a 1% increase in mass due to adducts could very easily lead to incorrect charge state assignments if based purely on matching peak positions to the nearest theoretical m/z values [28]. To contrast, the wide separation of the ion/ion product charge states in Fig. 2a enables unambiguous charge state assignment even with a mass error due to adducts. Increasing the IR laser power (Fig. S1c) had little to no further impact on the intact GroEL charge states, but rather induces dissociation of the complex. The resulting highly charged monomer and low charge (n–1)mer from IR multi-photon dissociation (IRMPD) is similar to collision-induced dissociation (CID) of natively sprayed protein complexes. [9,10,21,29] IRMPD of GroEL native charge states ranging from 61+ to 68+ resulted in monomer with charges ranging from 29+ to 40+ and tridecamer with charges ranging from 23+ to 37+, as seen in Fig. S2.

Fig. 5.

Spectra of GroEL charge states centered at ~22+ with 100 ms of (a) no IR activation, (b) 20% IR activation, and (c) 40% IR activation. Spectra were collected by scanning the trapping frequency from 300 to 45 kHz over 2 s (scan rate of 24,034 m/z s⁻¹) with ions ejected at q = 0.5 and calibrated using the spectrum in Fig. 3c.

IR activation following a limited amount of charge reduction produced similar results as IR activation of the native charge states, as shown in Fig. 4. Using 20% IR power led to desolvation with the mass error decreasing from +2.0% (Fig. S3a) to +0.3% (Fig. S3b). Again, note that the mass error can lead to incorrect charge state assignment by matching peak positions to theoretical m/z values. Increasing the IR power to 40% further decreased the mass error to +0.1% (Fig. S3c). Additionally, the complex dissociated into monomer with charges ranging from 18+ to 30+ and tridecamer with charges ranging from 19+ to 23+ as seen in Fig. S4.

Fig. 4.

Spectra of GroEL charge states centered at ~42+ with 100 ms of (a) no IR activation, (b) 20% IR activation, and (c) 40% IR activation. Insets of (a) and (b) show zoomed portions of the spectra. Spectra were collected by scanning the trapping frequency from 300 to 45 kHz over 2 s (scan rate of 24,034 m/z s⁻¹) with ions ejected at q = 0.5 and calibrated using the spectrum in Fig. 3c.

When the charge was further reduced to 17+ to 29+, desolvation from 20% IR activation similarly reduced the mass error from +2.0% to +0.3% (seen in Fig.5a–b); however, IRMPD using 40% IR activation did not lead to the same pattern of dissociation as with the charge states above 35+. There is no evidence of monomer in the low m/z of Fig. 5c and the charge reduced tetradecamer charge states are not defined. Previous studies have shown that CID of low charge complexes have stronger inter-unit interactions and collapse to compact structures and/or generate backbone fragments [30–33]. The “blurred” charges states in Fig. 5c could be explained by many overlapping masses due to a variety of backbone cleavages from the intact complex. New signal also appeared below m/z 1500 which could correspond to peptide fragments.

4. Conclusions

The combination of gas-phase proton transfer ion/ion reactions with IR activation in a digital ion trap allows for an exceptionally high degree of flexibility in the study of ions of biologically relevant complexes generated via nESI under native conditions. Gas-phase ion/ion reactions can generate any desired charge state lower than the natively sprayed charge states with high efficiency. Digital operation of an ion trap facilitates trapping and focusing of high m/z ions allowing, under the conditions used here, for the mass analysis of ions slightly in excess of m/z 400,000. Introducing IR activation with pulsed gas allows for efficient desolvation and fragmentation of the wide range of charge states generated with ion/ion reactions. Digital ion trap operation in conjunction with ion/ion reactions and IR activation has been demonstrated here to allow for the removal of loosely bound adducts from complexes as large as 800 kDa and as low in charge as +2. This combination of technologies therefore shows promise for applications in native MS and native tandem MS, particularly in cases in which extensive peak broadening and charge state overlap arising from extensive salt adduction is observed in the initial native mass spectrum.