Automated Control of Injection Times for Unattended Acquisition of Multiplexed Individual Ion Mass Spectra

Abstract

Charge detection mass spectrometry (CDMS) provides mass domain spectra of large and highly heterogeneous analytes. Over the past few years, we have multiplexed CDMS on Orbitrap instruments, an approach termed Individual Ion Mass Spectrometry (I²MS). Until now, I²MS required manual adjustment of injection times to collect spectra in the individual ion regime. To increase sample adaptability, enable online separations, and reduce the barrier for entry, we report an automated method for adjusting ion injection times in I²MS for image current detectors like the Orbitrap. Automatic Ion Control (AIC) utilizes the density of signals in the m/z domain to adjust an ensemble of ions down to the individual ion regime in real-time. The AIC technique was applied to both denatured and native proteins yielding high quality data without human intervention directly in the mass domain.

Graphical Abstract

Since their invention mass spectrometers have produced data reporting the mass-to-charge (m/z) ratio and abundance of detected ions. While this classic scheme allows for multiple unique species to be confounded in m/z space, there are strategies to place analytes in the mass domain and, therefore, remove ambiguity. For example, if resolution is sufficiently high as to resolve the 1 Da difference between predominantly ¹³C isotopes in proteins, the corresponding m/z spacing can be used to calculate charge and, therefore, intact mass.^1,2 If isotopic peaks are not resolved, the m/z spacing between resolved charge states can still be used to infer charge.³ However, if the analyte’s charge state distribution is not distinct and resolved, then its mass remains a mystery.⁴ This is true for not only large complexes well into the megadalton regime (for example, virus particles) but also for heterogeneous mixtures where many species overlap in m/z space (for example, advanced biologics or whole cell lysates). To alleviate spectral congestion, separation techniques such as liquid chromatography,^5,6 capillary electrophoresis,^7,8 and ion mobility^9,10 are often used. In many cases analytes are distinctly separated, but these techniques have their limitations depending on the number of proteoforms that coelute. To solve this problem, direct charge assignment on each ion can be determined with charge detection mass spectrometry (CDMS), resulting in mass domain spectra without a deconvolution step.^11–15

In the past, CDMS has only been available on specialized electrostatic linear ion trap (ELIT) instruments simultaneously measuring the m/z and charge (z) of individual ions.¹⁶ Recent innovations have extended the CDMS approach to Orbitrap analyzers, being the first to demonstrate and multiplex this capability with ~1000 ions per acquisition.^11,17 Coined Individual Ion Mass Spectrometry (I²MS), its potential creates the necessity for advanced logic to automatically adjust ion numbers into an I²MS-compatible regime by controlling injection times. What defines these optimal injection conditions are largely dependent on sample type, analyte concentration, instrument configuration, and dynamics at the ion source.

For ELITs, which operate best with just one ion or very few ions, there have been two approaches to control injection and acquisition for CDMS. For cases where an analyte ion has sufficient charge to produce a substantial signal from a single pass through the detector, it is possible to use this signal as a trigger to close the gate and start acquisition. For ions with fewer charges, a less precise method must be used, where the transmission and gate time of the instrument are adjusted in an attempt to detect just one ion. Since the arrival of ions at the ELIT is a random process, under ideal conditions a single ion will be detected 37% of the time, with the remaining 63% of acquisitions events being split equally between zero and multiple ions.¹² Empty acquisitions are obviously not useful, while multiple ion acquisitions may produce imprecise results due to undesirable ion–ion interactions. Recent efforts have attempted to mitigate the effects of multiple ions within the same ELIT acquisition, demonstrating acquisitions with up to 10–15 ions.¹⁸ Gating triggers based on image charge detection was also demonstrated, but throughput was similarly limited.¹⁹

Orbitrap analyzers differ from ELIT’s in their tolerance for operating with larger ion populations. Instead of being limited by ion–ion interactions, Orbitraps operated in I²MS mode are limited primarily by the possible appearance of coincidental ions at the same m/z, which would result in an ambiguity between multiple ions or a single ion carrying the sum of the charge of the multiple ions.²⁰ In other words, the optimal number of individual ions that can be acquired within one acquisition is limited by the number of analyte charge state and isotopic channels that can be populated without overlapping in m/z space. Overpopulating these channels (that is, creating multiple ion events) would lead to an uncertain charge and, therefore, an uncertain mass assignment for that signal. To minimize the possibility for ambiguous charge assignment, it is possible to operate the Orbitrap with short injection times producing sparse spectra with a low probability for coincidental ions. However, intentionally limiting the ion population will increase the number of acquisitions (that is, raise the experiment time) needed to produce a statistically relevant mass domain spectrum.

In dealing with large populations of ions, all commercial ion traps operate with some form of automated charge regulation. Automatic Gain Control (AGC) was originally developed to balance dynamic range and space charge effects by adjusting the ionization time for 3D ion traps operating with electron impact sources.²¹ This concept was later expanded to work with external ion injection.^22,23 At face value AGC may seem beneficial to I²MS, but the total charge contained in an ion trap is just one parameter which effects the probability of observing coincidental ions. For example, a spectrum consisting of numerous 50+ ions would have a lower probability of coincidental ions than a spectrum consisting of 25+ ions and the same total charge, since the latter would consist of twice as many individual ions.

In response to this technological gap, we describe Automatic Ion Control (AIC), a new method for balancing the total number of ions with the probability of observing coincidental ions. Instead of regulating on the basis of total charge, AIC considers the density of ion signals in m/z space, since this has a strong correlation with the probability of coincidental ions. As AIC is provided on instruments with Direct Mass Technology (a Thermo Fisher-commercialized form of I²MS) enabled, this automation lowers the barrier for entry for Orbitrap-based single ion experiments, and AIC opens the door for robust, large-scale experiments with a level of user involvement analogous to standard LC-MS runs today.

METHODS

All samples were subjected to AIC and recorded using I²MS as described previously with a 1 s transient acquisition.^11,28,29 The AIC procedure was embedded into the instrument internal software to directly interact with the mass spectrometer in real time and reduce bias between the AIC procedure and I²MS measurements. Below is a detailed description of our AIC implementation, and additional information on materials and methods can be found in the Supporting Information.

AIC attempts to maximize the number of ions within each spectrum while simultaneously avoiding the occurrence of interfering ions which overlap in m/z, which reduce the precision of charge state assignment.²⁰ The probability of observing coincidental ions increases as the density of signals in m/z space increases. The m/z density is somewhat analogous to LC peak capacity, in that it is representative of the number of signals, the width of an individual signal, and the span of the signals along the X-axis.

On a spectrum-by-spectrum basis, the algorithm measures the m/z density and adjusts the injection time for the subsequent spectrum in order to reach a target ion density. The regulation algorithm uses a sparse spectral assumption, where there are no interfering ions, and therefore the predicted m/z density is proportional to the injection time.

$$\text{target injection time}=\text{current injection time}\times \left(\text{target}m/z\text{density}/\text{actual}m/z\text{density}\right)$$ (1)

Since the probability of coincidental ions is related to spectral resolution, AIC measures density in terms of peak widths instead of simply measuring the number of ions per m/z unit. Since the appearance of ions in real samples is not uniformly distributed across m/z space, AIC also weights closer peaks more than distant peaks. To combat large instabilities within the control loop due to spray inconsistencies, injection times are increased or decreased in fractional steps.

RESULTS AND DISCUSSION

The most trivial implementation of ion control for I²MS would regulate the number of signals observed in a spectrum. This would provide some effectiveness but would be cumbersome since it would not account for the m/z span of the acquisition. A wider span can accommodate a larger number of signals, and therefore the optimal number of signals would need to be scaled with the m/z range. A more effective metric is the density of signals in the m/z domain, since this automatically accounts for the m/z span of the spectrum. However, signal density does not account for spectral resolution. Higher resolution settings will reduce the probability that two adjacent signals interfere with one another. Therefore, the density metric should use units which reflect the actual peak capacity of a spectrum.

$$\text{density}=\text{signal counts}\times \text{average peak width}/m/z\text{span}$$ (2)

This metric effectively represents the probability of observing interfering signals in a spectrum but only for the case where signals are randomly distributed across the m/z domain. Spectra of real samples show signals which tend to cluster together due to (1) ion charge being quantized (that is, z consisting of only integer values, constraining m/z variability) and (2) the many sources of subtle mass variability possible within the analytes: isotopic variability, post-translational modifications, adduction of cations and solvent, and so forth. This clustering can leave large regions of the spectrum completely vacant, especially if the acquisition range is larger than absolutely necessary.

Therefore, effective ion control for I²MS should consider only the densest regions of the spectrum. This is achieved by calculating the spacing between all adjacent peaks (in peak width units), sorting these spacings in ascending order, and then calculating the density of the first half of the spacings.

Using a metric implemented in this fashion, it has been empirically determined that a density of 5% provides an acceptable compromise between maximizing signal counts and minimizing interfering signals. This was done by raising the density until multiple ion signals became prevalent. Increasing the density for the same samples result in a higher injection time, which raises the chance for two ions to overlap in m/z space (Figure S1). Increasing amounts of interfering multiple ion signals result in both charge misassignment and increased overlap of adjacent isotope profile peaks, decreasing resolution and individual ion processing efficiency. Ideally, a single density value should be effective for a wide range of analytes and mass ranges.

Since ion densities are determined in real time, AIC is compatible with samples presented to the mass spectrometer via separations mechanisms which are expected to produce ion fluxes which vary in time. We demonstrate this potential by infusing reduced patient antibodies against SARS-CoV-2 (specifically, patient 1877 from Melani et al.)^24,25 via SampleStream^26,27 with a flow rate of 1 μL/min for I²MS analysis (Figure 1). Over the course of the protein elution profile, AIC adjusts the instrument injection time in correspondence with the varying signal density (Figure 1a). As the ion flux into the instrument increases over the first 750 acquisitions, more isotopic and charge state channels are populated, decreasing the spacing between each ion signal and increasing the spectral density. In contrast, as the ion flux into the instrument decreases after the first 750 acquisitions, fewer isotopic and charge state channels become populated, increasing the spacing between ion signals and subsequently decreasing measured ion density. As AIC acts on this fluctuating ion density over the course of protein elution, the changes in injection time maintain an individual ion regime that maximizes the rate of ion collection while minimizing multiple ion events (Figure 1b). In this example, both antibody light chain and heavy chains are in the individual ion regime despite their overall charge state distributions overlapping in m/z space (dotted lines). AIC optimizes the injection time more predominately for the most intensely detected species, in this case the light chain species. As a result, the heavy chain species signals, which are occupying numerous glycoform channels, are less populated with individual ions when compared to their light chain counterparts. Processing the individual ion data yields a spectrum of antibody component proteoforms that are isotopically resolved in the mass domain (light chains shown in Figure 1c).

Figure 1.

(a) AIC injection time adjustment (light purple trace) and corresponding SampleStream elution total ion current profile (dark purple trace) of reduced patient SARS-CoV-2 antibodies (patient 1877 from Melani et al.).²⁵ The three red ashed lines denote three time points where single spectra were examined in (b) with corresponding acquisition numbers (AN) and injection times (IT) listed for each spectrum. Green dashed lines denote individual ion shelves of the antibody heavy (top) and light (bottom) chains. (c) Final I²MS spectrum of detected light chain in the mass domain composed of 47 000 total ions. Inset shows an enlarged view of the most abundant light chain proteoform distribution.

While elongated elution profiles (30 min in Figure 1) are ideal for maximizing the sampling of a given ion population, shorter elution profiles can give fast and responsive insight on sample composition. Therefore, we applied AIC to shortened SampleStream elution profiles, accomplished by increasing elution flow rate over a range of 60 fold (Figure 2). Elutions were conducted at 1, 5, and 60 μL/min, resulting in full-width half-maxima of 14, 3.6, and 0.32 min, respectively (Figure 2a). In order to maintain a consistent ion spacing density, the injection times decreased while ion flux increased. On the latter half of the elution profile ion flux decreased and ion injection times correctly increased to stay nicely in the individual ion regime (Figure 2b). This automated fluctuation allows for the individual ion intensity shelf to be maintained throughout each experiment (Figure S2). Processing the data yields enolase (ID: P00924) spectra in the mass domain (Figure S3), the qualities of which improve over wider elution profiles from 550 to 120 000 individual ion events collected (Figure 2c). As expected, the number of detected ions that compose the I²MS data scale with number of spectra that are acquired over the elution profile, demonstrating that AIC maintains the target ion density regardless of ion flux.

Figure 2.

(a) Total ion current, (b) AIC-directed injection times, and (c) final I²MS spectra of 100 ng of enolase eluted from a SampleStream device at flow rates of 60 (black), 5 (blue), and 1 (purple) μL/min.

In order to demonstrate the flexibility of AIC, we coupled I²MS to native capillary electrophoresis, optimizing the conditions for elongated elution profiles (Figure 3).⁷ In a mixture of NIST monoclonal antibody, tetrameric pyruvate kinase, zinc-bound carbonic anhydrase, and tetrameric alcohol dehydrogenase, all four components were at least partially resolved (Figure 3a). Through AIC, the injection time was adjusted over the course of each proteins’ elution profiles, which maintained the individual ion intensity shelf for all four components (Figure 3b). While full isotopic distributions of the components were not acquired due to the brevity of the elutions (and thus less favorable ion statistics), all components yielded mass profiles, including notable features such as glycosylation patterns and lower abundance truncated forms (Figure 3c).

Figure 3.

(a) Total ion current (dark purple) and injection time (light purple) of a native protein mixture infused via capillary electrophoresis: NIST monoclonal antibody (mAb), tetrameric pyruvate kinase, zinc-bound carbonic anhydrase, and tetrameric alcohol dehydrogenase. (b) Single-acquisition mass spectra, the acquisition number (AN), and injection time (IT) illustrating individual ion collection for each native component. (c) Final I²MS spectra of the four components, including NIST mAb glycoforms and truncated pyruvate kinase (denoted by the truncated cartoon).

Underestimating the injection time leaves isotopic and charge state channels unfilled, increasing the amount of time required to sufficiently detect and analyze low-abundance species. Overestimating the injection time quickly introduces ion overlap, which leads to multiples of ions existing at the exact same m/z value in each acquisition (Figure S1). These “multiple ion events” can be confused for ions of a corresponding multiple of charge and, therefore, mass. In complex samples, multiple ion events can lead to the degradation of resultant mass spectra by presenting multiples of mass (that is, erroneous “dimers” or “trimers” of the species of interest). The antibody example exhibited complexity in the form of heavy (50 kDa) and light (25 kDa) chains coeluting, leading to two individual ion intensity shelves that overlap. Detecting multiple ion events by eye in this case would not be as simple as looking for signals of double intensity since a multiple ion event could consist of a single heavy chain and a single light chain, which would be lower in intensity than that of a two-heavy-chain event. Therefore, an automated solution such as AIC is vital in reducing human involvement thus lowering the barrier of entry and achieving automated CDMS using Fourier-Transform MS. As ions cannot overlap in m/z space for I²MS, the throughput of such analysis is ultimately limited. As such, I²MS benefits from longer elution profiles, as seen in the isotopic development (and low-abundance proteoform detection) going from 550 ions to 120 000 ions in the enolase example (Figure 2). Furthermore, despite injecting the same amount of sample across all three flow rates, the area underneath the total ion current profiles increased at decreasing flow rates: 2.0E9 arb. at 60 μL/min, 6.3E10 arb. at 5 μL/min, and 2.1E11 arb. at 1 μL/min (as calculated by Thermo Qual Browser). In other words, the sensitivity of the experiment increased with decreasing flow rates, both of which are vital for targeting low-abundance proteoforms. Regardless of flow rate or ionization efficiency, AIC maintains the same density from acquisition to acquisition. However, as demonstrated, higher flow rates result in shorter elution profiles and fewer ion signals being recorded overall.

As the presented iteration of AIC uses ion spacing in calculating density, isotopes of proteins with molecular weights under 50 kDa are severely “over resolved.” In other words, AIC calculates low-mass isotopic spacing as too low of a density value and drives the injection time up as a means to fill in the empty space between the isotopes. With the highly resolved isotope peaks unable to get any closer to each other, multiple ion events begin to appear. This influence is apparent in the native capillary electrophoresis example shown in Figure 3, where the smallest component (carbonic anhydrase) is borderline overpopulated while the larger molecular weight components (NIST mAb, alcohol dehydrogenase, pyruvate kinase) are predominantly populated with individual ion events. This indicates that the definition of ion density slightly overcompensates for highly resolved, low molecular weight species. This problem can be further addressed by utilizing a lower density target for smaller protein species. In any case, AIC was sufficient in obtaining mass profiles for all components tested, regardless of analyte mass.

CONCLUSIONS

AIC establishes a basis for which the broader scientific community can more easily optimize CDMS on Orbitraps for a multitude of applications. By demonstrating versatility on simple and complex mixtures in denatured or native modes, AIC makes I²MS accessible for automated experiments and large cohorts in a manner analogous to how AGC makes ensemble mass spectrometry accessible. Being able to automate the interface between sample introductory methods such as SampleStream and capillary electrophoresis coupled to I²MS translates to an ability to perform proteoform CDMS assays at unprecedented throughput and sensitivity with a substantial decrease in experimental complexity relative to LC-MS. This publication provides an initial benchmark as we begin to expand I²MS analysis to more hyphenated methods including chromatography.