Applications and advancements of FT-ICR-MS for interactome studies

Abstract

The set of all intra- and inter-molecular interactions collectively known as the interactome, is currently an unmet challenge for any analytical method, but if measured, could provide unparalleled insight on molecular function in living systems. Developments and applications of chemical cross-linking and high performance mass spectrometry technologies are beginning to reveal details on how proteins interact in cells and how protein conformations and interactions inside cells change with phenotype or during drug treatment or other perturbations. A major contributor to these advances is FT-ICR-MS technology and its implementation with accurate mass measurements on cross-linked peptide pair precursor and fragment ions to enable improved identification methods. However, these applications place increased demands on mass spectrometer performance in terms of high resolution spectral acquisition rates for on-line MSⁿ experiments. Moreover, FT-ICR-MS also offers unique opportunities to develop and implement parallel ICR cells for multiplexed signal acquisition and the potential to greatly advance accurate mass acquisition rates for interactome studies. This review highlights our efforts to exploit accurate mass FT-ICR-MS technologies with chemical cross-linking and developments being pursued to realize parallel MS array capabilities that will further advance visualization of the interactome.

Part 1 – Applications of FT-ICR for interactome measurements

Introduction

Understanding the structure/function relationship of proteins and their interactions within the context of living systems remains a critical unmet challenge in biological and biomedical research. With the rise of mass spectrometry-based proteomics over the last several decades it has become routine for researchers to apply the technique to obtain valuable information on amino acid sequences, abundances, and post-translational modifications on thousands of proteins originating from complex biological samples. Measurements of protein and selected modification level changes can offer new insight on biological pathways and how these are altered with conditions. However, all proteins that provide biological function necessary to support life have evolved to confer this function within a highly orchestrated, dynamic ensemble of conformations and interactions collectively referred to as the interactome (Sanchez et al. 1999). Changes in the interactome, independent of protein or known modification level changes, can dramatically alter the functional landscape in living systems (Vidal, 2001). The intricate relationship between the interactome and biological function has long served as strong motivation for development and of application of new technologies to enable interactome studies (Titeca et al., 2019). Large-scale interactome data holds unmatched potential for improving understanding life on a molecular level, overcoming diseases, and improving health and lifespan of mankind (Vidal et al., Cell, 2011).

The challenges associated with mapping the interactome have long been recognized (Sanchez et al., 1999). Since all physical protein-protein interactions have only the proximity or closeness of binding partners as a common attribute, establishing technologies that can generally provide this type of global information from cells has remained a major barrier. Genetic-based technologies, such as the yeast two-hybrid (Fields and Song, 1989; Uetz et al., 2000), in vivo FRET (Heim and Tsien, 1996) and computational methods (Huynen et al., 2000; Pazos and Valencia, 2002) were the major early contributors to interactome research. However, adaptation of these genetic and computational approaches to studies directly with animal or human samples or to dynamic measurements that could reveal molecular interaction changes critical to disease pathologies is less straight forward. This results in significant knowledge gaps regarding the interactome, how it contributes to phenotypic differences or responds to perturbations.

Mass spectrometry offers opportunities for interactome studies in a variety of ways. Perhaps the most widely applied mass spectrometry approach for studying protein-protein interactions (PPIs) is affinity purification mass spectrometry (AP-MS), which has been successfully used to identify tens of thousands of PPIs (Huttlin et al., 2017). AP-MS requires tagged proteins or antibodies for target protein purification and interaction information is acquired through lysate intermediate samples, limiting the types of interactome studies amenable to AP-MS. Proximity labeling techniques such as APEX (Rhee et al., 2013) and BioID (Roux et al., 2012) provide unique interactome information including transient and weak interactions that could be difficult to detect with AP-MS. The vast majority of interaction data that now exists in available databases such as IntAct (Hermjakob et al., 2004), BioGrid (Oughtred et al., 2019), and others (Licata et al., 2012; Salwinski et al., 2004; Szklarczyk et al., 2019) has been determined with yeast-two hybrid and affinity purification techniques. Generally, all these aforementioned techniques for studying the interactome provide limited or no structural information on protein conformations and interactions as they exist within cells. Thus, there currently exists a great need to develop and apply additional technologies that can help visualize interactomes from native systems and provide quantitative data to better understand molecular changes.

Chemical cross-linking with mass spectrometry (XL-MS), is an emerging technique for studying the interactome that is uniquely able to identify PPIs as well as provide structural information on interacting proteins. In this way, XL-MS provides critical and complementary information to that generated by the mainstay techniques of structural biology (x-ray crystallography, NMR and Cryo-EM) and all other techniques to measure PPIs on a large scale. The covalent bonds that form between a cross-linker molecule and amino acid residues in proteins that are identified in XL-MS studies provide low resolution structural information in the form of distance restraints between cross-linked atoms (Ihling et al., 2006; Kalkhof et al., 2005). These distance restraints can be used to generate structural models for proteins and protein complexes that were present in the sample during cross-linking reaction. Since cross-linking reactions are compatible with application to intact cells (Chavez et al., 2013; Weisbrod et al., 2013a; Zhang et al., 2008) and tissue samples (Chavez et al., 2018), XL-MS is also uniquely suited for studying protein structures and interactions as they exist in their native states, during drug treatment or other perturbations (Chavez et al., 2019; Chavez et al., 2016b), or in phenotypic comparisons (Chavez et al., 2015). This affords the possibility to capture interactions and conformations that are dependent on physiological conditions where the proper concentrations of substrates, interaction partners and integrity of lipid membranes are maintained.

Identification of cross-linked peptides presents new challenges and unique demands on mass spectrometry performance. The measured precursor mass of a cross-linked peptide pair consists of the sum of the individual cross-linked peptides plus any mass added by the cross-linker molecule. Without information on the individual masses of the cross-linked peptides, this becomes a quadratic problem for normal protein database search algorithms because one must consider every possible binary combination of cross-linkable peptides within a defined precursor mass tolerance Fig. 1A (Chen et al., 2001). The overwhelming number of resulting possibilities from this problem places a practical limit on sensitivity of identification and the number of proteins that can be considered during a database search of XL-MS data (Rinner et al., 2008). Even for a database containing only 18 proteins, a striking difference in the number of candidates in linear (proteome) and cross-linked (interactome) peptide databases exists Fig. 1B. For extension to large numbers of proteins such as an entire proteome, the number of candidate cross-linked peptides rapidly becomes excessive, causing unacceptably high false discovery rates, even with high mass measurement accuracy. One way that has been explored to circumvent this limitation for proteome-wide XL-MS applications involves development and implementation of cleavable cross-linkers (Tang et al., 2005) and reviewed by (Sinz, 2017). Cleavable cross-linkers have one or more low energy labile bonds that can be selectively cleaved during MS analysis, releasing intact cross-linked peptides. This allows for measurement of the individual peptide masses, thus reducing the database search space to that of a typical proteomics experiment. Regardless of the type of cross-linker used, XL-MS is highly reliant on the ability to make mass measurements with high resolution and accuracy. This arises because cross-linked peptides are typically two or more times higher in mass than single linear peptides and often require high resolution accurate mass measurements in MSⁿ analyses as well. This is highlighted in the first community wide comparative XL-MS study involving 32 internationally-diverse research groups, all of which utilized MS instrumentation capable of 60,000 – 120,000 resolving power (at m/z 200 or 400) (Iacobucci et al., 2019).

Figure 1 – Complexity of XL database vs. mass accuracy.

A) The size of a cross-linked peptide pair database increases quadratically with the number of proteins it contains. B) Number of candidate peptides (y-axis) vs. mass tolerance (x-axis) in a linear peptide database (blue line) and a cross-linked peptide database (red line) for a database consisting of 18 proteins. C-E) Stages of MS analysis during a ReACT cycle. C) FT-ICR MS1 measurement of a 4+ charged precursor ion with m/z 924.499. D) FT-ICR MS² of the precursor in C, in which the low energy CID labile PIR bonds have fragmented giving rise to two 1+ charged released peptide ions at m/z 1443.680 (green peptide) and m/z 1500.702 (blue peptide) and the reporter ion at m/z 752.412 (yellow). The sum of the neutral monoisotopic masses from the two peptides and reporter match the precursor mass with a mass error or 1.5 ppm. E) Low resolution ion trap MS³ spectra of the two released peptide ions measured in D used in a linear database search to determine the amino acid sequences.

Protein interaction reporter (PIR) based cross-linkers are a class of cleavable cross-linkers which are peptide based, modular in design and include two labile bonds within the cross-linker structure (Tang et al., 2005). Upon activation in the mass spectrometer, labile bonds in PIR cross-linked peptides fragment to release peptide and reporter ions in MS² spectra. The neutral masses of the cross-linked precursor, released peptide and reporter ions are related through the PIR mass relationship shown in Eqn. 1.

Equation 1 – PIR neutral mass relationship

Application of PIR technology to complex biological samples necessitated the development of new informatics capabilities to utilize Eqn. 1 to analyze the data. Initially, the program X-links was developed and applied to analyze high resolution spectra from successive MS scans acquired alternating between no applied collision energy and applied in-source CID during LC separation of cross-linked peptide samples. X-links was used in post-acquisition searches for solutions to the PIR mass relationship, with a specified relationship mass error tolerance (typically 10 ppm or less) (Anderson et al., 2007). This strategy achieved released peptide mass identification spectral pair (in-source off, in-source CID on) comparisons across the entire chromatogram. Extending this concept to include chromatographic information, the program BLinks was developed to refine relationships to include chromatographic overlap of precursor and released peptide ion peaks (Hoopmann et al., 2010). Development of a novel dual linear ion trap FT-ICR instrument (Weisbrod et al., 2013b) presented the opportunity to implement cross-linked peptide informatics on-line during LC/MSⁿ experiments. This hybrid FT-ICR-mass spectrometer was programmed to identify PIR mass relationships in real time with an algorithm utilizing Eqn. 1 on-the-fly called Real-time Analysis Cross-linking Technology (ReACT) (Weisbrod et al., 2013a). ReACT exploits detection of the mass relationship engineered into PIR cross-linked precursor and released peptides during MS² analysis. If the equation is satisfied within a user defined mass error tolerance (typically 20 ppm) the released peptides are subsequently individually isolated and subjected for MS³ analysis to determine their amino acid sequences, cross-linked site, and protein of origin Fig. 1C-E. Analysis of peptide sequence fragment ions would also benefit from high resolution/accurate mass measurements; however due to the relatively high time cost, the current implementation of ReACT relies on low resolution MS³ scans performed in the ion trap Fig. 1E. High resolution MS² spectra generated by ReACT can also be utilized to construct spectral libraries enabling identification of cross-linked peptides via spectral library searching (Schweppe et al., 2016a). Spectral libraries of PIR cross-linked peptide pairs can also be used for cross-platform identification, allowing for MS² libraries generated by FT-ICR-MS to be utilized to assign MS² spectra generated on orbitrap instruments (Schweppe et al., 2016a). Identification of the PIR mass relationship and cross-linked peptide pair sequences can also be assigned directly from MS² spectra, provided sufficient signal of PIR specific fragment ions and released peptide backbone y and b-type ions are detected (Mohr et al., 2018). ReACT has been successfully applied to acquire large scale XL-MS data from a diverse set of biological systems including bacteria (Weisbrod et al., 2013a) (Navare et al., 2015) (Wu et al., 2016), viruses (Alexander et al., 2017; Chavez et al., 2012; DeBlasio et al., 2016), human cell lines (Chavez et al., 2013) and animal tissues (Chavez et al., 2018). Currently there are over 60,000 non-redundant PIR cross-linked peptide pairs identified from 12 different species in the XLinkDB online database for XL-MS data (Keller et al., 2019; Schweppe et al., 2016b; Zheng et al., 2013). Of these, 40 % or 24,436 cross-linked peptides define 12,200 protein-protein interactions. Current estimates from large-scale Co-IP studies indicate there are likely more than 120,000 interactions among approximately 15,000 proteins in human cells (Huttlin et al., 2020). Therefore, considerable opportunities for improved interactome studies exist with in vivo cross-linking approaches that will require advanced mass spectrometry technologies.

Quantitative cross-linking measurements

Beyond identification of cross-linked peptide pairs with XL-MS, quantitative measurement of relative cross-linked peptide pair abundance levels with qXL-MS can provide unique information on changing protein conformations and interactions inside cells or living systems resultant from external stimuli such as drug treatment, or differing biological states (Chavez et al., 2019; Chavez et al., 2016b; Chavez et al., 2015). By incorporation of stable isotope labels into the cross-linker molecule (Zhong et al., 2017) or into proteins metabolically (Chavez et al., 2015), relative differential comparison with qXL-MS can provide insight into changing proteins structures and interactions within living systems. High resolution and accurate mass measurements are essential for detection and comparison of the light and heavy isotopic envelopes detected in the MS1 signal with qXL-MS. Furthermore, high resolution and accurate mass measurements made on the fragment ions generated in MS² of PIR cross-linked peptides can be utilized for targeted quantitative analysis by parallel reaction monitoring (PRM) analysis (Chavez et al., 2016a).

Despite the progress in adaptation of XL-MS to large-scale studies, current proteome scale cross-linking efforts are only scratching the surface of the complexity of the interactome, which for humans has been estimated to contain approximately 650,000 PPIs (Stumpf et al., 2008). The problem is magnified with consideration of proteoforms, individual molecular forms of proteins with varying PTMs and sequences, which have been estimated on the order of millions per cell type (Aebersold et al., 2018). This is particularly important given that interaction partners for minor protein isoforms can be as different as completely different genes and each proteoform can be functionally unique (Yang et al., 2016). As part of the quest to dig deeper into the interactome, technological advancements in mass spectrometric hardware are needed to allow greater sampling speed. Given the performance of current instrumentation (7T Velos-FT-ICR acquiring at 50,000 RP MS1 and MS²) running a ReACT method, less than a quarter of the high charge state precursor ions detectible by MS1 can effectively be sampled with MS² due to duty cycle limitations as illustrated in Fig. 2 A. In addition, only a fraction of the ions selected for MS² produce successful PIR relationships and lead to identified cross-linked peptide pairs Fig. 2A. This is further highlighted by the fact that approximately 20% new cross-linked peptide pairs are identified with each repeated injection of the same sample Fig. 2B. Due to the stochastic sampling process of precursor ions by ReACT, we would expect diminishing returns on new identifications with additional replicate injections, as has been observed with other data-dependent LC-MS methods for peptide identification in complex samples (Tabb et al., 2010). New ICR detector designs and acquisition strategies hold the promise for increased duty cycle and improved sensitivity which will allow for a greater depth of investigation into the interactome. To accompany continued collection of complex XL-MS interactome data, continued development of software tools to aide in the storage, processing and interpretation of the data are needed. To fill this need, XLinkDB was developed and has been continually updated specifically to serve as a public repository and interactive database for XL-MS data (Keller et al., 2019; Schweppe et al., 2016b; Zheng et al., 2013). Currently XLinkDB allows for generic input of XL-MS data and automates production interaction networks, maps to existing structural information in the Protein Data Bank, automates protein modeling and molecular docking, compares with existing interaction information, allows for visualization of quantitative XL-MS data and more. Continued development of XLinkDB to support the ever increasing size and complexity of interactome data sets will be a critical aspect along with advancements in FT-ICR-MS instrumentation for XL-MS experiments going forward.

Figure 2 – Duty cycle of ReACT analysis with identified XL-pairs.

A) Example total ion chromatogram (TIC) during one minute of a ReACT LC-MS analysis. Pie charts above indicate a total of 114 ions with charge 4+ or greater were detected by FT-ICR MS1 during this one-minute period. Only 26 of the 114 were sampled for MS² analysis (green), of which 12 did not satisfy the PIR mass relationship, 9 resulted in successful PIR relationship detection but did not lead to a fully identified cross-linked peptide pair, and 5 resulted in successfully identified cross-linked peptide pairs. B) Bar chart indicating the number of non-redundant cross-linked peptide pairs identified (y-axis) vs. the number of replicate injections (x-axis) for a set of 5-SCX fractions from a cross-linked HeLa cell sample. Each replicate injection leads to an additional ~ 20% new cross-linked peptide pair identifications.

II. FT-ICR-MS Instrumentation advancements for improved interactome studies

FT-ICR-MS is perhaps most uniquely suited to achieve the extreme high end performance in terms of resolving power and mass accuracy for which Marshall and others have most clearly demonstrated (Blair et al., 2017; Hendrickson et al., 2015; Smith et al., 2018). However, FT-ICR-MS also offers opportunities for analyzer growth in new areas to include arrays of detectors that can yield high resolution and mass accuracy spectra acquired in parallel providing improved measurement rates that scale with the number of analyzers. For interactome studies, mass analyzers that can achieve high resolution (100k) and part-per-million mass measurement accuracy during on-line liquid chromatography are essential, as described above. Since cross-linked peptide species typically are observed with mass two times or larger than single linear peptides, the demands on mass analyzer performance with interactome studies are even higher than that normally encountered in proteome studies. Thus, among mass analyzers utilized for interactome studies, FT-ICR and orbitrap instruments are most commonly employed. These instruments achieve high mass measurement accuracy and resolving power through image current detection of ion motion using superconducting magnetic fields or highly precise radial electric fields resultant from complex orbitrap electrode geometries. On the other hand, both require finite signal acquisition times to achieve high resolution accurate mass measurements which limits duty cycle, the number of spectra that can be collected and the number of cross-linked species that can be identified during liquid chromatography (LC) separation. However, unlike orbitrap instruments where the analyzer serves to provide both the detector and restraining forces needed to trap ions and cause oscillatory ion motion, FT-ICR-MS relies on externally-applied superconducting magnetic fields to provide the critical restraining forces that cause oscillatory ion motion that are independent of the ion motion detector. Thus, FT-ICR-MS offers opportunities to explore ICR cell design, advancements in magnetic fields, and other aspects that can be exploited to improve the high resolution spectral acquisition rates. While many of the advances we describe below may also be possible with orbitrap mass analyzers, the radial restoring forces with applied magnetic fields separate from the analyzer electrode design make such exploration of these concepts and developments like the MS array more practical with FT-ICR instrumentation.

II.a. Increased magnetic field strength increases spectral acquisition rates

The magnetic field is the primary determinant of achievable mass accuracy and resolving power for FT-ICR-MS. Cyclotron frequencies and consequently, mass resolving power increase linearly with the strength of the magnetic field as long as collisions with the background gas are minimal (Marshall and Guan, 1996). The relationship of magnetic field strength and cyclotron frequencies can be approximated by $\omega =\frac{q}{m}B$ where $\omega$ is cyclotron frequency, B is magnetic field, m is the mass of an ion and q is charge. The increased resolving power from a stronger magnetic field yields two useful experimental consequences. First, for a given transient length, a stronger magnetic field will produce a spectrum with increased mass resolving power. Second, with chromatographic time scales in mind, a stronger magnetic field enables a target resolution to be achieved with a shorter transient length, improving the ability to sample analytes of interest by increasing the duty cycle.

Beyond mass resolving power improvements, stronger magnetic fields have a myriad of additional benefits. A major benefit is that the number of ions that can be trapped scale quadratically with the magnetic field strength (Marshall and Guan, 1996). Larger ion populations experience stronger space-charge effects, which cause spectral distortions, frequency shifts, peak coalescence, and distortions in isotope distributions. Fortunately, a stronger magnetic field also reduces the effect of these spectral distortions (Senko et al., 1996), enabling larger populations of ions to be effectively sampled during a single scan, improving signal-to-noise.

Improvements in superconducting magnet technologies have been continuous, recently resulting in FT-ICR-MS instruments employing 21Tesla (T) magnets at national labs (Hendrickson et al., 2015; Shaw et al., 2016). The initial demonstration of a 21T FT-ICR showed that the 48+ charge state of intact bovine serum albumin (66 kDa) could be isotopically resolved with a 0.38 second transient (150k resolving power), and over 2 million resolving power could be achieved on the same species with a 12 second transient (Hendrickson et al., 2015). Additionally, on a mixture of peptides, the same instrument could achieve mass accuracy on the order of 100 ppb. These 21T systems have been used primarily to make advancements in problems requiring extremely high mass resolving power, such as top-down proteomics or analysis of complex organic matter (Blair et al., 2017; Smith et al., 2018). Top-down measurements from complex samples, like those used in interactome studies, will also be advanced with improved duty cycle high resolution, accurate MS and MSⁿ measurements.

II.b. Harmonic signal detection increases spectral acquisition rates

Conventional FT-ICR cell designs have consisted of single pairs of detection, excitation and trapping electrodes, with a variety of applied geometries and sizes (Marshall and Hendrickson, 2002; Marshall et al., 1998). Image current detection of ion motion with two electrodes produces oscillating time-domain sinusoidal signals with frequencies matching the ion cyclotron frequencies. These are transformed to the frequency domain and an applied calibration equation is used to yield measured spectra. Increasing the number of detection electrode pairs to 2, 3, 4…n results in detection of harmonic sinusoidal signals that are n-fold higher in frequency as compared with the fundamental ion cyclotron frequency (Grosshans and Marshall, 1991; Limbach et al., 1993; Nikolaev et al., 1990). This also results in acquisition of spectra that are n-fold higher in resolving power as compared with fundamental frequencies acquired over the same acquisition period. Conversely, harmonic signal detection enables an alternative method to increase scan rate n-fold as compared to fundamental frequency detection with equivalent resolving power as shown in Fig. 3A. Importantly, this improvement is achieved independently of advances possible with either increased magnetic fields or ICR cell arrays (Figs. 3B, 3C) or through use of absorption mode phase correction (Fig. 3D and discussed below). Additionally, harmonic signal detection can reduce required signal acquisition time with no loss of mass resolving power, or increase mass resolving power at the same scan rate without increasing the magnetic field (Grosshans and Marshall, 1991; Marshall and Hendrickson, 2002; Misharin and Zubarev, 2006; Misharin et al., 2010; Nagornov et al., 2014; Nikolaev et al., 1990; Nikolaev et al., 1996; Pan et al., 1988; Park et al., 2018; Shaw et al., 2018; Vorobyev et al., 2011). For example, an ICR cell with a single pair of detection electrodes can achieve 40,500 mass resolving power at m/z= 400 ion during a 300ms data acquisition period in 7T magnetic field. With 2 pairs of detection electrodes however, 2-times higher resolving power can be achieved during the same data acquisition period or 2-times faster scan rate can be achieved with no loss of mass resolving power. In general, the theoretical resolving power and scan rate linearly increase with the order of the detected harmonic signals if the ICR signal duration is the same. Over the past several decades, ICR cells with multiple detection electrodes for harmonic signal detection have been studied and demonstrated in many research fields. For example, ICR cells with 2^nd harmonic signal detectors were demonstrated for analysis of proteins and crude oil (Cho et al., 2017; Comisarow et al., 1978; Comisarow and Marshall, 1996; Vorobyev et al., 2011). For 2^nd harmonic signal detection, Vorobyev and co-workers connected both detection electrodes generally used for a dipole detector to a single input of the preamplifier, whereas the second preamplifier input was grounded. Cho and co-workers used excitation electrodes to measure 2^nd harmonic signals. The excitation electrodes were used for ion excitation and then switched to detection electrodes to enable second harmonic frequency detection. With the 2^nd harmonic signal detectors, they achieved equivalent resolving power with scan rate two times higher than that with dipole detection of fundamental frequencies. Further increased resolving power and scan rate performance was demonstrated by increasing the number of detection electrodes beyond two pairs (4 plates). 4^th harmonic signals using an ICR cell with 4 pairs of detection electrodes achieved 4-times higher resolving power for the same acquisition period, or increased scan rates for bimolecular sample as compared with a general dipole detector (Nagornov et al., 2014; Shaw et al., 2018). Our efforts using printed circuit board cells also demonstrated 4^th harmonic (Park et al., 2018) and more recently, 6^th harmonic signal acquisition (Park et al., 2020). In recent years, the benefit of a multiple harmonic detector was demonstrated with the combination of online separation techniques including LC (Kim et al., 2019) and gas chromatography (GC) (Thomas et al., 2019). The multiple harmonic detector showed an increase in the number of detected peaks in comparison to conventional LC- and GC-FT-ICR-MS.

Figure 3. Theoretically achieved scan rates as functions of harmonic orders (A), the number of ICR cells in the array (B), magnetic strengths (C) and phase correction modes (D).

The achieved scan rate increases with the order of detected harmonic signals, the number of ICR cells in the array, magnetic field strengths and absorption mode at a given resolving power.

Challenges associated with harmonic signal detection include low harmonic signal amplitude and unwanted detection of other harmonic and fundamental frequencies (Marshall and Hendrickson, 2002). The maximum amplitude of the nth harmonic signal is generally less than the fundamental signal obtained from a regular dipole detector. Excitation of ion motion to larger cyclotron orbits can improve the relative intensities of harmonics signals (Nagornov et al., 2014; Pan et al., 1988; Park et al., 2018). However, increased excitation amplitude can also increase axial motion of ions and increase the rate of dephasing of ion motion (Marshall and Hendrickson, 2002; Nikolaev et al., 1990). To improve harmonic signal magnitude, Misharin and co-workers developed a coaxial multi-electrode cell (O-trap) for detection of 3^rd harmonic signals (Misharin and Zubarev, 2006; Misharin et al., 2010). The O-trap was prepared by splitting the conventional ICR cell into two compartments; one primarily responsible for ion trapping and excitation, and the other one used for ion detection. The compartment for ion detection had internal coaxial electrodes around which the excited ions rotate and produced 3^rd harmonic signals. The coaxial cell achieved five times higher sensitivity than the conventional cell with ion cyclotron motion excited to half the cell radius.

In general, harmonic detection also generates undesired harmonic peaks that can complicate spectral interpretation. (Marshall and Hendrickson, 2002; Nikolaev et al., 1990). Our efforts resulted in development of an ICR cell with parallel dipole and harmonic signal detectors (Park et al., 2018). This configuration enables independent acquisition of fundamental and multiple harmonic signals in parallel from the same ions during the detection event (Park et al., 2018). Parallel signal acquisition with the ICR cell enabled direct comparison of mass spectra from the dipole and multiple harmonic detectors for unambiguous identification of all peaks with multiple harmonic detectors. Thus, using parallel detection of the fundamental frequencies enables unambiguous assignment of all multiple harmonic signals.

II.c. ICR cell arrays increase spectral acquisition rates

Since signal acquisition represents a major portion of the time spent for each FT-ICR mass spectrum that is produced, a significant route for increased spectral acquisition rates could include parallel spectral acquisition. As illustrated in Fig 4., installation of multiple high-resolution mass analyzers within a single superconducting magnet and operated in parallel can reduce spectral acquisition time by a factor equal to the number of mass analyzers because spectra from all cells are recorded at the same instance. Therefore, the acquisition time required per spectrum is decreased n-fold, where n is the number of FT-ICR cells. In pursuit of this concept, we explored two different types of ICR cell arrays, including linear (Park et al., 2016) and orthogonal ICR cell arrays (Park et al., 2017). In the linear and orthogonal arrays, the same value of trapping voltages (independent supply) were applied to each trapping electrodes to trap ions in each cell. After that, a single excitation waveform and amplitude was applied simultaneously to all cells for parallel spectral acquisition.

Figure 4. An ICR cell array to acquire parallel mass spectra with high resolving power.

Analysis workflow (A) for 3-ICR cell array (B); After quench, ions are filled in cell1, cell2, and cell3 by independent control of each trapping electrode. The trapped ions in each cell are excited at the same time. After excitation, ICR signals are simultaneously obtained from each cell by an array preamplifier.

Our initial investigations developed advanced multiple ICR cell arrays with up to 5 cells aligned with the central magnetic field axis and demonstrated up to 5X parallel spectral acquisition capabilities (Park et al., 2016). Ions were sequentially filled in each cell of the linear ICR cell array by independently controlling DC trapping voltages. After filling each ICR cell, all ions in all cells were simultaneously excited and detected to obtain parallel mass spectra with high mass resolving power (FWHM=185,000 for m/z 1183 ions during 4s data acquisition time). With this linear ICR cell array, we demonstrated feasibility to perform multiple stages of high resolution MS analysis simultaneously with the MS array (Park et al., 2016). These efforts employed the 1^st cell for HR MS1, and the 2^nd and 3^rd cells for HR MS2 spectra..

In addition to a linear cell array, we investigated the possibility to extend ICR cell arrays to multiple dimensions, since if possible, this could in principle extend MS array analysis significantly beyond 5-fold and offer increased operational flexibility of an ICR cell array. Our efforts to develop an orthogonal ICR cell array was based on the capability of cross-B field drift of ions (Nagornov et al., 2015; Wronka et al., 1986) by applying small DC voltages on a parallel pair of plates perpendicular to the magnetic field vector. The first or primary ICR cell was positioned on the central magnetic field axis and used for ion trapping during injection of ions from the LTQ. The secondary cell was located off the central axis, but directly adjacent and alongside the first ICR cell. In this orthogonal ICR cell array, ions from the primary cell were transferred to the second ICR cell with cross-B field drift of ions in 1.5 ms. A second population of ions was then injected in the primary cell and ions in both cells were then simultaneously excited and detected to obtain parallel high resolution spectra (60,000 FWHM) without no obvious discrimination in S/N during single data acquisition period. As both linear and orthogonal arrays are in principle compatible with one another, multidimensional ICR cell arrays with dramatically increased spectral acquisition rates may be possible in the future. Acquisition of multiplexed ICR spectra necessitates the development of new software tools for the processing and analysis of the resulting data. These include the ability to digitize and transform multiple transient channels and downstream tools to prepare the spectral data for protein database searching.

II.d. Other advancements

While advancements can be made through hardware improvements such as more powerful magnets and improved cell designs, there are also improvements that can be made through advancements in data acquisition and processing strategies.

1. Absorption mode

An absorption mode mass spectrum offers up to a 2-fold improvement in mass resolving power compared to the corresponding magnitude mode spectrum. Magnitude mode spectra result from direct interpretation of the complex frequency domain, which is expedient but produces lower resolution peaks. Absorption mode spectra are obtained by phase correction of the real and imaginary parts of the frequency domain, which has a numeric solution if cyclotron phase is known (Marshall, 1979). The first implemented solution to this problem was to utilize simultaneous excitation and detection, which allows direct computation of the absorption mode spectrum, but requires hardware modifications (Beu et al., 2004). Moving beyond hardware requirements, Marshall’s group produced software that scans through phase angle corrections to produce a full absorption mode spectrum (Xian et al., 2010), and later released a tool to correct for baseline deviations characteristic to absorption mode spectra (Xian et al., 2012). A purely algorithmic solution called Autophaser was later developed, which generates absorption mode spectra through iterative curve fitting, in addition to being able to correct for various phase correction artifacts (Qi et al., 2012). External data acquisition systems have now been developed using an updated version of the Autophaser algorithm to produce absorption mode spectra in a high-throughput manner concurrent with acquisition (van der Burgt et al., 2019). Recently, a general use external data acquisition system compatible with FT-ICR and orbitrap instruments has been developed by SpectroSwiss to facilitate absorption mode acquisition on a variety of instruments (Kooijman et al., 2019). Absorption mode analysis of mass spectra offers an improvement in mass resolving power additive without any hardware improvements and is applicable to any experiment through post-processing.

2. Improved data acquisition

Duty cycles can be improved by making efficient use of the time during the detection event, particularly on hybrid instruments where ion accumulation is external to the ICR cell. Kooijman et al modified an LTQ-FT with an external data acquisition system that starts recording following excitation and continues through the end of the ion accumulation for the next FT scan (Kooijman et al., 2019). This effectively extends the transient for any scan by the ion fill time for the next scan, allowing for a three-fold improvement in resolution with improved sensitivity without changing the total scan time in a desorption electrospray ionization (DESI) imaging experiment. The resolution enhancement is increased when the detection event is a small fraction of the duty cycle and is diminished when the duty cycle is dominated by the detection event. This approach will greatly benefit any high-resolution method with long ion fill times, where the data acquisition period would otherwise be underutilized.

III. Conclusions and future outlook

If mapped comprehensively, the interactome would reveal the entire set of intra- and inter-protein interactions that exist inside cells and fundamentally change knowledge about details of molecular function essential to support life. Even with the limited depth presently possible, quantitative interactome studies are able to provide new insight on interaction changes in cells with phenotypic and pharmacological comparisons. Cross-linking and mass spectrometry methods are making these new insights possible and in doing so, reveal new ways in which multidimensional accurate mass measurements can uniquely contribute. In turn, this new insight creates greater demand on mass analyzer performance in terms of sensitivity and acquisition speed that is compatible with LC/MSⁿ. The technological improvements in making high resolution mass spectrometry measurements as described above will undoubtedly improve interactome coverage, yielding new insights into the structure/function relationship of proteins and protein complexes. While complete coverage of the interactome may never be feasible advancements towards this goal are prerequisite to increased understanding of biological systems. Our efforts to develop and utilize single analyzer, fundamental frequency FT-ICR mass analysis for interactome studies over the past 5 years have resulted in identification of more than 60,000 non-redundant cross-linked peptide species. The combination of 6^th harmonic, parallel detectors in linear (5 X) and nonlinear (4 X) array formats would yield a multiplicative increase spectral acquisition rates by 120-fold. The increase in the number of acquired spectra will necessitate improved developments in spectral processing and XL-MS software tools that can take advantage of cluster-based computing. Parallel acquisition hardware advancements needed to acquire array spectra were described but improved software capabilities for processing array data for improved interactome studies are still needed. In principle, this advanced array technology could increase the number of cross-linked species that could be identified over a five-year period to millions of cross-linked peptides. Therefore, advancements in FT-ICR-MS array technologies together with ion source and ion storage improvements could in principle enable significant, if not comprehensive interactome mapping with highly complex biological systems.

ACRONYMS

AP-MS

affinity purification mass spectrometry

APEX

Enhanced ascorbate peroxidase

BioID

proximity-dependent biotin identification

CID

collision induced dissociation

Co-IP

co-immunoprecipitation

Cryo-EM

cryo-electron microscopy

DC

direct current

DESI

desorption electrospray ionization

FRET

Förster resonance energy transfer

FT-ICR

Fourier transform ion cyclotron resonance

FWHM

full width half maximum

GC

gas chromatography

LC

liquid chromatography

MS

mass spectrometry

NMR

nuclear magnetic resonance

PIR

protein interaction reporter

PPIs

protein-protein interactions

PRM

parallel reaction monitoring

PTM

post-translational modification

ReACT

Realtime analysis cross-linking technology

RP

resolving power

S/N

signal to noise

T

Tesla

TIC

total ion chromatogram

XL-MS

Chemical cross-linking with mass spectrometry