Effects of Nano-Electrospray Ionization Emitter Position on Unintentional In-Source Activation of Peptide and Protein Ions

Abstract

Native ion mobility-mass spectrometry (IM-MS) typically introduces protein ions into the gas phase through nano-electrospray ionization (nESI). Many nESI setups have mobile stages for tuning ion signal and extent of co-solute and salt adduction. However, tuning the position of the emitter capillary in nESI can have unintended downstream consequences for Collision Induced Unfolding or Collision Induced Dissociation (CIU/D) experiments. Here, we show that relatively small variations in nESI emitter position can shift the mid-point (commonly called the “CID50” or “CIU50”) potential of CID breakdown curves and CIU transitions by as much as 8 V on commercial instruments. A spatial “map” of the shift in CID50 for loss of heme from holomyoglobin onto emitter position on a Waters Synapt G2-Si mass spectrometer shows that emitter positions closer to the instrument inlet can result in significantly greater in-source activation, whereas different effects are found on an Agilent 6545XT instrument for the ions studied. A similar effect is observed for CID of singly protonated leucine enkephalin peptide and Shiga toxin 1 subunit B homopentamer on the Synapt G2-Si. In-source activation effects on a Synapt G2-Si are also investigated by examining RMSD between CIU fingerprints acquired at different emitter positions and shifts in CIU50 for structural transitions of bovine serum albumin and NIST monoclonal antibody.

Graphical Abstract

Introduction

Native mass spectrometry (and its variant coupled to ion mobility spectrometry, IM-MS) can be a powerful tool to probe protein structures and interactions. Nano-electrospray ionization (nESI) is widely used for native IM-MS because it can transfer kinetically trapped native-like structures to the gas phase for structure analysis. nESI has been shown to be capable of retaining secondary, tertiary, and quaternary structure in many protein complexes both experimentally^1–4 and computationally.^5–7 By deliberately adding internal energy to native-like ions in the gas phase, collision induced unfolding (CIU) and/or dissociation (CID) can be used to identify differences in biomolecule structure, augmenting the capabilities of native IM-MS.^8–12 These structural differences range from relatively large, such as those in high-order structure or ligand binding sites,^13–17 to much more subtle, such as those due to differences in proteins with nearly identical collision cross section distributions in the absence of deliberate activation.¹⁸ Structural information determined by native IM-MS often agrees very well with results from more traditional biophysical and biochemical techniques.^{3,4,11,15,16,19–29}

Although nESI can kinetically trap biomolecular complexes in native-like structures, it is also known that the chemical and physical properties of the nESI droplet are dynamic.^24,30–35 For example, even with the use of volatile buffer salts such as ammonium acetate, the effective pH of droplets can change by as much as 2 pH units as the droplets evaporate,^36,37 and, over time, nESI may acidify neutral ammonium acetate solutions in positive ion mode nESI.³⁸ Other examples of protein disruption in nESI include redox reactions associated with nESI,^39,40 intrinsic native ECD,⁴¹ annealing of the nascent analyte ion,⁴² electrothermal supercharging,⁴² and nozzle-skimmer dissociation^43–45 (a type of in-source CID).

Collisional activation (i.e., nozzle-skimmer dissociation) has long been used to deliberately activate and fragment nascent biomolecular ions in MS as they enter the instrument, but collisional activation (CID and CIU) inside the instrument is more commonly used in modern instruments due to more precise control of energy deposition and the ability to first isolate ions of interest.¹ In CIU/D, modern instruments typically use an adjustable electric potential (the “injection potential”) to accelerate the analyte ion into a region filled with buffer gas. Collisions between the analyte ion and the gas heat the analyte, causing dissociation and/or unfolding. The fraction of remaining precursor ion can be plotted as a function of injection potential to produce a breakdown curve. A common metric to compare breakdown curves generated by these methods are CID50 and CIU50 values, i.e., the injection potential at which 50% of the precursor ion has undergone a measurable dissociation or change in collision cross section, respectively.^46–48 Because collision gas and pressures, collision cell design, and other properties can differ greatly between instruments and may be user-controlled, CIU50 and CID50 values can be challenging to compare between instruments or experimental conditions. For example, within a single instrument, activation early on in the ion path, e.g., in the StepWave of a Waters Synapt G2-Si instrument, can shift breakdown curves and CIU fingerprints along the injection potential axis.^3,49 This shift can be “additive” in the sense that CIU50 values for successive structural transitions are all shifted to higher voltage by approximately the same amount on this instrument, but it is not generally known whether this is the case for other instruments, where ions may experience very different buffer gas pressure profiles and electric fields.

A handful of previous studies have shown that the position and orientation of the emitter in conventional electrospray ionization and nESI can affect the observed mass spectrum and in-source activation of protein-ligand complexes,^35,50,51 and other studies have investigated the internal energy distribution of ions in ESI.^33,52–56 These studies show that emitter position, emitter orientation, and electrospray potential can individually help desolvate and desalt ions and increase mass accuracy, but some combinations of them can lead to dissociation. These studies also demonstrate that greater desalting can occur when there are larger distances between the emitter opening and the instrument inlet.

Control and reproducibility of ion charge state distributions⁵⁷ and structures using nESI can be a hurdle to obtaining highly consistent CIU and CID data, both over time and between laboratories. nESI emitters, which can vary in both material and physical dimensions, are often pulled to different opening diameters by different methods in different laboratories,^58–60 likely resulting in different initial nESI droplet sizes. Laboratories often further optimize source temperature and emitter position to optimize signal strength and reduce the degree of unwanted co-solute or salt adduction, though the precise position of the emitter is rarely reported in publications and may vary from user to user or even spectrum to spectrum. A small number of reports in the literature examine variability in CID/U data and explore design principles to increase reproducibility. Loo and coworkers attributed variation in mass spectra and ion behavior between users without a clear explanation to different users having a different “touch” for ESI.³⁵ They reported that each user, even on the same day, observed differences in mass spectra of protein and protein complex analytes generated from ESI. In contrast, a collaborative effort between three laboratories showed that an “in-source CIU” modification for Agilent Technologies 6560C IM-MS instruments can reduce CIU fingerprint RMSD values for the same analyte between source hardware to less than 5%.⁶¹

Here, effects of nESI emitter position on in-source ion activation in CID and CIU experiments are systematically studied on two different commercial instruments (a Waters Synapt G2-Si quadrupole–IM–time-of-flight, Q-IM-TOF, instrument and an Agilent 6545XT Q-TOF instrument) for several protein and peptide ions. Based on these results, strategies are provided for use in nESI-(IM)-MS experiments to minimize unintentional in-source activation and increase reproducibility.

Methods

All peptide and protein samples were obtained from Sigma Aldrich (St. Louis, MO, USA) as lyophilized proteins and used without further purification. Equine heart myoglobin (SKU: M1882), leucine enkephalin acetate salt (SKU: L9133), bovine serum albumin (BSA, SKU: 05470), Shiga toxin 1 subunit B (SKU: SML0562), and NIST IgG1κ monoclonal antibody (NIST mAb, SKU: 8671) were reconstituted in water and buffer swapped using Micro Bio-spin 6 columns (Bio-Rad, Hercules, CA) into pH 7.4, 200 mM ammonium acetate solution to generate native protein solutions. Partially denatured myoglobin solution, as described elsewhere,⁵¹ was also prepared to yield a myoglobin sample with a single dominant CID loss channel. For this sample, reconstituted holomyoglobin was buffer swapped into 18 MΩ·cm water and diluted with methanol (Thermo Scientific, Waltham, MA) to yield an 80/20 vol/vol water/methanol solution. All final sample concentrations ranged from 5–15 μM with the exception of a high-concentration native myoglobin sample prepared as 35 μM. The nESI source on both instruments was operated in positive ion mode under static-nESI conditions using 1.0/0.78 mm o.d./i.d. borosilicate glass nESI emitters pulled to ~2-micrometer i.d. openings using a P97 Flaming-Brown micropipette puller (Sutter Instrument, Novato, CA, USA) without further modification. nESI was initiated by applying a 0.9–1.5 kV potential to a platinum wire inserted into the solution inside the emitter without use of an external syringe pump. Throughout the manuscript the term “close” refers to the emitter position in which the tip of the emitter is as close to the entrance of the instrument as possible without breaking the tip of the emitter on the instrument inlet cone (an xyz position of 0 ± 0.2, 0.5 ± 0.5, 0 ± 1 mm relative to the center of the inlet, where x, y, and z represent the vertical [above/below the center of the inlet cone], horizontal along inlet axis, and horizontal perpendicular to inlet axis distances relative to the inlet). “Far” positions varied. A complete description of all positions used in included in Table S1.

To assess variability in nESI emitter dimensions, a field-emission, variable-pressure scanning electron microscope (Thermo Fisher Apreo 2S Lo, Thermo Scientific, Waltham, MA) was used. The i.d. of the pulled emitter openings overall was 2.2 ± 0.6 μm, and the distance from the onset of the taper at the end of the emitter to the emitter opening was 3.3 ± 0.1 mm. Further information on emitter reproducibility and settings is included in the SI (see Appendix S1 and Figure S1).

All (IM-)MS data were analyzed using Unidec,⁶² Python, and CIUSuite 2 (using settings listed in Table S2).⁶³ For CIU experiments, RMSD “average of pairwise root mean square deviation” (RMSD_AofP) is used to refer to the average of two or more experiment-to-experiment pairwise RMSD values and RMSD_PofA is used to refer to the pairwise RMSD of two averaged CIU fingerprints. Further discussion of the differences between these metrics is included in the supporting information (Appendix S2). Data for CIU experiments were collected in quintuplicate at each position with 3 of the 5 trials collected using the self-same emitter without moving it.

Results and Discussion

Choice of samples.

CID can be used to fragment covalent bonds in both peptides (e.g., in bottom-up and middle-down proteomics) and proteins (e.g., in top-down proteomics) as well as noncovalent bonds in native MS. In order to survey effects of nESI emitter position in both CID and CIU, we used the common MS standard peptide leucine enkephalin (amino acid sequence YGGFL) and the proteins equine holomyoglobin (which dissociates by loss of a noncovalently bound heme ligand), Shiga toxin 1 subunit B pentamer (as an example of a homooligomer), bovine serum albumin (BSA, which has multiple accessible CIU transitions), and NIST humanized IgG1κ monoclonal antibody (NIST mAb, as a representative of this class of mAb that is of major current interest in CIU applications).¹⁸ Holomyoglobin, notably, is known to exhibit two competitive CID channels (loss of positively charged heme and loss of neutral heme)³⁹ when prepared by nESI from aqueous ammonium acetate, whereas it dissociates almost exclusively by loss of charged heme when prepared by ESI from 80/20 water/methanol solution. The latter behavior is attributed to partial denaturation of the protein in the water/methanol solution, consistent with its slightly shifted charge state distribution in positive ion mode (see Figure 1). However, the extent of denaturation is not sufficient to lose the heme in solution, and the precise BIRD values obtained for the partially denatured ions in studies by the Williams group indicates that the ions’ structures are still relatively homogeneous.⁶⁴ Furthermore, it is known that the oxidation state of native holomyoglobin and the relative prevalence of the neutral vs. charged heme loss channels upon CID are correlated.³⁹ Loss of heme⁰ from holomyoglobin upon CID has been associated with precursor holomyoglobin containing Fe(II), and loss of a heme¹⁺ is associated with precursor holomyoglobin containing Fe(III). The latter oxidation state can result from oxidation of the nESI solution after many minutes of continuous spray, but it has been shown that other heme-containing proteins can undergo intrinsic native ECD at high concentrations to produce monomers containing Fe(III).⁴¹

Figure 1.

Mass spectra for myoglobin ions from 80/20 water/methanol (a) and aqueous ammonium acetate (b). The primary dissociation channels for each solution condition are shown in (c) water/methanol (loss of heme¹⁺) and (d) ammonium acetate (competitive loss of heme⁰ and heme¹⁺).

Figure 1 shows example mass spectra for CID of holomyoglobin ions prepared by nESI from 80/20 water/methanol (partially denatured, Figure 1a) and from neutral 100 mM aqueous ammonium acetate (native-like, Figure 1b). CID of holomyoglobin from water/methanol is strongly dominated by a single dissociation channel via loss of heme¹⁺ (Figure 1c), while CID of holomyoglobin from ammonium acetate exhibits both this loss channel and loss of heme⁰ (Figure 1d). The single CID channel for holomyglobin from 80/20 water/methanol solution facilitated determination of CID50 values when using different nESI emitter positions in many of the experiments described below.

Assessment of protein CIU and CID repeatability using a single nESI emitter and fixed emitter position.

In both CID and CIU experiments on native-like proteins, it is often necessary to repeat measurements with different nESI emitters, and, as described in the Introduction, the emitter position is often tuned to some extent to maximize signal and remove salt adducts. To assess the contributions of using different emitters and emitter positions to variability in CID and CIU, it was necessary to first characterize variability associated with the case in which a single emitter is used without changing its position. For CIU, we used metrics commonly used in comparing CIU fingerprints. CIU fingerprints are often acquired in triplicate, and the triplicate-average root mean square deviation (RMSD) between each pair of maximum-intensity-normalized fingerprints is reported as a percentage (“average pairwise RMSD”, hereafter RMSD_AofP) to assess variability. Here, CIU fingerprints of NIST mAb²⁵⁺ were acquired in triplicate (Figure 2 a–c) using a single nESI emitter in a fixed position and TWIMS settings of 650 m/s and 30 V. When averaged (Figure 2d), these fingerprints show a very low RMSD_AofP (1.19%, Figure 2e and 2f).

Figure 2.

CIU fingerprints of NIST mAb²⁵⁺ collected in triplicate (a,b,c) using a single nESI emitter in a fixed position. The average of these fingerprints is shown in (d). Average (green bar) and single standard deviation (error bars) in pairwise RMSD of CIU fingerprints for all native charge states of NIST mAb (22–26+) and BSA (12–14+) and for only the most abundant charge state (NIST mAb²⁵⁺ and BSA¹⁶⁺) (e). Average CIU difference plot for the fingerprints shown in (a,b,c) for NIST mAb²⁵⁺ (f).

A similar set of CIU experiments was performed for BSA¹⁶⁺, and a very low RMSD_AofP (0.98%, data not shown) was measured. Additional single-emitter RMSD_AofP values for NIST mAb²²⁻²⁶⁺ and BSA¹⁴⁻¹⁶⁺, averaged over all charge states, are shown in Figure 2f, alongside values for the most abundant charge state of each protein. These results show that CIU fingerprints collected sequentially using a single emitter can be highly repeatable (RMSD_AofP ~4%), especially for high abundance charge states (RMSD_AofP ~3%).

Reproducibility of CID and CIU using fixed vs. uncontrolled nESI emitter position.

CID of singly protonated leucine enkephalin peptide (555 Da) generated by nESI from a solution of pH 7 aqueous ammonium acetate was performed on the Synapt G2-Si using injection potentials between 5 and 35 V. CID of holomyoglobin⁷⁻¹⁰⁺ ions (17,549 Da) generated by nESI from a solution of 80/20 water/methanol (vol/vol) was performed with injection potentials between 5 V and 65 V. Figure 3 shows CID breakdown curves for leucine enkephalin (Figures 3a and 3e) and holomyoglobin^9,10+ in water/methanol (Figures 3b, 3c, 3f, and 3g) in uncontrolled (Figures 3a–c) and fixed (Figures 3e–g) emitter positions across multiple days.

Figure 3.

CID breakdown curves (a-c, e-g) and CIU fingerprint differences (d,h) in uncontrolled (a-d) and fixed (e-h) nESI emitter positions on a Waters Synapt G2-Si are shown for leucine enkephalin (a,e), holomyoglobin⁹⁺ (b,f), holomyoglobin¹⁰⁺ (c,g), and BSA¹⁵⁺ (d,h). Statistics for CID50 are summarized in (i), with error bars representing the standard deviation of CID50 value across all trials. Statistics for CIU50 of BSA¹⁵⁺ are summarized in (j), with error bars representing the standard deviation in CIU50 value across three trials.

In uncontrolled nESI emitter position experiments, the position of the nESI emitter was coarsely tuned by a single user for high precursor signal while minimizing sodium adduction. The CID50 for all three ions (leucine enkephalin and holomyoglobin^9,10+) had a roughly inverse correlation with the signal strength. (Figure S2 shows example data for holomyoglobin⁹⁺ from 80/20 water/methanol solution.) That is, higher precursor signal corresponded to greater in-source activation (lower CID50), and positions with lower signal corresponded to less in-source activation (higher CID50; see Figure S2). In contrast, controlling emitter position by using identical nESI stage micrometer settings between days and emitters significantly reduces variability of the replicate-average CID50 for these ions (Figure 3i). The standard deviation in the CID50 for the fixed emitter position experiments was lower by a factor of ~3–6 for all three analytes as compared to that for the uncontrolled emitter experiments. Thus, as for CIU (see above), repeatability in CID experiments can be very high when using a fixed emitter position. These results show that controlling emitter position can improve reproducibility of CID experiments. In principle, use of controlled emitter position can thus increase confidence in interpreting CID50 differences between proteins as structural or stability differences, especially if emitter position is recorded and reported alongside other experimental parameters.

An analogous set of experiments was used to assess effects of nESI emitter position on CIU of BSA¹⁵⁺ on a Waters Synapt G2-Si using injection potentials from 5–200 V. RMSD values were calculated for uncontrolled emitter positions across multiple days (Figure 3d) and for a fixed emitter position using multiple capillaries (three on the same day; Figure 3h). For BSA¹⁵⁺, the RMSD_AofP for the uncontrolled emitter position experiments (27.91%) is far higher than the accepted standard for reproducibility (< 5%).⁶¹ In sharp contrast, the RMSD_AofP for the fixed emitter position triplicate collected on the same day is 4.30%, which falls within this standard and is clearly higher than that obtained for the triplicate using a single emitter (see above). An effect of similar magnitude of exchanging emitters was observed for all charge states of BSA (14–16+) and NIST mAb (22–26+) that were investigated. The overall RMSD_AofP for both proteins was determined to be 6 ± 3% across all charge states and 4 ± 2% for the most abundant charge states. Because variation in CIU fingerprints acquired using different emitters in the same position can be larger than that using a single emitter, it is therefore recommended to use multiple emitters to accurately survey CIU variability for a fixed emitter position. In contrast to the CID experiments described above, use of uncontrolled emitter position to optimize for high signal and low salt adduction in CIU resulted in a similar amount of variation (as measured by RMSD_AofP) as did exchanging emitters in a fixed position (Figure 3j).

Taken together, the above results show that controlling nESI emitter position can significantly reduce variability in CID50 and reduce CIU RMSD. Noting and reporting emitter position (e.g., the nESI stage micrometer positions and stand-off of the nESI emitter tip from the stage) can thus facilitate reproducibility between laboratories and improve traceability, as well provide crucial information for literature meta-analysis.

Spatial map of extent of in-source activation due to emitter position in CID experiments.

Motivated by the above-described influence of nESI emitter position on variability of CID and CIU data, we investigated whether there was a spatial mapping of this effect onto emitter position and to what extent this effect may vary between different instruments and analytes. Figure 4a shows a spatial map of CID50 onto nESI emitter position for partially denatured holomyoglobin⁹⁺ on the Waters Synapt G2-Si. (An analogous map for holomyoglobin¹⁰⁺ is shown in Figure S3.) Representative CID breakdown curves in three emitter positions (“far”, “below”, and “close”; see Figure 4b) show shifts of the entire breakdown curve along the injection potential axis. On the spatial map, emitter positions for which the CID50 is shifted to lower values indicate greater in-source activation for this ion. As can be seen in the spatial map, emitter positions that cause higher in-source activation are located roughly along the inlet axis out to ~3 mm in front of the inlet and then follow a roughly 90° turn toward the axis of the emitter holder. The most activating of these are positions closest to the inlet. Other, more “peripheral” emitter positions cause less in-source activation.

Figure 4.

(a) A map of the effect of position on CID50 of holomygolobin in water/methanol on a Waters Synapt G2-Si. Blue cubes show the highest CID50 potential required while red cubes show the lowest. (b) CID breakdown curves of holomyoglobin water/methanol at the “close” (red; 0, 0.5, 0.5 mm), “far” (blue; 0, 3, 5 mm), and “below” (black; −3, 1, 1 mm) positions. These positions are shown with arrows in (a). (c) CID breakdown curves of holomyoglobin in ammonium acetate on the Waters Synapt G2-Si at the close (red) and far (blue) positions. (d) CID breakdown curves of holomyoglobin in ammonium acetate on the Agilent 6545XT Q-TOF Mass Spectrometer at a “close” (red; 0.5, 2 mm) and “far” (blue; 2, 5 mm) position. See Methods section for explanation of emitter position coordinates.

Based on this map, CID was also performed at “high” and “low” activation positions for singly-protonated leucine enkephalin (Figure S4a–d, representative mass spectra Figure S5), native Shiga toxin 1 subunit B pentamer (Figure S4e), and native holomyoglobin⁹⁺ (Figure 4c) to determine whether a similar spatial pattern holds for other analytes. Differences of ~3 and 4 V were observed for leucine enkephalin and Shiga toxin 1 subunit B pentamer, respectively, with greater in-source activation at the emitter position close to the inlet (Figure S4). A smaller difference of 0.8 V was observed for native myoglobin⁹⁺ (Figure 4c). Results for all three of these analytes qualitatively agree with the map for partially denatured holomyoglobin⁹⁺.

To ascertain whether a similar spatial dependence applies to a different instrument source, the influence of position on an Agilent 6545XT Q-TOF mass spectrometer was studied using native myoglobin and leucine enkephalin with the nESI emitter located close (~1–3 mm) and far (~4–7 mm) from the instrument inlet. On this instrument, no significant difference in CID50 was observed for leucine enkephalin (Figure S6) or moderate concentration (8 μM) myoglobin (Figure 4d) between the two emitter positions. However, when using a high concentration (35 μM) native myoglobin sample on the Agilent 6545XT mass spectrometer (located on axis with the instrument inlet on the other side of the nESI emitter), use of the far emitter position resulted in greater competition between the heme^0/1+ loss channels upon CID. Almost all CID occurred by loss of heme¹⁺ in the close position (see Figure S7). These results illustrate that effects of nESI emitter position on CID can be instrument dependent. While, for the Waters Synapt G2-Si, measurably different amounts of in-source activation occurred at close vs far positions for all three analytes, only small (average of < 0.5 V) differences were observed for the Agilent 6545XT for a similar distance between emitter positions for two of the three tested analytes.

Intriguingly, a large difference in CID branching ratio (heme⁰ vs. heme¹⁺ loss channels) on the 6545XT was observed for high- (but not low-) concentration holomyoglobin solutions with a large (~7 V) concomitant shift in CID50 (see Figure S7). This effect may be due to electrochemical oxidation in solution³⁹ due to the higher electric fields present in the closer emitter position, or it may be due to redox activity within weakly bound dimers at the high concentration. Brandner et al. recently reported “intrinsic native electron capture dissociation” of (heme-containing) cytochrome c, that only occurs at high (≥ 37.5 μM) concentrations and requires no deliberate collisional activation.⁴¹ They attributed this to inter-monomer electron transfer in cytochrome c dimers somewhere within the instrument source region. Here, it appears that emitter-position-dependent activation differences on the Agilent 6545XT are insufficient to significantly affect CID of either leucine enkephalin or native holomyoglobin monomers but are sufficient to cause observable differences for holomyoglobin at high concentration. CID50 values for these experiments are summarized in Figure S8.

The Waters Synapt G2-Si and the Agilent 6545XT mass spectrometer (see photographs and schematics in Figure S9) have significantly different source designs, which may explain the observed differences in nESI emitter position-dependent activation. One major difference is that, inside the Synapt G2-Si source, ions travel through a relatively high-conductance (~15 mm diameter) ion guide into the “Stepwave”, whereas, for the Agilent 6545XT, they pass through a narrow (0.55 mm diameter, 19 cm long) inlet capillary followed by a supersonic expansion into the 4-Torr Front Funnel. The nESI potential needed for stable spray across all emitter positions on the Agilent 6545XT (1.2–1.5 kV) was slightly higher than that used on the Waters Synapt G2-Si (0.9–1.3 kV). Nitrogen drying gas is used in the Agilent 6545XT source, whereas no drying gas is used in these Waters Synapt G2-Si experiments. The range of emitter-to-inlet distances and the source temperatures (24–31 °C) used in the two sources are similar. Together, these design differences suggest that the above-described differences in CID behavior arise primarily from differences in the sharpness of the transition from atmospheric to low pressure and the amount of time ions spend beyond the source inlet at relatively high pressure in the two instruments.

Variation of protein CIU fingerprints at fixed emitter positions as measured using RMSD.

Above, it was shown that using a fixed emitter position with a single emitter can result in very low variability in CID50, CIU50, and CIU RMSD between replicates. To determine whether different emitter positions also cause different measurable in-source activation in CIU experiments, we first investigated whether RMSD for replicates using a single emitter at a fixed position varied with emitter position. Generating CIU fingerprints with low RMSD is a widely accepted criterion for high reproducibility/repeatability and is necessary to confidently distinguish structures of ions based on differences in their CIU fingerprints.

RMSD for CIU fingerprints acquired at different emitter positions were calculated to assess whether these data meet the < 5% acceptance criterion and quantify their variability. These data were collected in a same-day quintuplicate using BSA and NIST mAb (mass spectra Figures S10 and S11) as described in the Methods section. BSA fingerprints were collected in this way on 3 separate days, and NIST mAb fingerprints were collected on one day (representative data shown in Figures S12–14). Results in this section for BSA are reported as the average across those 3 days. Figure 5 summarizes the measured effects of different capillaries at positions close to and far from the instrument inlet. nESI emitter position effects on the repeatability (using a single emitter) and reproducibility (between different emitters) of CIU fingerprint replicates are measured by the RMSD calculated in CIUSuite2 for BSA (Figure 5a) and NIST mAb (Figure 5b). The average RMSD across 3 days and all 5 fixed-position replicates for BSA is less than 6% (and less than 5% for the highly-abundant 15 and 16+ charge states; see Figure 5a). The average RMSD for NIST mAb exceeds 10% for two of the low-abundance abundant native charge states (22 and 23+) under some conditions, which is outside the acceptance criterion established by Ruotolo and coworkers. However, for high-abundance charge states (25 and 26+) at both close and far positions it is less than 7% overall and for most fingerprints is less than 5% (Figure 5b).

Figure 5.

Statistics for RMSD of quintuplicates of BSA collected on three separate days (a) and quintuplicates of NIST mAb collected on a single day (b). In (a) colored bars represent average, and error bars represent the combination of triplicate and day to day variation. In (b) colored bars represent average, and error bars represent the triplicate variation.

Loo and coworkers observed that results from native MS experiments can to some extent depend on the “touch” of the user,³⁵ including tuning of emitter position and other user-tuned source conditions. Comparison of the RMSD at different emitter positions in the above experiments sheds some light on the origins of this user-dependent “touch” for nESI-IM-MS. Figure 5a shows that variability (as measured by RMSD) for BSA CIU fingerprints collected using a single emitter in each position is higher in the “close” position than in the “far” position. For NIST mAb, CIU fingerprints collected in the close position have a higher RMSD than those collected in the far position when the same emitter is used (Figure 5b). Using a single emitter in a “far” position rather than a “close” position led to more stable spray and a lower (< 2%) RMSD value for these ions in most replicates.

However, whereas single-emitter experiments show a clear influence of emitter position on CIU fingerprint repeatability, assessment of reproducibility using different emitters is more meaningful in characterizing expected variation of CIU over time within a single lab or between laboratories using the same type of instrument. For the ions studied here, exchanging nESI emitters increased the RMSD by a factor of 1–3 with respect to single emitter experiments and largely masked any differences in CIU fingerprint variability (RMSD) between the close and far positions (Figure 5). As a caution, it was observed that 1 of 8 emitters tested had an unusually large RMSD of 8.7%, while the others consistently had RMSDs below 4% (analysis with this trial excluded shown in Figure S15).

Differences in CIU fingerprints collected at different emitter positions as measured by RMSD.

In addition to its uses as a metric of variability for a single analyte under a single set of experimental conditions, RMSD is also used as a metric to assess structural differences of ions prepared under different experimental conditions or differences between analytes with similar structures. Here, we apply RMSD_AofP (average of all pairwise RMSD values between fingerprints in each trial acquired under different experimental conditions) and RMSD_PofA (the pairwise RMSD between the average of all trials for each condition, commonly used to assess structural differences between analytes), to assess the differences between fingerprints of the same charge state collected at a close and a far position. For BSA, the RMSD_PofA and RMSD_AofP between the fingerprints acquired at the close and far positions are roughly twice those for data acquired at fixed positions (Figure 5a), indicating a clear effect of emitter position on CIU behavior for this ion. In contrast, due to the high RMSD between different capillaries in the far position for NIST mAb, it is less clear whether the same effect occurs for this ion (Figure 5b).

Differences in CIU fingerprints at different emitter positions as measured using CIU50.

Instead of RMSD, it might seem that the more intuitive property in CIU experiments to compare to the above CID50 results for CID experiments would be the CIU50. The BSA and NIST mAb data described above were analyzed using CIUSuite2 to identify features and CIU50 of transitions for each individual trial (fit parameters in Table S2). The results of this analysis are summarized in Figure 6 for both BSA (Figure 6a) and NIST mAb (Figure 6b). For both NIST mAb and BSA a small (~2 V average, ~1 V median) shift in CIU50 is observed at a close position. This indicates the close emitter position causes more in-source activation than does the far position, in agreement with the CID results. However, the same-day standard deviation in the CIU50 due to exchanging nESI emitters was found to be 1–2 V for these replicates, and the day-to-day standard deviation was ~2 V. Thus, the shift in CIU50 between the two emitter positions was similar in magnitude to realistic variability of CIU50 caused by exchanging emitters and/or day-to-day instrument drift for the Waters Synapt G2-Si.

Figure 6.

Statistics for CIU50 transitions in far (blue) and close (red) positions for BSA collected on three separate days (a) and NIST mAb collected on a single day (b). In (a) error bars represent day-to-day variation. In (b) error bars represent variation across the triplicate. Insets show box-and-whiskers plots for differences in average CIU50 measured at close and far positions for all charge states and all observed transitions. The unusually high CIU50 difference value in the inset of (a) represents data for one replicate of BSA¹⁴⁺ with very low signal.

One explanation for the apparently greater sensitivity of RMSD to emitter position relative to CIU50 is that the latter are determined by integrating IM-MS data over ranges of drift times corresponding to each CIU feature. This integration may average out differences in the shape of drift time distributions that explicitly increase the RMSD. Thus, RMSD (or other similar metrics that directly account for drift time distribution differences)⁶⁵ may be a more sensitive tool than CIU50 for assessing variability of in-source activation for different experimental conditions when these differences are small. There may also be a kinetic contribution, i.e., energy barriers for the observed structural transitions for these ions and internal energy required to observe them on the experimental timescale may differ.

Conclusions

The data presented here indicate that tuning emitter position between experiments can be a significant source of day-to-day and emitter-to-emitter differences between CID breakdown curves and CIU fingerprints. However, recording and using identical, micrometer-adjusted nESI emitter positions on a Waters Synapt G2-Si instrument can increase repeatability and reproducibility of CIU/D data. Systematically mapping the emitter-position-dependent shift in CID50 values for holomyoglobin provided a straightforward way to assess which emitter positions lead to greater in-source activation, with positions close to the instrument inlet typically exhibiting greater activation (and variability in activation) than far and peripheral positions. In contrast, emitter position was shown to have a very small influence on CID50 for most analytes on an Agilent 6545XT Q-TOF instrument. This difference between the two instruments likely arises from differences in collisional heating and cooling experienced by the ions as they pass into and through the source regions, which have significantly different pressure profiles and gas dynamics. Modeling this ion heating and cooling as a function of source design, nESI potential, and gas dynamics is the subject of future investigation.

For CIU fingerprints on the Waters Synapt G2-Si, RMSD values indicate that use of close vs. far emitter positions can result in measurable differences in BSA CIU fingerprints. CIU50 values, by contrast, were less useful for characterizing emitter position effects due to high variability in these values when exchanging emitters and due to instrument drift from day to day. For NIST mAb, RMSD and CIU50 differences between close and far emitter positions were both similar in magnitude to the those at fixed position between emitters and days. These results illustrate that it can be important to characterize emitter position effects on CIU/D for an analyte of interest before choosing which emitter position to use, and that acquiring replicates at a fixed position as well as recording emitter position can be useful in increasing reproducibility. To support reproducibility and traceability within and between laboratories for native IM-MS experiments, including CID/U experiments, an example table reporting nESI emitter positions and other relevant experimental settings for these data is included in the SI (see Table S1), and an associated table template is available for download on the Prell group Github website (https://github.com/prellgroup).