Voltage Rollercoaster Filtering of Low-Mass Contaminants During Native Protein Analysis

Abstract

Intact protein mass spectrometry (MS) via electrospray-based methods is often degraded by low-mass contaminants, which can suppress the spectral quality of the analyte of interest via space-charge effects. Consequently, selective removal of contaminants by their mobilities would benefit native MS if achieved without additional hardware and before the mass analyzer regions used for selection, analyte readout, or tandem MS. Here, we use the high-pressure multipole within the source of an Orbitrap Tribrid as the foundation for a coarse ion filter. Using this method, we show complete filtration of 2 mM polyethylene glycol (PEG-1000) during native MS of SILu mAb antibody present at a 200× lower concentration. We also show the generality of the process by rescuing 10 μM tetrameric pyruvate kinase from 2 mM PEG-1000, asserting this voltage rollercoaster filtering (VRF) method for use in native MS as an efficient alternative to conventional purification methods.

Graphical ABSTRACT

INTRODUCTION

Contaminants such as detergents and polymers are persistent obstacles in protein mass spectrometry.^1–3 In some cases, filtering contaminants inside the mass spectrometer is feasible. Assuming no m/z overlap, filtering is of course possible via time-of-flight,⁴ ion trap,⁵ or quadrupole selection.⁶ While external sample purification methods such as pulldowns, liquid chromatography, and/or other hardware-centric options like FAIMS are all options available to filter contaminants,^3,7–9 each has trade-offs regarding workflow compatibility and demands for additional hardware. Extremely high concentrations of common contaminants like poly(ethylene glycol) (PEG) can introduce the additional challenge of space-charge limitations for in-instrument filtering options.^10,11 In Orbitrap Tribrids,¹² the C-trap, ion trap, and ion routing multipole are trapping devices, susceptible to space-charge biases toward low-m/z ions.^13,14 Additionally, if the analyte is outside of the isolation range of the instrument’s quadrupole (e.g., m/z 2000 for the Orbitrap Eclipse), there is no commercially available method to filter contaminants inside Orbitrap instruments from proteins and their complexes in native-mode electrospray.

Given our experiences with polymeric contaminants during direct infusion, we sought to address such issues using the early part of the ion path of Orbitrap Tribrid instruments. This family of instruments shares similar hardware: an inlet capillary, a stacked ring ion guide (SRIG), an injection flatapole (MP00 for Tribrids), an interflatapole lens to separate pressure regimes (L0), and a bent flatapole (MP0) (Figure 1, left panel; Figures S1 and S2; Table S1). Collisions with the background gas in the high-pressure multipole largely dissipate ions’ kinetic energy (KE) for improved containment within the ion path. The steady voltage decrease across these optics prevents ion accumulation.

Figure 1.

Scheme of voltage rollercoaster filtering and default voltage profiles. Default parameters transmit contaminants and adducts, leading to possible space-charge issues and low baseline-subtracted signal-to-noise ratios (SNR). Introducing in-source collision-induced dissociation (IS-CID) can knock off adducts but give larger analytes too much kinetic energy to be properly directed by MP0. The VRF method can balance adduct removal with kinetic energy management to maximize clean analyte transmission and, by extension, SNR.

For the purpose of filtering out low-mass contaminants from >100 kDa, native analytes, an alternative source voltage scheme was developed (Figure 1). A coarse filter based on the residual KE of ions postcollisional-cooling was established by lowering the DC voltages of MP00 and L0 below the voltage of MP0. Ions from larger, natively sprayed molecules enter the region with enough KE to reach MP0 despite any energy losses in MP00. In contrast, smaller (≪100 kDa) analytes lose too much KE to enter MP0 and persist within the ion path, leading to a coarse “voltage rollercoaster” filtering (VRF) effect (Figure 1). Consequently, the VRF profile borrows concepts from trapped ion mobility spectrometry (TIMS), specifically the elution phase.^15,16 However, VRF does not trap ions in the gas phase, and the VRF profile is static, unlike TIMS. Furthermore, VRF is demonstrated on an instrument not designed for mobility-based separations. Here, VRF is applied before any trapping device in the instrument and thus minimizes space-charge effects that severely limit analysis of high m/z species. This is accomplished without external hardware augmentations like FAIMS and yet accomplishes a similar filtering effect. We demonstrate how VRF can circumvent spectral dominance of millimolar levels of PEG-1000 to obtain high-quality spectra of an antibody or 240 kDa protein complex.

EXPERIMENTAL SECTION

SILu mAb antibody (Sigma-Aldrich, MSQC4) (henceforth “antibody”) and tetrameric pyruvate kinase (Sigma-Aldrich, #10128155001) were prepared according to standardized procedure¹⁷ for a final concentration of 10 μM in 100 mM ammonium acetate, pH 7.0. Adulterated samples also contained 2 mM PEG-1000 (Sigma-Aldrich, #20242–8) (“adulterated”) to initially suppress all native analyte detection (Figure S3). Samples were sprayed using a Nanospray Flex static source (Thermo Fisher Scientific) and medium-sized borosilicate-coated emitters (Thermo Fisher Scientific, ES387).

Experiments were conducted on an Orbitrap Eclipse (Thermo Fisher Scientific) implementing nonstandard values for voltages via scripts written in Lua without any hardware alterations. For the purpose of tuning VRF parameters (SRIG, MP00, and L0 voltages), the response value was the total area corresponding to the largest analyte peak within the m/z 5000–7000 range. As a result, though the tracked peak may change over the course of tuning, the maximum possible transmission for a single charge state within the search window would be retained. Signal optimizations consisted of MP00 DC ramps at varying SRIG DC voltages, and the L0 DC voltage was maintained at 2 V below that of MP00 to prevent ion accumulation. Note that optics that are not explicitly defined (except for L0 which is always 2 V below any custom MP00 value) will exhibit their standard tune values (Table S1). Also note that while the MP00 and L0 voltage values will be expressed relative to the MP0 set value, in-source collision-induced dissociation (IS-CID) is not an optic and therefore refers to the added magnitude of the voltage drop between the SRIG and MP00. Acquisitions employing default and VRF voltage profiles were taken back-to-back within the same data collection. Spectral acquisitions had an injection time maximum of 100 ms, an AGC target of 3 × 10⁵, a resolving power of 7500 at m/z 200 (unless otherwise specified), 5 microscans per spectrum, and were averages of 10–20 acquisitions. All spectral data were examined in Thermo Xcalibur Qual Browser 4.1 (Thermo Fisher Scientific).

RESULTS AND DISCUSSION

Standard analysis of 10 μM antibody adulterated with 2 mM PEG-1000 showed no detected ions above m/z 1680 (Figure 2a and inset). Application of IS-CID—which increases the magnitude of the voltage drop leading into MP00 (Figure 1)—attenuated adulterant signal to the point where antibody was detectable with a signal-to-noise ratio (SNR) of 37 (Figure 2b, inset), albeit with an unusual charge state distribution.^18,19 Attempting isolation of antibody charge states from PEG via the ion trap did not reveal any discernible antibody peaks (Figure 2c). The combination of IS-CID and ion trap isolation revealed a more standard antibody charge state distribution (Figure 2d), but the low SNR of 17 (for charge state 23+) prevented this method from being viable. These methods were also ineffective for reclaiming tetrameric pyruvate kinase, as the addition of high IS-CID resulted in significant ejection of complex monomers (Figure S4). In other words, heavily adulterated samples did not benefit from conventional in-instrument filtering or contaminant suppression methods given the analyte was beyond the isolation range of the quadrupole.

Figure 2.

Observed PEG space-charge effects on SILu mAb antibody (10 μM antibody, 2 mM PEG-1000). (a) Default voltages; (b) 250 V IS-CID; (c) ion trap isolation (target m/z 7000, width m/z 4000); (d) the combination of 250 V of IS-CID and ion trap isolation (target m/z 7000, width m/z 4000). See Table S2 for additional ion currents.

To filter PEG from the sample above via VRF, optimizations were conducted on the adulterated antibody and showed maximum antibody transmission at an MP00 voltage of −8 V (relative to MP0) and 200 V of IS-CID (Figure 3a). Signals for the antibody in the adulterated sample showed a significant bias toward higher IS-CID as evidenced by the ~51% signal decrease from 200 to 160 V IS-CID at an MP00 setting of −8 V. However, the unadulterated samples showed no such bias (Figure S5). Implementing the optimum VRF parameters on the same sample filtered PEG ions away and allowed for clear, baseline resolution (SNR of 943 for charge state 23+) for five antibody glycoforms (Figure 3b). Therefore, applying VRF to an adulterated antibody allowed for complete filtration of PEG in a way not previously demonstrated in Orbitrap instruments, enabling more productive use of available space-charge capacity downstream.

Figure 3.

Optimization and application of VRF on PEG-adulterated SILu mAb antibody and pyruvate kinase tetramer. Optimizations, adulterated runs, and filtered runs for a given species were all conducted in the same spray. (a,c) Ramp data for antibody (a) and pyruvate kinase tetramer (c) allows for optimal profile-tuning. Some source activation voltages have been withheld to maintain clarity of graphs (refer to Figures S5 and S6 for complete profiles). Error bars are standard deviations (n = 10). (b,d) PEG-adulterated antibody (b) and pyruvate kinase (d) are shown with VRF in effect. (b) Source activation = 200 V, relative MP00 = −8 V, relative L0 = −10 V. (c) Source activation = 200 V, relative MP00 = −26 V, relative L0 = −28 V. See Table S2 for additional ion currents.

Similar to the antibody case using VRF above, adulterated pyruvate kinase had a strong bias toward high source activation, with the maximum mean signal being achieved at 200 V IS-CID and a relative MP00 voltage of −26 V (Figures 3c and S6). Implementing this optimum lens voltage profile resulted in complete filtration of free PEG-1000, but the spectrum had a low SNR of 53 (for charge state 33+) (Figure 3d). This was, in part, due to signal splitting from singly and doubly PEG-bound pyruvate kinase tetramers at ~47 and ~18% relative abundance by peak intensity, respectively. In addition, the signal of monomer ejected from the complex reached approximately 57% relative abundance by peak intensity. Higher activation energy would eject additional PEG but also more monomeric subunits. However, the filtration of PEG and partial preservation of the complex demonstrate that VRF can filter out contaminants while maintaining noncovalent interactions.

The application of VRF was effective on the proteins shown in part due to the large KE difference between contaminant and analyte upon attempted entry into MP0. This difference originates in the added acceleration and deceleration of ions via IS-CID and the VRF profile. Trials without IS-CID had optima in the default voltage profile region (Figures S5 and S6), and these optima migrated toward the alternative profile with increasing IS-CID (Figure S7). Higher IS-CID also seemed to increase the discrepancies in optima among different analyte charge states (Figure S8). However, the contaminant was consistently filtered past an increase of 1–2 V from MP00 to MP0 regardless of IS-CID, likely neutralizing on L0 due to failing the energy requirement. Free PEG does not appreciably transmit under the VRF profile (Figure S9). Therefore, major differences in signal optima from the adulterated versus unadulterated samples in the negative MP00 regime may be attributed to disruption of PEG noncovalently bound to the analyte. The analytes were able to avoid filtration from the ion path through their native (constrained) tertiary/quaternary structures, which prevent excess collisions and the kinetic energy loss that those collisions would cause.^20,21 Additionally, the analytes’ higher charge states and masses led to larger kinetic energy “budgets” via voltage drops and entrainment in gas-expansion regions, respectively. Consequently, ions with an optimum in the default voltage regime—such as from a small size (≪100 kDa) or from a denaturing environment imparting a larger collisional cross section—may achieve limited success with the VRF voltage profile. Early tests on native carbonic anhydrase (30 kDa) showed a strong preference for the default voltage profile (Figure S10). While filtering was still possible due to nonzero transmission in a shallower form of the VRF voltage profile, the analyte transmission was reduced by roughly order of magnitude in the process (Figure S11). Note that “rollercoaster filtering” should be viable on instruments with similar hardware. The Q Exactive UHMR falls under this category, and its quadrupole filters up to m/z 25 000. The VRF approach could filter low-mass contaminants prior to isolation of charge states of high-mass complexes even from adulterated mixtures.²² Thus, the efficacy of this method for large, native analytes over default voltage profiles positions VRF as (1) a useful and lightweight alternative to external purification methods and (2) an important resource for enhancing IS-CID in native MS as shown for unadulterated samples (Figures S5–S6).

CONCLUSIONS

Altering the voltage profile of front-end optics in Orbitrap Tribrid instruments allowed for the reclamation of select samples during native MS from concentrated low-mass adulterants in addition to transmission optimization for both adulterated and unadulterated analytes. While optimization profiles for source parameters were nonuniform between the analytes shown, a required tuning time of several minutes makes this method accessible for direct infusion experiments. Furthermore, as proteins with similar weights, charge profiles, and collisional cross sections are expected to use similar VRF settings, this method should be viable for automated characterization of antibodies. Turning attention to a deeper understanding of theory and a more diverse series of test cases, it may be possible to adopt these principles for applications with tighter filtering requirements in the future.

Supplementary Material

Supplemental Figures

Fig. S1. Hardware schematic of an Orbitrap Tribrid; Fig. S2. Tune tree examples with marked pertinent voltages; Table S1. Nomenclature and default values used for testing; Fig. S3. Spectra of carbonic anhydrase at multiple levels of adulteration; Table S2. Average extracted ion current (EIC) and total ion current (TIC) of presented figures; Fig. S4. Observed PEG space-charge effects on pyruvate kinase; Fig. S5. Ramped optimizations of adulterated and unadulterated SILu MAb antibody; Fig. S6. Ramped optimizations of adulterated and unadulterated tetrameric pyruvate kinase; Figure S7. Favorable MP00 values by native species and IS-CID activation; Figure S8. Extracted ion current for unadulterated, tetrameric pyruvate kinase during voltage optimization ramps; Fig. S9. Ramped optimizations of the adulterant PEG-1000; Fig. S10. Ramped optimizations of adulterated and unadulterated carbonic anhydrase; Fig. S11. Spectra of unadulterated and adulterated but filtered carbonic anhydrase (PDF)