While there have been many excellent studies in the metabolomics discipline conducted using direct infusion mass spectrometry (Maleki et al., 2018; Dou et al., 2023; Sun et al., 2023; Wolthuis et al., 2023) and flow injection mass spectrometry (Zang et al., 2018), chromatographic separations are commonly hyphenated to mass spectrometry. This coupling adds a further dimension of separation, beyond mass spectrometry itself, which in essence is a separation.

The main aim of applying chromatographic separations up-front of a mass spectral analysis is to reduce the complexity of highly complex biological samples, in an effort to ameliorate the effects of dynamic range (Want et al., 2005) and ion suppression (Furey et al., 2013). Furthermore, chromatography provides a very valuable concentration of each analyte as it leaves the column as a peak, substantially improving its signal-to-noise ratio and making it more likely to be detected and quantified, as opposed to direct infusion mass spectrometry.

One of the main limitations of chromatography is its lack of reproducibility and its heavy reliance on a few common stationary phases. Reversed-phase chromatography, for example, is now the most well-developed and ubiquitous separation medium in analytical chromatography, and rather than use a different column, it is common to use ion-pairing reagents to adapt the chemistry of an analyte to a reversed-phase system (Gong, 2015). Unfortunately, the vast majority of ion-pairing reagents are not mass spectrometry compatible, comprising (as many do) inorganic salts or highly ionizable organic compounds in millimolar concentrations that overwhelm the detection of less abundant endogenous compounds in the system being studied. Various powerful metabolomics methods exploit these properties to perform more targeted analyses, but for broad-based untargeted metabolomics, analysts are typically limited to RP or HILIC with volatile organic acids, bases and salts used as separation systems. Thus, a major limitation of liquid chromatography, as applied to mass spectrometry is that the separation gradient is responsible for much of the selectivity of the system, which can lead to situations where global separations are poorly suited to the separation of closely related compounds.

There are four major types of ion mobility spectrometry (Figure 1). The most common and oldest ion mobility analysers consist of a drift tube containing an inert gas, typically dubbed ‘drift tube ion mobility spectrometry or DTIMS’ (Kirk et al., 2019), as described above. Travelling wave ion mobility spectrometry (TWIMS) is a more recent modification of the traditional drift tube on similar hardware, with the addition of an electrostatic pulse waveform that propagates along the tube, allowing nonlinear resolution of ions at the expense of ion heating (Shvartsburg and Smith, 2008). Ion heating is one of the more significant limitations of TWIMS as ion temperatures can reach 551–774 K (Merenbloom et al., 2012) and can lead to dissociation in small molecules (Morsa et al., 2011). More recent developments in this technology include structures for lossless ion manipulation (SLIM) and cyclic ion mobility (Giles et al., 2019). The SLIM architecture consists of a sandwich structure, with parallel, mirror image electrode arrays that can be configured for either conventional or travelling wave ion mobility (Ibrahim et al., 2017). Currently, available commercial instruments based on the technology configure the system in a serpentine array, significantly extending the ion mobility path length to a total of 13 m utilizing 44 U-turns. Resolving power varies depending on the configuration of the ion optics and the voltages applied, but resolutions of around 300 are available (May et al., 2021). The cyclic IMS instrument incorporates a doughnut-shaped cyclic ion mobility (cIM) device that allows ions to be separated via a user-designated number of repetitive 98 cm paths. The drawback to the instrument is that faster, lighter ions eventually end up ‘lapping’ slower ones, and to this end the instrument brackets the cIM with ion traps, allowing isolation and selection of pre-and post-separation ions. This allows some unprecedented and complex ion mobility experiments, such as the one described by (Sisley et al., 2020), to be performed.

Another type of ion mobility analyser is high-field asymmetric ion mobility spectrometry (FAIMS). This relies on the use of a changing compensation voltage as a filter across a gas flow counter to the direction of the ions. FAIMS was originally published by Buryakov et al. (1993), and has since found its way into several commercial instruments. It is generally employed as a filter (Canterbury et al., 2008), rather than a separation device and has therefore been sparsely used in liquid chromatography coupled to ion mobility multidimensional separations.

Recently, the Park group introduced the trapped ion mobility spectrometry (TIMS) device (Ridgeway et al., 2016). Unlike the previous systems, in TIMS the ions are held in a trapping device and exposed to a moving column of gas, based typically on a modification of ion funnel technology. This trap and release substantially reduce the dimensions of the ion mobility device. Ions are separated based on their physicochemical properties as in conventional drift tube ion mobility, however, the movement of the gas and ions are reversed. In TIMS, ramping the electric field gradient (E) releases ions in descending order of their mobility (K). For an excellent review of TIMS, see (Ridgeway et al., 2018). Parallel accumulation, serial fragmentation (PASEF) is an experimental methodology available on TIMS instruments that uses the unique capabilities of the ion mobility trapping device to accumulate ions, and then sequentially pulse them into a quadrupole for isolation and fragmentation (Meier et al., 2018). Thus, in a single drift spectrum, fragment patterns for each MSMS spectrum are resolved via ion mobility. This dramatically increases the acquisition rate of data dependent acquisition experiments (Meier et al., 2015). It has recently been modified to support data-independent acquisition (dia-PASEF) for proteomics applications (Demichev et al., 2022), which provides an intriguing possibility for improving metabolomics analysis.

Regardless of the specific type of ion mobility device, there are two main advantages of ion mobility spectrometry. Firstly, it allows the separation of isomeric analytes. As discussed in Section 2.2, one of the major limitations of chromatography is that while an optimal liquid chromatography gradient can separate closely related ions, the generally applicable rapid gradients used with untargeted analysis often preclude the separation of isomeric compounds with closely related structures. While this is an issue for chromatography, modern ion mobility instruments with resolutions greater than 100 can now separate many isomeric ions within milliseconds. By coupling ion mobility to chromatography, multidimensional separation can be performed, allowing unprecedented resolution of complex mixtures without extending separation times or employing complex fractionation strategies.

Similar to the first, the second advantage of ion mobility comes from the modern high-resolution IM instruments. Collisional cross section is a fundamental property of an ion, and it can therefore be calculated with greater accuracy than predicting retention times on columns. While ion mobility devices measure ‘drift times’ rather than CCS directly, all modern instruments can calculate CCS from drift times, and this directly addresses one of the biggest challenges in metabolomics—the unambiguous identification of metabolites.

This latter poses an important challenge for the application of IMS to metabolomics. CCSs are able to be determined from experimental parameters via the Mason-Schamp equation (Siems et al., 2012) in DTIMS instruments, while for other types of IMS, calibration standards with known CCS values (derived from DTIMS experiments) must be used (Dodds and Baker, 2019). In practice, even on DTIMS instruments, a typical use for metabolomics, e.g., single field multiplexed analysis, require calibration standards derived from slower, more complex stepped-field methodologies (Dodds and Baker, 2019). Due to the development of large-scale, experimentally determined libraries (Zheng et al., 2017; Picache et al., 2019), as well as appropriate calibration, CCS features provide an additional criterion for metabolite annotation.

Sensitivity has been the perennial issue with ion mobility mass spectrometry. In standard ion mobility analysis, analytes are introduced to the ion mobility device as a packet, followed by ion separation and concomitant detection in the mass spectrometer. Before the next packet of ions can be introduced to the ion mobility device, the slowest ion in the current packet must pass through and exit the IM device. This means that during ion mobility separation, all ions that would otherwise be analysed are lost. Furthermore, due to imperfect electronics, ions of interest can also be lost during the ion mobility separation itself. There have been many attempts to improve sensitivity, such as improving overall instrument design as outlined in 3.1 (Deng et al., 2016), advances in the construction of the instrument with improved lens designs and electric field generation, and the use of multiplexing. Multiplexing is a technique used to decrease ion packet losses in ion mobility mass spectrometry. Multiple ion packets are pulsed into a single ion mobility device in a pseudo-random order. This creates a patterning effect in the ion mobility analysis that can be deconvoluted by the use of a sliding window algorithm during data analysis. This technique has been reported to reduce sensitivity losses from 99% to 50% (Reinecke et al., 2019).

Current, commercially available ion mobility instruments possess resolving power in the hundreds, compared to the tens of thousands of theoretical plates typical in chromatography (for a fascinating discussion on the relationship between resolution parameters for chromatography and spectrometry see (Rokushika et al., 1985)). Consequently, reproducibility is a key parameter to confidently assign collisional cross sections to analytes. Interlaboratory studies have been performed, demonstrating RSDs of 0.14% and ∼1.5% in TWIMS-based devices (Hernández-Mesa et al., 2020; Righetti et al., 2020) and 0.29% for DTIMS-based devices (Stow et al., 2017), which compares favourably to chromatography retention time RSD of 5%–20% depending on analyte (Madji Hounoum et al., 2015).

A further limitation to ion mobility is that it does not get around the problem of matrix effect, a somewhat complex term in ‘omics disciplines, particularly metabolomics. Typically, the “matrix” corresponds to all of the components of a sample one does not want to analyse, while the “analytes” are the components one wants to detect. While all sample preparation methodologies are biased, a good ‘targeted’ analysis aims to strip out the vast majority of the “matrix” while leaving the “analytes” in place. In untargeted metabolomics one does not have this luxury, and the “matrix” and the analytes are (apart from contaminants, buffers, salts, etc.) mostly the same thing. The interaction between these contaminants, as well as the more abundant analytes within the ion mobility device can result in ion suppression (or ion enhancement).

Ion suppression is commonly associated with electrospray ionization, although matrix effects and suppression of ionization are also found in electron ionization used with gas chromatography (Yarita et al., 2015) and nanoelectrospray ionization (Kourtchev et al., 2020). Ion suppression occurs in the ion source, but matrix effects such as ion-ion interactions and space charging can occur in trapping instruments (Hohenester et al., 2020). Thus, ion mobility separations, occurring as they do, after ionization, provide no reduction on in-source ion suppression. Consequently, suppression resulting from, flow injection or direct infusion mass spectrometry will not be ameliorated by the use of IMS, and could result in additional artifacts in measurement (Levin et al., 2014).